Discovery of 21 New Changing-look AGNs in the Northern Sky

2.1.1. SDSS Spectroscopy

2.1.2. LAMOST Spectroscopy

CL AGNs showed continuum flux changes in the optical and mid-infrared. Apart from objects with repeat spectroscopy, we carry out additional searches of CL AGNs based on imaging data. We briefly introduce the imaging data as follows.

The SDSS imaging survey scanned the sky in five filters, ugriz (Fukugita et al. 1996), covering 11,663 deg² in SDSS-I/II from 2000 to 2007 (Abazajian et al. 2009) and an additional 3,000 deg² in SDSS-III in 2008.

The Pan-STARRS1 (PS1; Chambers et al. 2016) survey used a 1.8-meter telescope with a 1.4 Gigapixel camera to image the sky in five broadband filters (grizy). The observations cover three-quarters of the sky several times per filter. We use the PS1 magnitudes in the stack catalog, from co-added images made from the multiple exposures. Using a quasar composite spectrum from Vanden Berk et al. (2001) convolved with the PS1 and SDSS g-band filter curves (Equation (2) in Wu et al. 2004), the magnitude difference, δ(g_PS1 − g_SDSS), is between −0.065 mag and 0.008 mag at redshift z < 2.

The DESI Legacy imaging survey (DELS; Dey et al. 2018) commenced imaging surveys, including the DECam Legacy Survey (DECaLS) g-, r-, and z-bands, the g- and r-bands of the Beijing–Arizona Sky Survey (BASS; Zou et al. 2017), and the Mayall z-band Legacy Survey (MzLS). The imaging survey commenced from 2014 and will conclude in 2018, covering 14000 deg² sky. The magnitude difference, δ(g_DELS − g_SDSS), is between −0.053 mag and 0.005 mag at z < 2, using the composite spectrum convolved with the DELS and SDSS g-band filter curves.

The Wide-field Infrared Survey Explorer (WISE; Wright et al. 2010) mapped the all-sky from 2010 January to July in four bands centered at wavelengths of 3.4, 4.6, 12, and 22 μm (W1, W2, W3, and W4). The secondary cryogen survey and Near-Earth Object Wide-field Infrared Survey Explorer (NEOWISE; Mainzer et al. 2011) Post-Cryogenic Mission mapped the sky from 2010 August to 2011 February. The NEOWISE Reactivation Mission (NEOWISE-R; Mainzer et al. 2014) surveys the sky in the W1 and W2 bands from 2013. We use WISE multi-epoch photometry from individual single-exposure images in the W1 and W2 bands.

The Catalina Real-time Transient Survey (CRTS; Drake et al. 2009) uses data from the Catalina Sky Survey, which repeatedly covers 26,000 deg² on the sky. The CRTS photometric data are unfiltered and calibrated to V-band magnitude. We apply a constant offset to the CRTS magnitudes to match the simultaneous g-band PS1 magnitude.

The Palomar Transient Factory (PTF; Law et al. 2009) is a wide-field survey covering approximately 30,000 deg² in the g- and r- bands from 2009 to 2012.

We carried out a systematic investigation in the SDSS DR14 spectral archive, in which 87% of spectra are good quality (zWarning = 0 or 16). Cross-matching the spectra with a radius of 2'' results in 350,609 objects repeatedly observed (≥2 epochs). There were 175,575 repeatedly observed objects classified as "GALAXY" by the SDSS pipeline in at least one epoch spectrum. Among them, 2,023 objects were classified as "GALAXY" in one epoch of the spectrum and "QSO" in another epoch of the spectrum. We visually checked all the spectra of these objects to search for CL AGNs.

Because the definitions of AGN spectral types depend on the strength of broad emission lines, we qualitatively define a detection of a broad emission line if the signal-to-noise ratio (S/N) of the emission line is higher than 5 or as a non-detection if the S/N is lower than 1. Specifically, a weak detection is when the S/N is between 1 and 3, and an intermediate detection is when the S/N is between 3 and 5. Visual checks cannot accurately discern the quantitative changes of emission lines, but they can qualitatively distinguish the cases with dramatic changes. More specifically, with a visual inspection it is easier to distinguish >5σ and <1σ detections than to distinguish 3σ–5σ and 1σ–3σ detections. Therefore, the visual inspection process is inclined to select AGNs that changed between Type 1 and Type 2 (or Type 1.9), with an appearing or disappearing broad Hα (or Hβ) emission line. Nevertheless, a visual check selects AGNs that changed between Type 1 and Type 1.8 with a lower completeness. Objects with distinct companions (within 2'') in the SDSS image were also excluded in the visual check process. From the visual inspection, we selected 9 CL AGNs. Four of the selected CL AGNs were reported previously, including the CL quasar J0159+0033 in LaMassa et al. (2015), J0126−0839 and J2336+0017 in Ruan et al. (2016), and J1011+5442 in Runnoe et al. (2016). We recovered all the changing-look quasars in these three works. The new CL AGNs found in the SDSS archive are J1104+6343, J1118+3203, J1150+3632, J1358+4934, and J1533+0110. We describe the objects rejected by visual inspection in the Appendix. In Section 4.1, we describe the fitting of the Hα and Hβ lines of these objects to quantitatively understand the changes of CL AGNs selected from visual inspection. The ratio of CL AGNs with an appearance or disappearance of broad Balmer emission lines is roughly 0.006%, for objects that were identified as galaxies and repeatedly observed. This ratio may be affected by some issues, for example, the time intervals of the repeat spectra, spectroscopic survey selection bias, spectral quality, and possible selection bias of visual inspection. The selection steps are summarized in Table 1.

We cross-matched the LAMOST spectral archive with the SDSS spectra archive. The selection steps are summarized in Table 2. There were 155,220 objects that were classified as "GALAXY" (75%) or "QSO" (25%) with good spectral quality in the SDSS, or were observed in LAMOST. Comparing the SDSS spectra with the LAMOST spectra, we fit the Balmer emission lines in spectra from the SDSS and the LAMOST with Gaussian profiles. We selected the objects with changes in the emission-line flux density larger than 2 × 10⁻¹⁸ erg s⁻¹ cm⁻² Å⁻¹, which is a conservative criterion avoiding missing CL AGNs at this step. There were 8,181 objects with potential emission-line variation. We visually inspected all these objects, and identified 10 CL AGNs, including 8 turn-on CL AGNs (J0831+3646, J0909+4747, J0937+2602, J1115+0544, J1132+0357, J1447+2833, J1545 +2511, and J1552+2737) and 2 turn-off AGNs (J0849+2747 and J1152+3209). We obtained flux calibrations to match the narrow emission-line flux of the SDSS spectra, assuming a constant narrow emission-line flux within a few tens of years. We describe the objects visually rejected in the Appendix. In this survey, about 0.007% of the galaxies from the LAMOST and SDSS cross-matched sample were proven to be CL AGNs with the appearance of broad Balmer emission lines.

Most of the reported CL AGNs were discovered by repeat spectroscopy. Meanwhile, the optical and mid-infrared flux of CL AGNs varies following the type transition. We conducted additional CL AGN searches based on photometric variability. We searched CL AGN candidates from objects that were spectroscopically identified as "GALAXY" but brightened in later photometric data, and from objects that were spectroscopically identified as "QSO" but significantly dimmed.

3.3.1. Turn-on CL AGNs Selected from Imaging Data

We conducted a survey of turn-off CL AGNs from SDSS quasar catalogs (shown in Table 4). There are 346,464 quasars (Yang et al. 2017) in the SDSS Data Release 7 Quasar catalog (Schneider et al. 2010) and the Data Release 12 Quasar catalog (Pâris et al. 2017). Cross-matching with a radius of 2'', 94% of quasars were detected in the WISE multi-epoch data. There were 6,847 quasars that dimmed (ΔW1 > 0.2 mag) and became bluer (Δ(W1 − W2) < −0.1 mag) in the mid-infrared.

On a timescale of months to years, the continuum variability of quasars is typically 0.2 mag in the optical (e.g., Vanden Berk et al. 2004; Wilhite et al. 2005; MacLeod et al. 2012). Among the quasars with mid-infrared variability, 232 quasars also dimmed significantly in the optical (more than 1 mag in the g-band). We recovered two turn-off CL AGNs, including J0849+2747 and J1011+5442. In 2017 April, we observed one candidate (with the XLT using a 23 slit), J1104+0118, with g-band and W1-band magnitudes dimmed by more than 1.8 mag and 0.6 mag, respectively. The continuum flux of J1104+0118 obviously became much redder than that in the SDSS spectrum, and the broad Mg ii emission line was not detected in the new spectrum. Higher-quality spectra are needed for a more detailed analysis A more complete survey of CL AGNs will require more spectroscopic follow-up.

In total, there are 21 new CL AGNs found in our surveys (summarized in Table 5). Among the new CL AGNs, 5 were found by repeat spectra from the SDSS, 10 were discovered based on repeat spectra from SDSS and LAMOST, and 6 were selected from photometric variability in optical and mid-infrared imaging data, and confirmed by new spectroscopy. The five known CL AGNs that we recovered are also listed in Table 5. Figure 1 shows an example of the light curves (left panels), and spectra (right panels) of a new CL AGN, J0831+3646. The light curves and spectra of the other 20 new CL AGNs are listed in the Appendix. In Figure 1, the spectra separating for 13 years show that there was no Hβ emission in 2002 in the SDSS (black), while the broad Hβ, as well as broad Hα, emission lines emerged in the spectra taken in 2015 in LAMOST (blue). The residual spectrum (gray spectrum at the bottom of the right panel), between the bright epoch and faint epoch spectra, distinctly shows the emerging of broad Hβ and Hα. The left panel in Figure 1 shows the photometric data from several surveys, including SDSS, PS1, DELS, WISE, CRTS, and PTF. The imaging data taken in separated epochs help identify the variability of objects.

Zoom In Zoom Out Reset image size

We fit the spectra of all the CL AGNs with a quasar spectral fitting procedure (QSfit, Calderone et al. 2017), considering the contribution from the AGN continuum, Balmer continuum, host galaxy, blended iron lines, broad emission lines, and narrow emission lines. The spectroscopic measurements are summarized in Table 6. The later epoch spectrum is calibrated to the earlier SDSS spectrum assuming a constant narrow emission line during a few years. Only in 4 out of these 21 objects is the S/N of the faint-state spectrum lower than that of the bright state spectrum. The weaker Balmer emission-line intensity at the faint state is not a result of a worse observational condition.

Table 6.  Spectral Type Transition of CL AGNs

Name	Redshift	λL5100	S/N1	S/N2	LHβ,1	LHβ,2	LHα,1	LHα,2	${\rm{S}}/{{\rm{N}}}_{{\rm{H}}{\beta }_{1}}$	${\rm{S}}/{{\rm{N}}}_{{\rm{H}}{\beta }_{2}}$	${\rm{S}}/{{\rm{N}}}_{{\rm{H}}{\alpha }_{1}}$	${\rm{S}}/{{\rm{N}}}_{{\rm{H}}{\alpha }_{2}}$	E1	E2	Type
		(1042 erg s−1)			(1041 erg s−1)	(1041 erg s−1)	(1041 erg s−1)	(1041 erg s−1)
J0831+3646	0.19501	18.23 ± 0.24	12.3	10.9	−4.05 ± 0.33	2.85 ± 0.39	3.60 ± 1.00	10.19 ± 0.42	−12.1	7.3	3.6	24.2	2	1	on
J0849+2747	0.29854	63.42 ± 1.48	9.3	2.7	8.56 ± 0.39	−2.81 ± 2.03	22.76 ± 1.95	2.49 ± 0.97	22.2	−1.4	11.7	2.6	1	2	off
J0909+4747	0.11694	6.97 ± 0.20	11.0	8.5	−1.88 ± 0.13	1.32 ± 0.18	1.48 ± 0.13	7.39 ± 0.13	−14.9	7.5	11.8	57.1	1.9	1	on
J0937+2602	0.16219	6.85 ± 0.32	20.9	7.1	−1.71 ± 0.13	1.75 ± 0.20	3.66 ± 0.24	11.93 ± 2.04	−12.8	8.6	15.4	5.9	1.9	1	on
J1003+3525	0.11886	45.77 ± 1.64	15.6	6.9	−1.40 ± 0.12	10.30 ± 1.02	2.34 ± 0.57	12.58 ± 0.63	−11.7	10.1	4.1	19.9	2	1	on
J1104+0118	0.57514	302.41 ± 6.30	7.0	0.7	79.89 ± 5.19	−43.38 ± 13.61	⋯	⋯	15.4	−3.2	⋯	⋯	1	2	off
J1104+6343	0.16427	4.46 ± 0.49	6.0	6.1	1.14 ± 0.14	0.33 ± 0.11	4.90 ± 0.16	0.79 ± 0.13	7.9	3.0	31.6	6.2	1	1.8	off
J1110−0003	0.21922	48.61 ± 1.34	8.0	6.1	0.27 ± 0.10	6.70 ± 0.92	1.78 ± 0.71	11.98 ± 1.01	2.8	7.3	2.5	11.9	2	1	on
J1115+0544	0.08995	17.27 ± 0.27	18.9	8.7	−1.23 ± 0.07	2.46 ± 0.18	0.05 ± 0.07	7.92 ± 1.11	−18.4	14.0	0.7	7.2	2	1	on
J1118+3203	0.3651	56.48 ± 2.29	4.9	4.9	11.76 ± 1.11	2.90 ± 0.51	⋯	10.31 ± 0.57	10.6	5.7	⋯	18.0	1	1	off*
J1132+0357	0.09089	16.26 ± 0.38	17.3	7.4	−3.23 ± 0.08	4.74 ± 0.40	0.78 ± 0.05	2.40 ± 0.17	−40.2	11.8	15.9	14.1	1.9	1	on
J1150+3632	0.34004	39.01 ± 2.28	4.9	5.2	5.02 ± 0.53	−3.29 ± 0.49	22.24 ± 1.25	3.18 ± 0.40	9.4	−6.7	17.8	7.9	1	1.9	off
J1152+3209	0.37432	138.05 ± 2.68	11.9	3.7	38.31 ± 1.16	10.53 ± 1.29	⋯	⋯	33.0	8.2	⋯	⋯	1	1	off*
J1259+5515	0.19865	17.95 ± 0.91	7.4	3.5	0.02 ± 0.24	5.53 ± 0.82	4.30 ± 0.38	9.21 ± 3.88	0.1	6.8	11.4	2.4	1.9	1	on
J1319+6753	0.16643	30.90 ± 1.46	12.1	7.3	0.36 ± 0.33	7.96 ± 1.16	2.37 ± 0.27	5.90 ± 4.43	1.1	6.8	8.6	1.3	1.9	1	on
J1358+4934	0.11592	5.50 ± 0.19	7.8	13.1	0.36 ± 0.09	0.96 ± 0.10	0.74 ± 0.06	1.98 ± 0.10	3.8	9.2	13.4	19.0	1.9	1	on
J1447+2833	0.16344	66.68 ± 0.97	7.8	13.1	−1.16 ± 0.29	2.49 ± 0.24	2.27 ± 0.19	11.63 ± 0.25	−3.9	10.6	12.2	47.2	1.9	1	on
J1533+0110	0.14268	<1.00	14.2	13.4	−3.98 ± 0.17	0.54 ± 0.12	1.00 ± 0.15	3.97 ± 0.29	−23.6	4.4	6.9	13.7	1.9	1.5	on
J1545+2511	0.11696	6.81 ± 0.06	19.6	17.1	−2.16 ± 0.15	0.71 ± 0.11	1.69 ± 0.12	5.90 ± 0.14	−14.4	6.7	14.4	41.3	1.9	1	on
J1550+4139	0.22014	28.26 ± 1.30	11.6	8.1	−4.41 ± 0.47	15.71 ± 1.48	2.47 ± 0.40	14.01 ± 0.89	−9.5	10.6	6.2	15.8	1.9	1	on
J1552+2737	0.08648	<1.00	12.3	7.8	−3.67 ± 0.05	−1.34 ± 0.10	0.38 ± 0.04	2.70 ± 0.05	−69.2	−13.9	9.8	59.1	1.9	1.9	on*

Note. λL₅₁₀₀ is the continuum luminosity at 5100 Å in the bright epoch spectrum. J1533+0110 and J1552+2737 are too red to fit a power-law continuum, with a upper limit of 1.00 × 10⁴² erg s⁻¹. S/N₁ and S/N₂ are the median S/N pixel⁻¹ of the former and recent epoch spectra, respectively. L_Hβ,1, (L_Hα,1) and L_Hβ,2 (L_Hα,2) are the luminosities of the broad Hβ (Hα) component in the former and recent epoch spectra. The negative luminosity means that there is absorption instead of emission lines. There is no Hα data when Hα moves out of the range of the spectrum. ${\rm{S}}/{{\rm{N}}}_{{\rm{H}}{\beta }_{1}}$ (${\rm{S}}/{{\rm{N}}}_{{\rm{H}}{\alpha }_{1}}$) and ${\rm{S}}/{{\rm{N}}}_{{\rm{H}}{\beta }_{2}}$ (${\rm{S}}/{{\rm{N}}}_{{\rm{H}}{\alpha }_{2}}$) are the S/Ns of broad Hβ (Hα) components in the former and recent epoch spectra. E1 and E2 describe the spectral types of the former and recent epoch spectra. Type describes the transition type.

Download table as:  ASCIITypeset image

According to the definition of AGN spectral types, we mark the spectral type as Type 1 when there are (>5σ) broad Hβ components detected and Type 2 when the S/N of the broad Hα is lower than 5. Specifically, when (>5σ) Hα is detected, we mark the spectal type as Type 1.9 (1.8 or 1.5) when there is no, <1σ, (weak, 1σ–3σ, or intermediate, 3σ–5σ) broad Hβ detected. As the Balmer emission lines disappeared in some spectra, we freed the parameter limit of the amplitude of the line fitting, which was assumed to be positive. The fitting results are shown in Table 6. The negative luminosity means that there is absorption instead of emission lines. As shown in Table 6, six CL AGNs transitioned between Type 1 and Type 2 classes, and 10 CL AGNs transitioned between Type 1 and Type 1.9 classes. In addition, the fitting procedure recognizes >5σ broad Hβ emission lines in both bright and faint-state spectra of J1118+3203 and J1152+3209, and <1σ broad Hβ lines in both bright and faint-state spectra of J1552+2737. Therefore, these three AGNs fail to fit the quantitative definition above. While the dramatic Hβ luminosity changes show that Hβ faded in J1118+3203 and J1152+3209 (marked as "off*" in Table 6) and enhanced in J1152+3209 (marked as "on*" in Table 6). Among them, 15 AGNs turned on, with broad Balmer emission lines emerging (or enhancing), as well as increased broadband flux. The broad Balmer lines of 6 CL AGNs disappeared (or faded), accompanied by dimming in the continuum. Among the 21 CL AGNs, 17 CL AGNs showed broad Hα, Hβ, Hγ, or even Hδ appeared or disappeared. Only broad Hα and Hβ changes were detected in J0831+3646, J1319+6753, and J1552+2737; clear emerging of broad Hα and Hβ emission lines, and increased flux in the optical and mid-infrared, confirmed their changes.

The continuum flux varies along with the CL AGN type transition. Table 7 summarizes the photometric variability of CL AGNs. We obtain the g-band variability, Δg, from the magnitude offset between the PS1 g-band magnitude, or the DELS g-band magnitude when available, and the SDSS g-band magnitude. We use the first epoch SDSS photometry if there are data at more than one epoch. The uncertainties of Δg are calculated from propagation of SDSS g-band magnitude uncertainty and PS1 g-band magnitude uncertainty, or the DELS g-band magnitude uncertainty when available. Similarly, the g − r color variation, Δ(g − r), is calculated by the color offset between PS1 g − r color, or the DELS g − r color when available, and the SDSS g − r color. The g-band variability ranges from −1.89 to 0.52 mag in this sample. As in some CL AGN cases, the PS1 or DELS images were taken before the type transition. We also calculate the spectrophotometry with the spectra convolved with the SDSS filters. The spectrophotometry are calibrated by setting the first epoch SDSS spectrophotometry equal to the SDSS photometry at the closest epoch. The variations of spectrophotometry in the g-band and g − r color are also listed in Table 7. In the mid-infrared, there is a series of exposures within one day with a long interval of half a year in WISE multi-epoch data. We calculate the 3σ-clipped mean magnitude of WISE in W1 and W2 every half a year. The uncertainties of W1 and W2 are obtained from the 3σ-clipped standard deviation of W1 and W2 magnitude every half a year. There are less than 10 epochs of WISE mean magnitude for a single object. For the turn-on objects, the WISE variability is calculated from the offset between the brightest epoch photometry and the first epoch photometry. For turn-off objects, the WISE variability is calculated from the offset between the last epoch photometry and the brightest epoch photometry. The magnitude and color variability are calculated between the later epoch and the former epoch. Therefore, the magnitude variability is negative if an object becomes brighter, and the color variability is negative if the object becomes bluer. Figure 2 shows the optical broadband flux changes along with mid-infrared flux. In our sample, the mid-infrared variability ΔW1 ranges from −1.01 to 0.68, and 15 of them varied for more than 0.2 mag ($| {\rm{\Delta }}W1| \gt 0.2$ mag). The mid-infrared flux is not significantly affected by dust extinction (Weingartner & Draine 2001). In the scenario of variable obscuration, the variation in the W1 band due to dust extinction yields a factor of ∼21 variability in g-band magnitude, according to the extinction curve in the optical and mid-infrared even considering micrometer-sized grains (Wang et al. 2015). A variability of 0.2 mag in the W1 band suggests approximately 4.2 mag variability in the g-band. The optical variability, from photometric and spectrophotometric data, is not consistent with the scenario of variable obscuration in 10 CL AGNs in our sample at more than the 3σ confidence level (J0831+3646, J0937+2602, J1003+3525, J1104+0118, J1110−0003, J1115+0544, J1132+0357, J1259+5515, J1447+2833, and J1550+4139) and in eight CL AGNs between the 1σ and 3σ confidence levels (J0849+2747, J1104+6343, J1150+3632, J1152+3209, J1319+6753, J1533+0110, J1545+2511, and J1552+2737). The accuracy of the spectrophotometry is lower than that of the photometry. Following-up with high-accuracy photometric data after the type transition can better constrain the mechanism of type transition. Using mid-infrared variability to select CL AGNs, we were inclined to select CL AGNs with intrinsic changes instead of varying obscuration.

Zoom In Zoom Out Reset image size

Table 7.  Variability of Changing-look AGNs

Name	Survey	Δgphot	${\rm{\Delta }}{(g-r)}_{\mathrm{phot}}$	Δgspec	${\rm{\Delta }}{(g-r)}_{\mathrm{spec}}$	ΔW1	Δ(W1 − W2)	δCRTS
		(mag)	(mag)	(mag)	(mag)	(mag)	(mag)	(mag)
J0831+3646	DELS	−0.41 ± 0.01	−0.35 ± 0.02	−0.57 ± 0.37	−0.48 ± 0.49	−0.35 ± 0.10	0.13 ± 0.14	1.15
J0849+2747	DELS	0.31 ± 0.01	0.29 ± 0.02	1.00 ± 0.65	0.69 ± 0.80	0.28 ± 0.13	−0.24 ± 0.21	1.17
J0909+4747	PS1	0.06 ± 0.01	−0.03 ± 0.02	−0.71 ± 0.29	−0.39 ± 0.39	−0.42 ± 0.10	0.10 ± 0.18	0.90
J0937+2602	DELS	−0.17 ± 0.01	−0.06 ± 0.01	−0.29 ± 0.32	−0.24 ± 0.44	−0.41 ± 0.10	0.07 ± 0.15	0.92
J1003+3525	DELS	−0.52 ± 0.01	−0.27 ± 0.01	−1.17 ± 0.21	−0.54 ± 0.29	−0.58 ± 0.16	0.16 ± 0.26	0.92
J1104+0118	DELS	1.89 ± 0.03	1.48 ± 0.04	6.14 ± 111.71	4.98 ± 111.72	0.68 ± 0.11	−0.56 ± 0.26	1.68
J1104+6343	PS1	0.55 ± 0.02	0.44 ± 0.03	0.08 ± 0.46	0.11 ± 0.53	0.31 ± 0.24	0.03 ± 0.36	0.93
J1110−0003	DELS	−0.42 ± 0.02	−0.33 ± 0.02	−0.87 ± 0.40	−0.43 ± 0.53	−0.65 ± 0.12	0.11 ± 0.27	0.95
J1115+0544	DELS	−0.21 ± 0.01	−0.11 ± 0.01	−0.82 ± 0.22	−0.44 ± 0.30	−1.01 ± 0.10	0.59 ± 0.16	0.97
J1118+3203	PS1	0.67 ± 0.04	0.49 ± 0.05	0.80 ± 0.33	0.30 ± 0.43	0.15 ± 0.41	0.20 ± 0.58	1.50
J1132+0357	PS1	−0.11 ± 0.01	−0.15 ± 0.01	−1.16 ± 0.19	−0.55 ± 0.26	−0.56 ± 0.05	0.19 ± 0.12	0.91
J1150+3632	DELS	1.59 ± 0.02	0.92 ± 0.03	0.95 ± 0.41	0.50 ± 0.47	0.30 ± 0.24	0.18 ± 0.37	1.38
J1152+3209	PS1	1.36 ± 0.03	0.99 ± 0.03	1.39 ± 0.60	0.84 ± 0.76	0.19 ± 0.08	−0.27 ± 0.13	⋯
J1259+5515	DELS	−0.06 ± 0.02	−0.05 ± 0.02	−0.77 ± 0.93	−0.40 ± 0.99	−0.54 ± 0.10	0.26 ± 0.26	1.06
J1319+6753	PS1	−0.28 ± 0.01	−0.26 ± 0.01	−0.46 ± 0.40	−0.17 ± 0.47	−0.23 ± 0.07	0.10 ± 0.12	0.70
J1358+4934	DELS	0.41 ± 0.01	0.20 ± 0.02	−0.46 ± 0.20	−0.15 ± 0.24	0.00 ± 0.16	0.00 ± 0.23	0.80
J1447+2833	PS1	−0.15 ± 0.01	−0.18 ± 0.01	−0.72 ± 0.20	−0.46 ± 0.29	−0.45 ± 0.05	0.06 ± 0.08	0.85
J1533+0110	DELS	−0.14 ± 0.01	−0.05 ± 0.01	−0.28 ± 0.18	−0.21 ± 0.20	−0.12 ± 0.06	0.05 ± 0.15	1.13
J1545+2511	PS1	−0.16 ± 0.01	−0.14 ± 0.01	−0.16 ± 0.48	0.03 ± 0.52	−0.10 ± 0.04	0.14 ± 0.07	0.84
J1550+4139	DELS	−0.25 ± 0.01	−0.13 ± 0.02	−0.55 ± 0.49	−0.24 ± 0.57	−0.34 ± 0.06	0.12 ± 0.15	1.14
J1552+2737	PS1	0.03 ± 0.01	−0.26 ± 0.01	−0.45 ± 0.36	−0.37 ± 0.47	−0.25 ± 0.08	0.12 ± 0.17	0.89
J0126−0839	PS1	0.43 ± 0.01	0.15 ± 0.02	0.29 ± 0.18	0.25 ± 0.21	0.00 ± 0.09	0.00 ± 0.27	0.91
J0159+0033	DELS	0.28 ± 0.02	0.22 ± 0.02	1.15 ± 0.24	0.60 ± 0.28	0.12 ± 0.21	−0.01 ± 0.41	0.95
J1011+5442	PS1	1.04 ± 0.02	0.53 ± 0.02	1.91 ± 0.12	0.46 ± 0.15	1.30 ± 0.21	−0.50 ± 0.36	0.86
J1554+3629	DELS	−0.56 ± 0.01	−0.42 ± 0.02	−1.37 ± 0.39	−0.67 ± 0.49	−0.74 ± 0.08	0.16 ± 0.17	1.22
J2336+0017	DELS	0.41 ± 0.02	0.30 ± 0.02	0.41 ± 0.40	0.22 ± 0.45	0.14 ± 0.28	−0.17 ± 0.42	1.06

Note. Δg_phot and ${\rm{\Delta }}{(g-r)}_{\mathrm{phot}}$ are the g-band variability and g − r color variability from imaging data, while Δg_spec and ${\rm{\Delta }}{(g-r)}_{\mathrm{spec}}$ are from spectrophotometry.

Download table as:  ASCIITypeset image

The optical and mid-infrared colors vary following flux changes. Figure 3 shows the color variability versus the magnitude variability in the optical (left panel) and mid-infrared (right panel). The bluer-when-brighter trend is a known trend in the optical (e.g., Wilhite et al. 2005; Schmidt et al. 2012; Zuo et al. 2012; Ruan et al. 2014). In the mid-infrared, there is a trend in which W1–W2 is redder when brighter. A similar W1–W2 color dependence on AGN luminosity is reported in the Swift/Burst Alert Telescope AGN sample (Ichikawa et al. 2017).

Zoom In Zoom Out Reset image size

The global SED variation helps to clarify the color changes in the infrared. We construct the infrared SED using the WISE data and the near-infrared data from the Two Micron All Sky Survey (2MASS; Skrutskie et al. 2006), which scanned the entire sky from 1997 to 2001 in three bands: J (1.25 μm), H (1.65 μm), and K_s (2.17 μm). Because there are no multi-epoch near-infrared and WISE W3 and W4 data, we show examples of two different objects in Figure 4. The two objects are J1115+0544 (left panel) at the faint state before it turned on and J0849+2747 (right panel) before it turned off.

Zoom In Zoom Out Reset image size

We fit the SED with the stellar and dust torus emission components. The simple stellar synthesis model by Bruzual & Charlot (2003, hereafter BC03) with an age of 5 Gyr and solar metallicity is used for the stellar emission. The dust torus component is described by a new version of the radiative transfer model CAT3D (Hönig & Kishimoto 2017). Due to limitations of the data, we adopt the templates without the wind component. We perform the SED fitting with a new Markov chain Monte Carlo (MCMC) method (Shangguan et al. 2018) combining the stellar and dust torus models.

In the SED before the AGN turned on, the W1 and W2 bands generally follow the stellar emission. In the SED after the AGN turned on, the W1 and W2 bands, especially W2, were strongly affected by the dust torus radiation. Therefore, it is possible that the W1–W2 shows as blue without a strong AGN contribution, and turns out to be red after the AGN turning on. Thus the mid-infrared W1–W2 color of CL AGNs turns to be redder when brighter.

In the optical, the color–magnitude variation relation (dashed line in left panel in Figure 3) of CL AGNs (listed in Table 5) is fitted by a least-squares fitting algorithm as

$$\begin{eqnarray}&&{\rm{\Delta }}(g-r)=(0.650\pm 0.042){\rm{\Delta }}g-(0.034\pm 0.017).\end{eqnarray} \tag{ 1 }$$

The CL AGNs selected from repeat spectroscopy are not biased by WISE variability selection criteria. For CL AGNs selected from repeat spectroscopy, the mid-infrared color–magnitude relation (dotted line in left panel in Figure 3) is fitted as

$$\begin{eqnarray}&&{\rm{\Delta }}(W1-W2)\\ &&\quad =\,(-0.404\pm 0.078){\rm{\Delta }}W1-(0.016\pm 0.033).\end{eqnarray} \tag{ 2 }$$

Thus, a selection criterion based on mid-infrared variability as $| {\rm{\Delta }}(W1-W2)| \gt 0.1$ when $| {\rm{\Delta }}W1| \gt 0.2$, is reasonable. With all the CL AGNs (in Table 5), the color–magnitude relation in the mid-infrared (dashed line in the left panel in Figure 3) is fitted as

$$\begin{eqnarray}&&{\rm{\Delta }}(W1-W2)\\ &&\quad =\,(-0.432\pm 0.066){\rm{\Delta }}W1-(0.028\pm 0.028).\end{eqnarray} \tag{ 3 }$$

The CL AGNs, selected from WISE variability, obey the color–magnitude relation well. The mid-infrared variability and color variability are good tracers for CL AGNs with intrinsic changes.

The current data coverage is not good enough to measure the timescale of transition. Therefore, we only roughly estimate the upper limit of type transition as follows. The upper limit of the transition timescale is obtained by the time interval between the former and recent spectroscopic epoch (Δt_spec in the rest-frame in Table 8). We further prescribe a limit to the timescale with the light curve changes approaching the transition. For example, although the two spectra of J1115+0544 were separated by 14 years, there was no variability detected in SDSS, PS1, and CRTS imaging data. The rapid flux brightening was detected by WISE in 2015. The W1 of J1115+0544 increased for more than 1 mag in a short interval, less than one year in the rest-frame. The light curves of CL AGNs keep quiescent at faint states, namely before the transition of turn-on CL AGNs or after the transition of turn-off CL AGNs. The CL AGNs usually vary in their AGN phase. In some cases (for example, J0831+3646, J1110−0003, J1115+0544, J1259+5515, and J1319+6753), the WISE flux increased when the AGN turned on, and decreased with a smaller amplitude later on, possibly due to the variability of the AGN accretion rate. On the other hand, it is also possible that the flux increase is due to the TDEs instead of AGN activity. In this scenario, the later light curve should continually decrease following the typical light curve of TDEs, which is proportional to t^−5/3 (Rees 1988; Lodato et al. 2009; Guillochon & Ramirez-Ruiz2013). Follow-up photometric data are needed to distinguish the scenarios of AGN accretion rate variability and TDEs. The epoch approaching the type transition is obtained by a recent quiescent image epoch at the faint state, or the brightest image epoch in their AGN phase. In Table 8, Epoch1* and Epoch2* show the imaging data epochs (and imaging survey) approaching the transition before and after the type transition. Δt, obtained from photometric variation approaching the type transition when available, is a better upper limit of the transition timescale in the rest-frame than Δt_spec. Therefore, the upper limit of the type transition timescales ranges from 0.9 to 12.6 years in the rest-frame.

The short transition timescale is not consistent with the scenario of variable obscuration (e.g., LaMassa et al. 2015; Gezari et al. 2017; Sheng et al. 2017). In a scenario of variable accretion rate, the CL timescale in this sample is much shorter than the inflow timescale of gas in the inner parts of the accretion disk, as discussed in previous works (LaMassa et al. 2015; MacLeod et al. 2016; Runnoe et al. 2016; Gezari et al. 2017). So modeling CL AGNs via changes in accretion rates is far from a settled matter. A detailed report on the mechanisms of the type transition will be presented in a subsequent paper.