Discovery of an Mg ii Changing-look Active Galactic Nucleus and Its Implications for a Unification Sequence of Changing-look Active Galactic Nuclei

All the spectra in this work are obtained from the public SDSS DR14 database (Abolfathi et al. 2018), which covers 14,555 deg². A benefit of its ∼20 yr cumulative data is the fact that extensive multi-epoch spectra are quite suitable for investigating the AGN spectral variability. The multi-epoch spectroscopic observations are mainly from three parts: (1) the overlapped survey areas between adjacent plates; (2) dedicated programs, e.g., the Time Domain Spectroscopic Survey (Ruan et al. 2016a) and SDSS reverberation mapping (Shen et al. 2015); (3) re-observed plates due to insufficient signal-to-noise ratio (S/N). The spectral wavelength coverage for SDSS I&II (SDSS III) is 3800–9200 (3600–10400) Å with spectral resolution R ∼ 1850–2200, and the five-band ugriz magnitudes have typical errors of about 0.03 mag in depth to 22.0, 22.2, 22.2, 21.3, 20.5 mag (Abazajian et al. 2009).

Although the optical light curves are not used to select Mg ii CL AGNs in Section 2.3, they are still useful for understanding the origins of the CL behavior. CSS (Drake et al. 2009) repeatedly covered 26,000 deg² on the sky using a 0.7 m Schmidt telescope with a wide field of view of 8.1 deg². The photometric data were unfiltered and calibrated to V-band magnitude, to a depth of ∼20 mag.

Previous observations of CL AGNs indicate that the so-called Mg ii-emitters are likely to be the faint states of Hβ CL AGNs (MacLeod et al. 2016; Yang et al. 2018; MacLeod et al. 2019). In order to search both Mg ii and Hβ CL AGNs, we define the Mg ii-emitters as those with prominent broad Mg ii but no broad Hβ component (FWHM_Hβ < 1000 km s⁻¹), similar to Roig et al. (2014). The advantage of this approach is being able to discover both Mg ii and Hβ CL AGNs based on the multi-epoch spectra when AGNs turn off or on. Compared with the widely used variability-color selection (MacLeod et al. 2016, 2019; Sheng et al. 2017; Yang et al. 2018), our method servers as a tailored approach for searching Mg ii CL AGNs at low luminosity end, as all of the Mg ii-emitters are very faint (see below).

We start with all spectra (4.8 million) in SDSS DR14 database (Abolfathi et al. 2018). Following are the selection criteria for Mg ii-emitter candidates.

• 1.  
  Redshift: 0.4 < z < 0.8, zWarning = 0.
• 2.  
  Class = "QSO" or "GALAXY."
• 3.  
  Mg ii flux > 0, and (FWHM_Hβ < 1000 km s⁻¹ or Hβ flux < 0).
• 4.  
  SN_spe > 2, SN_Hβ > 2 and SN_{Mg ii} > 1.

To include the Mg ii line, Hβ–[O iii] complex of AGNs, and simultaneously avoid the emission lines that are too close to the spectral edges resulting in low S/Ns, Criteria 1 and 2 are applied. In addition, Criterion 3 ensures that the Mg ii (Hβ) is emission line (narrow emission line or absorption line) based on the measurements from SDSS automatic pipeline (Bolton et al. 2012). As shown in left panel of Figure 1, the S/Ns of continuum and lines (Criterion 4) are needed to ensure the spectral qualities of the candidates. We note that all these candidates are very faint with a typical flux density of 10⁻¹⁷ erg cm⁻² s⁻¹ Å⁻¹ at 3000 Å, thus the S/Ns are lower than the typical values of ordinary quasars (e.g., 5 ∼ 10 in Shen et al. 2011). These four criteria yield ∼16,000 Mg ii-emitter candidates.

Zoom In Zoom Out Reset image size

Then we use PyQSOFit¹⁰ (Guo et al. 2018; Shen et al. 2019) to perform the local fit for the Mg ii region ([2700, 2900] Å) with a power-law continuum, iron template and up to three Gaussian profiles to extract the line properties. To exclude the potential Type II AGNs with narrow Mg ii doublets and also alleviate the noise fitting, we select the candidates with

• 1.  
  2000 km s⁻¹ < FWHM_{Mg ii} < 20000 km s⁻¹ and EW_{Mg ii} > 10 Å.

This leaves ∼800 objects (red and gray dots), shown in the middle panel of Figure 1.

Next, we visually inspect each spectrum to confirm that the spectral fitting for Mg ii line and the measurements of Hβ line from SDSS automatic pipeline are reasonable. About half of ∼800 objects, usually located in the lower-right portion of FWHM–EW diagram, were excluded because of the extremely week broad Mg ii line blending with the continuum, which can significantly affect the spectral decomposition. Another ∼50 objects showing significant broad Hβ components¹¹ are also excluded, which are usually with large EW_{Mg ii} located in the upper portion of the FWHM–EW diagram. This process leaves us a sample of 361 (with a detection rate of ∼0.02% in ∼2 million galaxies/quasars) unique Mg ii-emitters,¹² including 52 objects with multi-epoch observations. Their spectra also usually show significant galactic features, e.g., strong absorption lines (e.g., Ca H+K), 4000 Å break and weak power-law continuum.

Finally, we refit the brightest and faintest epochs of these 52 Mg ii-emitters, and find 10 objects with significant Mg ii variability (>3σ) in the right panel of Figure 1. Rejecting seven ordinary sources due to normal broad Mg ii variation without CL behavior, this leaves three Mg ii CL AGNs (see Figures 1–3 and Table 1), i.e., a detection rate of ∼0.001% (3/52 × 0.02%) based on Mg ii-emitters, which is consistent with that of Hβ CL AGNs in Yang et al. (2018). We also discovered new Hβ CL AGNs, which will be shown in a future paper.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Figure 2 shows an unambiguous turn-off Mg ii CL AGN (J1525+2920) at z = 0.449. This object was first selected as a Mg ii-emitter in the bright state by our work, which was targeted as a luminous red galaxy by SDSS. J1525+2920 shows a dramatic change in Mg ii EW, i.e., EW_{Mg ii} = 110 ±26 Å to 0 (or Δf_{Mg ii} = 103 ± 25 erg cm⁻² s⁻¹ at 4σ level), with a factor of 2 continuum variation blueward of rest frame 4000 Å, which rules out the dust-reddening scenario for CL behavior as this model expects a constant line EW_{Mg ii}. Moreover, the accompanied disappearing of helium and iron lines at 3191 and 3581 Å, as typical features of CL AGNs and TDEs (Brown et al. 2016; Yan et al. 2019), further supports that this is a genuine Mg ii CL event rather than a false alarm due to the calibration problem.¹³ The self-consistence of two faint epochs¹⁴ also suggests that the SDSS flux calibration is robust for this object. The residual spectrum (bright-faint) is well fitted by a power-law continuum f_λ = λ^{−1.24±0.05}, which may indicate a possible AGN origin of the varying component.¹⁵ Its rest-fame transition timescale is less than 286 days, which is consistent with other normal Hβ CL AGNs (MacLeod et al. 2016; Yang et al. 2018).

By convolving with the SDSS filters, this source shows the variations in g- and r-band are 0.54 ± 0.02 mag and 0.1 ± 0.01 mag, respectively, which would be missed by conventional variability selections (e.g., Δg > 1 mag, MacLeod et al. 2016; Rumbaugh et al. 2018).

The seasonally averaged CSS light curve in the upper panel of Figure 2, together with these r-band synthetic magnitudes obtained from three spectra and SDSS-r, indicates a weak variability (ΔV_max−min = 0.17 ± 0.05 mag) over 10 yr. This strongly disfavors the TDE scenario, which typically exhibits a rapid raising phase with several magnitudes and a slow decay by ∼t^−5/3 within at most several years (Rees 1988; Evans & Kochanek 1989), as well as some temporal supernova-driven broad emission lines (Simmonds et al. 2016). The slight difference in the photometric systems of SDSS and CSS can be safely ignored for our purposes.

Given the FWHM_{Mg ii} = 12,700 ± 300 km s⁻¹, we estimate the black hole mass of M_BH = 10^8.0±0.1 M_⊙ (1σ statistical error) according to Shen et al. (2011), and hence the averaged Eddington ratio ${\lambda }_{\mathrm{Edd}}={L}_{\mathrm{bol},(\mathrm{bright}+\mathrm{faint})/2}/{L}_{\mathrm{Edd}}\sim 3.3\times {10}^{-3}$, where L_Edd = 1.26 × 10³⁸M_BH/M_⊙ erg s⁻¹. Here we emphasize that the Mg ii line usually does not follow the breathing mode (Shen 2013; Yang et al. 2019); i.e., the line width may not change with continuum variation, and whether there is an intrinsic R–L is still unclear. Thus, the black hole mass based on Mg ii bears a larger uncertainty compared to Hβ.

Figure 3 exhibits two tentative turn-on Mg ii CL AGNs, and both of their bright states are selected as Mg ii-emitters. Together with J1525+2920, we find that all three Mg ii transitions occurred when f_{3000 Å} is below 10⁻¹⁷ erg cm⁻² s⁻¹ Å⁻¹ (or L_{3000 Å} < 10^43.5 erg s⁻¹), which is much fainter than normal Hβ CL AGNs (MacLeod et al. 2016; Yang et al. 2018).

J0948+0050. The Mg ii variability in this object is very significant with Δf_{Mg ii} = 224 ± 18 erg cm⁻² s⁻¹ (>5σ). However, in the faint state, there is still a small remnant of the broad Mg ii. The light curve shows a strong variability of ∼1 mag over a timescale of ∼10 yr.

J2244+0043. The Mg ii variability in this object is relatively weak with Δf_{Mg ii} = 73 ± 23 erg cm⁻² s⁻¹ (>3σ). The EW_{Mg ii} = 18 ± 7 Å is also small, which almost reaches the lower EW boundary of Mg ii-emitters in Figure 1. The residuals of bright and faint epochs show an insignificant broad Hβ line. The light curve indicates that it varies ∼0.5 mag over a timescale of ∼10 yr.

Due to the lack of obvious accompanied transitions (e.g., iron and helium lines) and verification from multi-epoch spectra, together with the concerns above, we classified them as tentative Mg ii CL AGNs. If they keep brightening further, we would expect the appearance of a broad Hβ component.

Recently, Guo et al. (2019) demonstrated that the dramatic changes in broad Hα/Hβ emission in the observationally rare CL quasars are fully consistent with their photoionization model, and the theoretical CL sequence predicted by their model provides natural explanations for the persistence of broad Mg ii in CL quasars defined on Hα/Hβ and the rare population of broad Mg ii-emitters (see their Figure 8).

Here we recovered an observed CL sequence with three real CL AGNs at different temporal evolution stages to confirm their prediction of the photoionization model. In Figure 4, two known multi-epoch Hβ CL AGNs (J141324.27+530526.9, gray lines, z = 0.457 in Dexter & Begelman 2019 and J022556.07+003026.7, green lines, z = 0.504 in MacLeod et al. 2016), together with our Mg ii CL AGN (J1525+2920, red lines, z = 0.449), are selected to construct the observed CL sequence in bright, intermediate, and faint stages. The Mg ii lines share a similar profile for the faintest epoch of intermediate Hβ CL AGN and the brightest epoch of the Mg ii CL AGN, as well as the Mg ii and Balmer lines in between the bright and intermediate CL AGNs. This allows for temporal stages in three different CL AGNs to link together, mimicking the variability evolution in a CL object and compensating for the lack of multi-epoch spectroscopy across the full sequence in a single object. Although, the exact broad line width may be different or slightly affected by non-response effect in Mg ii (Guo et al. 2019), it would be trivial for demonstrating the concept of broad line disappearance.

Zoom In Zoom Out Reset image size

As shown in Figure 4, when the broad Balmer lines almost disappear (e.g., become undetectable), the broad Mg ii emission is still substantial. While the continuum luminosity continues to drop, the broad Mg ii eventually becomes too weak to be detectable. In this sequence, some faint epochs of the intermediate Hβ CL AGN and brightest epoch of Mg ii CL AGN are the so-called Mg ii-emitters. Thus, we speculate that the Mg ii-emitter is more likely to be the transition quasar population, where the quasar continuum and broad Balmer-line flux had recently dropped by a large factor but the broad Mg ii flux is still detectable on top of the stellar continuum. We also notice that the Mg ii line always disappears later than Hα because Mg ii has both less variability and suffers less contamination from the host galaxy. All these features are consistent with the theoretical CL sequence predicated by the photoionization model in Guo et al. (2019).