Äîêóìåíò âçÿò èç êýøà ïîèñêîâîé ìàøèíû. Àäðåñ îðèãèíàëüíîãî äîêóìåíòà : http://hea-www.harvard.edu/~pgreen/Papers/RuanCLAGN2015.pdf Äàòà èçìåíåíèÿ: Sun Sep 27 22:40:41 2015 Äàòà èíäåêñèðîâàíèÿ: Sun Apr 10 10:00:03 2016 Êîäèðîâêà: Ïîèñêîâûå ñëîâà: ï ï ï ï ï ï ï ï ï ï ï

Draft version September 15, 2015
A Preprint typ eset using L TEX style emulateap j v. 5/2/11

TOWARDS AN UNDERSTANDING OF CHANGING-LOOK QUASARS WITH A STATISTICAL SAMPLE: AN ARCHIVAL SPECTROSCOPIC SEARCH IN SDSS
John J. Ruan1,2 , Scott F. Anderson2 , Sabrina L. Cales3,4 , Michael Eracleous5 , Paul J. Green6 , Eric Morganson6 , Jessie C. Runnoe5 , Yue Shen7,8 , Tessa D. Wilkinson2 , Michael R. Blanton9 , Tom Dwelly10 , Antonis Georgakakis10 , Jenny E. Greene11 , Stephanie M. LaMassa3 , Andrea Merloni10 , Donald P. Schneider5
Draft version September 15, 2015

arXiv:1509.03634v1 [astro-ph.HE] 11 Sep 2015

ABSTRACT The uncertain origin of the recently-discovered `changing-looking' quasar phenomenon ­ in which a luminous quasar dims significantly to a quiescent state in repeat spectroscopy over 10 year timescales ­ may present unexpected challenges to our understanding of quasar accretion. To better understand this phenomenon, we take a first step to building a statistical sample of changing-look quasars with a systematic but simple archival search for these ob jects in the Sloan Digital Sky Survey Data Release 12. By leveraging the >10 year baselines for ob jects with repeat spectroscopy, we uncover two new changing-look quasars. Decomposition of the multi-epoch spectra and analysis of the broad emission lines suggest that the quasar accretion disk emission dims due to rapidly decreasing accretion rates, while disfavoring changes in intrinsic dust extinction. Narrow emission line energetics also support intrinsic dimming of quasar emission as the origin for this phenomenon rather than transient tidal disruption events. Although our search criteria included quasars at all redshifts and quasar transitions from either quasar-like to galaxy-like states or the reverse, all the most confident changing-look quasars discovered thus far have been relatively low-redshift (z 0.2 - 0.3) and only exhibit quasar-like to galaxy-like transitions. Subject headings: galaxies: active, quasars: emission lines, quasars: general
1. INTRODUCTION The quasar phenomenon is thought to be a relatively short-lived stage of galaxy evolution involving rapid accretion onto the central supermassive black hole (SMBH). Observational constraints on lifetimes show that quasar phases in galaxies generally last for a total of 107-8 years (Martini & Weinberg 2001; Kelly et al. 2010), after which the accretion rate drops dramatically and the active nucleus transitions to a low-luminosity active galactic nucleus (AGN) or quiescent galaxy state (Churazov et al. 2005). Cosmological simulations of galaxy formation that include sub-grid models of SMBH growth and feedback have suggested that the accretion history of SMBHs may be episodic, where luminous quasar phases
Corresponding author: jruan@astro.washington.edu Department of Astronomy, University of Washington, Box 351580, Seattle, WA 98195, USA 3 Yale Center for Astronomy and Astrophysics, Physics Department, Yale University, New Haven, CT, 06511 4 Department of Astronomy, University of Concep cion, Concep cion, Chile 5 Department of Astronomy & Astrophysics and Institute for Gravitation and the Cosmos, 525 Davey Lab, The Pennsylvania State University, University Park, PA 16802, USA 6 Harvard Smithsonian Center for Astrophysics, 60 Garden St, Cambridge, MA 02138, USA 7 Kavli Institute for Astronomy and Astrophysics, Peking University, Beijing 100871, China 8 Carnegie Observatories, 813 Santa Barbara Street, Pasadena, CA 91101, USA 9 Center for Cosmology and Particle Physics, New York University, Department of Physics, New York University, 4 Washington Pl, New York, NY 10003 10 Max-Planck-Institut fur extraterrestrische Physik (MPE), Giessenbachstrasse 1, D-85748, Garching bei Munchen, Ger¨ many 11 Department of Astrophysical Sciences, Princeton University, Peyton Hall, Princeton, NJ 08544, USA
1 2

are regulated by quasar feedback processes (e.g., Di Matteo et al. 2005; Hopkins et al. 2005; Springel et al. 2005). Although the exact characteristics of quasar light curves over cosmic time are difficult to infer observationally, indirect arguments based on AGN populations have also suggested that AGNs dramatically `flicker' in luminosity between luminous quasar and quiescent galaxy phases (Schawinski et al. 2015). However, direct observations of such transitions in luminous quasars have thus far been scarce. It has been suggested that the transition from quasars to low-luminosity AGN or quiescent galaxies for individual ob jects may not be observable due to the long timescales expected for this process. Such dramatic changes in the accretion state are commonly observed in X-ray binaries, which can undergo spectral state transitions between the high-luminosity/soft-spectrum and low-luminosity/hard-spectrum states in the X-rays (e.g., Homan & Belloni 2005). Scaling the hours-long timescales observed for these spectral state transitions in X-ray binaries to 108 M mass SMBHs predicts transition timescales in quasars of 104-5 years (Sobolewska et al. 2011). Indirect evidence for a luminous quasar transition in an individual ob ject was previously provided by observations of Hanny's Voorwerp, a serendipitously discovered ionized emission-line gas cloud lying 20 kpc away from the quiescent galaxy IC2497 (Lintott et al. 2009). Based on multi-wavelength observations, it is argued that this gas cloud could only have been ionized by a AGN continuum with luminosity 1045 erg s-1 ; this implies that the nearby quiescent galaxy was recently in a quasar state, with a transition timescale of 104 years and consistent with expectations (Schawinski et al. 2010; Keel et al. 2012a). Since this discovery, many other candidate fading AGNs with extended emission-line regions


2

Ruan et al. to AGN accretion and structure, and warrant additional observations. Since the discovery of LaMassa et al. (2015), at least two more spectroscopic CL quasars have been serendipitously discovered (Cales et al. in preparation; Runnoe et al. in preparation), primarily through visual inspection of early spectroscopic data from the Time-Domain Spectroscopic Survey (Morganson et al. 2015) in the Sloan Digital Sky Survey-IV (SDSS-IV). These results suggest that CL quasars may be surprisingly common, and can be found by mining spectral data sets with large numbers of repeat quasar and galaxy spectra over a sufficiently long baseline. Motivated by these results, we present a simple archival search for CL quasars in SDSS-I/II/III Data Release 12 (DR12, Alam et al. 2015), which includes a total of 4 â 106 optical/near-IR spectra over 10,000 deg2 of sky. This data set includes a significant number of repeat spectra spanning a 10 year baseline, and many of its various science programs specifically targeted quasars and galaxies. We aim to cast a wide net in this investigation, and include quasar and galaxies at all redshifts while remaining impartial in our search with regard to whether each ob ject transitions from a quasar-like to galaxy-like state or vice versa; this approach could not only yield a large number of CL quasars, but also potentially the first spectroscopic quiescent galaxy to luminous quasar transition, which would have intriguing implications for their origin. The outline of this paper is as follows: Section 2, presents the data sets and criteria used in our search. In Section 3, we describe the changing quasars found in our search, and modeling of their broad emission lines. Section 4 evaluates evidence from our sample favoring various scenarios for the origin of CL quasars. We summarize and conclude in Section 5. Throughout this paper, we assume a standard CDM cosmology with m = 0.309, = 0.691, and H0 = 67.7 km s-1 Mpc-1 , consistent with the Planck full-mission results of Planck Collaboration et al. (2015).
2. AN ARCHIVAL SPECTROSCOPIC SEARCH

have been found and investigated, resulting in similar inferred transition timescales (e.g. Keel et al. 2012b, 2014). Previously, transitions of Seyfert 1 galaxies to Seyfert 1.8/1.9 (and vice versa) have been directly observed in repeat spectroscopy over timescales of 10 years (Goodrich 1995; Shappee et al. 2014; Denney et al. 2014). Goodrich (1995) showed that the origin of the observed transition in some of these AGNs are consistent with intrinsic changes in the AGN continuum emission, while variations in dust obscuration along the line of slight is favored for others, although these two effects may occur in concert if dust is embedded in the narrow-line region gas (Netzer & Laor 1993). The recent discovery of the first "changing-look" (CL) quasar by LaMassa et al. (2015) extends this transitional phenomenon to AGNs in new luminosity and redshift regimes (see Figure 1 of LaMassa et al. 2015). Repeat spectroscopy of this luminous quasar (SDSS J015957.64+003310.5) shows a dramatic decrease in the quasar continuum emission, accompanied by disappearance of the broad H line and strong dimming of the broad H line, showing that a simple orientation-based view of AGN unification is incomplete. Surprisingly, the observed transition in this CL quasar occurred over restframe timescales of 7 years; this is much shorter than the 104 year timescales expected for this transition to occur from previous arguments based on on X-ray binaries and extended emission line regions surrounding quiescent galaxies. The origin of changing-look behavior in luminous quasars is uncertain. LaMassa et al. (2015) demonstrated that the observed dimming of the quasar continuum in J0159+0033 coincides with broadening of the broad Balmer emission lines, such that the derived black hole mass (estimated through single-epoch spectroscopic black hole mass methods) is preserved. This behavior is consistent with intrinsic dimming of the quasar continuum emission, while a scenario in which the continuum and emission line dimming is caused by an increase in dust extinction is disfavored through modeling of the spectral changes. Furthermore, LaMassa et al. (2015) also argue that obscuration by a dust cloud outside the broad line region in a circular Keplerian orbit is unlikely since its crossing time across the broad line region would be much longer than the observed transition timescale. If the dimming of quasar emission in CL quasars is intrinsic, then the observed behavior can be caused by dramatic changes in the accretion flow, which can occur during transitions between radiatively efficient and inefficient accretion regimes (Ichimaru 1977; Rees et al. 1982; Narayan & Yi 1994). Thermal and dynamical instabilities in the accretion disk may also produce strong changes in the disk emission on even shorter timescales (Lin & Shields 1986; Siemiginowska et al. 1996). Merloni et al. (2015) argued that the LaMassa et al. (2015) CL quasar may instead be a transient stellar tidal disruption event (TDE) near the central SMBH (see also Eracleous et al. 1995), which would cause a luminous nuclear flare, followed by a slow dimming over the fewyears timescales observed. This scenario is supported by their difference-imaging light curves of this quasar, which show that the time evolution of the broadband nuclear emission is consistent with that expected from TDEs. In any of these scenarios, CL quasars represent an intriguing new phenomenon that can provide unique insights

2.1. Search Criteria We utilize the list of all 4,355,202 spectra in SDSS (York et al. 2000) DR12, and perform the selection cuts detailed below to produce a final sample of 117 CL quasar candidates. These spectra were taken by the SDSS 2.5m telescope (Gunn et al. 2006) using the SDSS-I/II and Baryon Oscillation Spectroscopic Survey (BOSS, Eisenstein et al. 2011; Dawson et al. 2013) spectrographs (Smee et al. 2013), and compiled in the `spAll' files produced by the SDSS spectroscopic reduction pipeline (Bolton et al. 2012). For a CL quasar to show a convincing transition, its multi-epoch spectra must clearly possess quasar-like spectral features in one epoch (powerlaw continuum and broad emission lines), and galaxy-like features in another epoch (absorption spectra and narrow emission lines if star-formation or nuclear activity is present). While a sophisticated method of detecting this transition in repeat spectra of each ob ject is likely to be more sensitive to subtle changes, our current goal is to search only for the most obvious and convincing cases of CL quasars. Thus our simple approach relies on the automated SDSS pipeline to classify each spectrum as quasar-


Archival Search for Changing-Look Quasars
Table 1 Measured SDSS sp ectral properties of the changing-lo ok quasars in our sample. SDSS Ob ject J015957.64+003310.5
b

3

z 0.312 0.198 0.243

MJD 51871 55201 52163 54465 52096c 55449

H FWHMa [km s-1 ] 3788 ± 5954 ± 4121 ± .... 6289 ± 7209 ± 163 857 223 1180 1367

log10 LH a [erg s-1 ] 42.36 ± 41.72 ± 42.00 ± .... 41.86 ± 41.48 ± 0.04 0.11 0.04 0.20 0.22

H FWHMa [km s-1 ] 4714 ± 682 .... 4297 ± 1165 .... 6993 ± 2271 ...

log10 LH a [erg s-1 ] 41.88 ± 0.63 .... 41.55 ± 0.20 .... 41.28 ± 0.20 ...

log10 L5100 [erg s-1 ] 43.52 ± 0.05 43.27 ± 0.06 43.43 ± 0.03 ... 43.04 ± 0.09 42.56 ± 0.18

J012648.08-083948.0 J233602.98+001728.7

a b c

These measurements of the luminosities and widths are for the broad comp onents of these Balmer lines. Changing-look quasar previously found by LaMassa et al. (2015) and also disused in Merloni et al. (2015). This MJD is the mean of four closely-spaced ep ochs of sp ectra that have been stacked (see discussion in Section 3.1).

Table 2 Inferred SDSS sp ectral properties of the changing-lo ok quasars in our sample. SDSS Ob ject J015957.64+003310.5
a

MJD 51871 55201 52163 54465 52096b 55449

log10 MBH,H [M ]



log10 MBH,H [M ]



log10 (Lbol /L

Edd,H

)

log10 (Lbol /L

Edd,H

)

J012648.08-083948.0 J233602.98+001728.7

7.93 ± 0.10 8.20 ± 0.26 7.96 ± 0.10 .... 8.13 ± 0.29 8.00 ± 0.30

8.02 ± 0.33 .... 7.89 ± 0.84 .... 8.11 ± 0.53 ...

-1.6 ± 0. -2.1 ± 1. -1.7 ± 0. .... -2.3 ± 0. -2.7 ± 2.

1 6 4 4 6

-1.7 ± 1.3 .... -1.7 ± 0.7 .... -2.3 ± 0.7 ...

a b

Changing-look quasar previously found by LaMassa et al. (2015) and discussed in Merloni et al. (2015). This MJD is the mean of four closely-spaced ep ochs of sp ectra that have been stacked (see discussion in Section 3.1).

like or galaxy-like. Specifically, using the CLASS spectral classification provided for each spectrum in our sample (which is based on fitting to a set of galaxy, quasar, and stellar eigenspectra, see Bolton et al. 2012), we create two subsamples: a galaxy-like sample of 2,510,060 spectra where CLASS = `GALAXY', and a quasar-like sample of 587,306 where CLASS = `QSO'. In both these subsamples, sky fibers have been removed using the sourcetype targeting keyword. Although it is well-known that these automated pipeline classifications occasionally fail to accurately classify the observed spectra, we emphasize that we are focusing on the most obvious and convincing CL quasars, for which the pipeline classifications will suffice for this initial archival search. For each spectrum in our galaxy-like sample, we positionally match to the quasar-like sample using a 1 matching radius to identify ob jects with repeat spectra and disparate classifications. We impose additional quality-control conditions on this search: the difference in the pipeline redshifts between the two epochs must be |z | < 0.01, and the absolute value of the rest-frame time lag between the two epochs is |trest-frame | > 4 years. These conditions remove a significant number of false-positives in which the pipeline catastrophically fails to fit the spectrum in one of the epochs. This search results in 180 pairs of repeat spectra of 117 unique ob jects (a few ob jects have more than two epochs of spectra), and we visually inspect all spectra of each of these CL quasar candidates. From the visual inspection, we find three clear cases

of CL quasars (listed in Table 1), which include the LaMassa et al. (2015) CL quasar (J01595+0033) and two additional new convincing cases (J01264-0839 and J2336+00172), both of which exhibit quasar-like to galaxy-like transitions. Specifically, the broad H emission in all three CL quasars disappears, while the broad H emission dims significantly (and disappears in J01264-0839), accompanied by dimming of blue quasar continuum emission. We additionally find one ambiguous CL quasar where the latest epoch of SDSS spectra appears to show a galaxy-like spectrum at blue wavelengths, but this SDSS spectrum is corrupted at the redder wavelengths. In the Appendix, we present additional recent non-SDSS spectroscopy which demonstrates that this ob ject does not transition to a galaxy-like state at the epoch of the latest spectrum; this behavior instead likely stems from known instrumental issues affecting this particular fiber. In the visual inspection of the multi-epoch spectra, the vast ma jority of the false-positives from our search were cases where the pipeline switched between CLASS = `GALAXY' and CLASS = `QSO' classifications in repeat spectra despite little change in the spectral properties. Often, this occurs for AGN at redshifts of z 0.4 in which the broad H emission line is redward of the smaller wavelength coverage of the SDSS spectrograph in the earlier epoch (leading to a CLASS = `GALAXY' classification), but visible in the later epoch from the BOSS spectrograph due to its slightly larger wavelength coverage (leading to a CLASS = `QSO' classification). As


4

Ruan et al. (2004a), and 170, 000 SDSS galaxy spectra from Yip et al. (2004b), respectively. Specifically, we fit combinations of the first five quasar and first five galaxy eigenspectra, with their amplitudes (i.e., PCA coefficients) as 10 free parameters. These fits are performed through a simple 2 minimization. Since the Petrosian radius measured in SDSS imaging for our three CL quasars are between 1.45 to 2.58 , they are spatially extended. The host-galaxy and quasar contributions in each resulting SDSS spectrum are thus dependent on the fiber diameter. SDSS-I/II spectra were obtained using 3 diameter fibers, while SDSS-III (BOSS) spectra were acquired using 2 diameter fibers. For J0159+0033, the earlier spectrum was obtained with a 3 fiber, while the later spectrum was obtained with a 2 fiber, thus we decompose the two spectra separately. In contrast, both epochs of spectra for J0126-0839 were obtained with 3 fibers, therefore the host-galaxy contribution should be constant between the two spectra. For this ob ject, we decompose both epochs of spectra simultaneously and impose the additional constraint of constant galaxy parameters between the two epochs of spectra. Finally, for J2336+0017, a total of five epochs of SDSS spectra are available, including four early epochs (with 3 fibers) within a 2-year timespan in the observed frame during its quasar-like phase (MJD of 51783, 51877, 52199, 52525), and an epoch 9 years later with a 2 fiber. Since the spectral changes in the four early epochs of this ob ject are relatively small, we simply use the mean spectrum of these four epochs in our spectral decomposition to achieve higher signal-to-noise, and our quoted MJD of 52096 is thus actually the mean MJD of the four early epochs. To accommodate the different fiber diameter of the early and later spectra, the galaxy parameters in the fitting are allowed to vary between the mean earlier spectrum and the later spectrum. In Section 4.2, we discuss the evolution in the continuum luminosity of this ob ject over its five separate spectroscopic epochs. The results of our spectral decomposition for each of the three CL quasars are shown in Figures 1-3, which display both observed epochs of each CL quasar, as well as their best-fit quasar and galaxy components from eigenspectra. Following previous conventions, we refer to the fitted quasar and galaxy spectra shown in Figures 1-3 as the `reconstructed' quasar and galaxy spectra, while the `decomposed' quasar spectra (not shown) are the original spectra with their corresponding reconstructed galaxy spectra subtracted. All of our broad emission line analysis in Section 4.2 is performed on the decomposed quasar spectra rather than the reconstructed quasar spectra, since this allows us to use the uncertainties on the flux densities from the original spectrum. Yip et al. (2004a,b) demonstrated that the first five eigenspectra in their PCA analysis captured 98.29% and 98.37% of the variance in their quasar and galaxy spectra samples, respectively. Not surprisingly, we find that extending the spectral decomposition to the first 10 quasar and galaxy eigenspectra did not noticeably improve the resulting fits to the observed spectra, and so we only utilize the first five in our analysis for simplicity. 3.2. Broad Emission Line Analysis Using our decomposed quasar spectra in both epochs for each of our CL quasars, we measure the proper-

part of the visual inspection, the fiber plugging positions and targeting flags for each pair of repeat spectra was compared to ensure that there is no offset in the fiber position between the two spectra, artificially leading to more host-galaxy emission in the SDSS spectrum. This offset can occur since some SDSS fibers were part of a SDSS program to test the redshift recovery of the spectroscopic pipeline in SDSS-III relative to in SDSS-I/II, and are identified using the PROGRAM = `APBIAS' target flag in the spectra as well as their disparate fiber plugging positions in repeat spectra. One of these ob jects was recovered in our search, and was removed from our sample.
3. SPECTRAL PROPERTIES OF

CHANGING-LOOK QUASARS 3.1. Spectral Decomposition We decompose the quasar and host-galaxy components of both epochs of spectra for each of the three CL quasars found in our search. For the spectral decomposition, we follow the general method of Vanden Berk et al. (2006) and Shen et al. (2015), with only minor modifications, which is based on fitting the spectra using a mix of quasar and galaxy eigenspectra created from a principal component analysis of large samples of SDSS spectra. This decomposition method using eigenspectra differs from the approach of LaMassa et al. (2015), who instead fit a model of power-law quasar continuum emission, host-galaxy emission generated from a stellar population synthesis model, and emission lines to the multiepoch spectra. The main advantage of our approach is that the continuum emission is fitted and decomposed empirically without having to rely on the accuracy of a power-law and stellar population synthesis models for the quasar and galaxy continuum emission, respectively. For example, a mixture of eigenspectra can more accurately describe a galaxy spectrum consisting of stars with a continuous range of ages in comparison to the simple star formation histories assumed by stellar population synthesis models. The primary disadvantage of our approach is that because narrow emission lines are present in both the quasar and galaxy eigenspectra, separating the narrow line emission from the continuum emission is less straightforward (this does not adversely affect the broad emission lines). Although the CL quasar J0159+0033 recovered in our search was previously discovered and analyzed by LaMassa et al. (2015), we nevertheless include it in our analysis below to demonstrate whether our independent spectral decomposition and broad emission line fitting for this ob ject produces results that are consistent with their published values. We first correct all spectra in our sample for Galactic extinction, using the maps of Schlafly & Finkbeiner (2011) and the Milky Way reddening law of Cardelli et al. (1989). To facilitate the spectral decomposition, we resample all our spectra and the egienspectra to a common wavelength grid of the form log10 = 3.35 + 0.001a, for integer a from 0 to 5,914. The wavelength coverage of this common wavelength grid is wide enough to accommodate all spectra in our sample, and is similar to the native SDSS resolution. For our spectral decomposition, we utilize the eigenspectra from the principal component analysis of 17,000 SDSS quasar spectra from Yip et al.


Archival Search for Changing-Look Quasars

5

SDSS J015957.64+003310.5, z = 0.312 25 MJD = 51871 20 15 10 5 0 25 3500 4000 4500 5000 5500 6000 6500 7000 MJD = 55201 20 15 10 5 0 3500 4000 4500 5000 5500 6000 6500 7000
rest-frame wavelength [ ]

F [1 0

-17

erg cm-2 s-1

-1

]

H 12 10 MJD = 51871 8 6 4 2 0 12 4750 4800 4850 4900 4950 5000 5050 wa 10 MJD = 55201velength [ ] 8 6 4 2 0 4750 4800 4850 4900 4950 5000 5050
rest-frame wavelength [ ]

14 12 MJD = 51871 10 8 6 4 2 0 14 6450 6500 6550 6600 6650 6700 6750 wa 12 MJD = 55201velength [ ] 10 8 6 4 2 0 6450 6500 6550 6600 6650 6700 6750
rest-frame wavelength [ ]

H

]

-1

erg cm-2 s-1

7

-1

Figure 1. Top: Sp ectral decomposition for the two ep ochs of sp ectra of SDSS J015957.64+003310.5 (see LaMassa et al. 2015), in the rest-frame. The black lines are the observed spectra, and the green and blue lines are the reconstructed quasar and host-galaxy spectra from the eigenspectra decomp osition, resp ectively. The b est-fit model to the observed sp ectrum from the decomp osition (i.e. sum of the green and blue lines) is the red line. The dramatic dimming in the quasar continuum and broad Balmer emission lines are consistent with intrinsic dimming of the accretion disk emission rather than dust extinction. Bottom left: Fitting of the H line region in the decomposed quasar spectrum, for the two epo chs of sp ectra. The decomp osed quasar sp ectra are the black lines, the b est-fit broad and narrow H emission lines are the blue lines, and the total fits to the decomposed quasar sp ectra (including quasar continuum and all emission lines) are shown in red. Although narrow emission lines are included in the fit, their amplitudes in the decomposed sp ectrum are not equivalent to the narrow emission lines in the observed spectrum since they are partially subtracted as part of the host-galaxy sp ectrum (see Section 3.2). Bottom right: Similar to the bottom left panels, but for the H emission lines.

F [1 0

F [1 0

-1

7

erg cm-2 s-1

-1

]


6

Ruan et al.

40 30 20 10 0 3500 4000 4500 5000 5500 6000 6500 7000 MJD = 54465 wavelength [ ] 40 30 20 10 0 3500 4000 4500 5000 5500 6000 6500 7000
H
rest-frame wavelength [ ]

SDSS J012648.08-083948.0, z = 0.198 MJD = 52163

F [1 0

-17

erg cm-2 s-1

-1

]

14 12 MJD = 52163 10 8 6 4 2 0 14 4750 4800 4850 4900 4950 5000 5050 wa 12 MJD = 54465velength [ ] 10 8 6 4 2 0 4750 4800 4850 4900 4950 5000 5050
rest-frame wavelength [ ]

H

30 25 MJD = 52163 20 15 10 5 0 6550 6600 30 6450 6500wavelength 6650 6700 6750 [] 25 MJD = 54465 20 15 10 5 0 6450 6500 6550 6600 6650 6700 6750
rest-frame wavelength [ ]

]

-1

erg cm-2 s-1

7

-1

Figure 2. Spectral decomp osition for the two ep ochs of sp ectra of SDSS J012648.08-083948.0 in the rest-frame, same format as Figure 1.

ties of the H and H broad emission lines to estimate black hole masses MBH and bolometric Eddington ratios Lbol /LEdd . Our broad emission line fitting procedure generally follows the method of Shen et al. (2011), in which the single-epoch virial MBH estimates are based on the broad emission line Full-Width Half Maximum (FWHMs), as well as a radius-luminosity relation for the broad line region from reverberation mapping of lowredshift AGNs. We emphasize that the narrow emission lines in the decomposed quasar spectra (observed spec-

F [1 0

F [1 0

-1

7

erg cm-2 s-1

-1

]

trum minus the best-fit galaxy spectrum) we use for the broad emission line fitting are not equivalent to those in the original observed spectrum. This is because narrow emission lines are also present in the galaxy eigenspectra (and thus the best-fit host-galaxy spectrum), which is subtracted to obtain the decomposed quasar spectra. However, we include the narrow emission lines in our analysis below to avoid biases in fitting the broad emission lines. For the H region of each ob ject, we use the de-


Archival Search for Changing-Look Quasars

7

SDSS J233602.98+001728.7, z = 0.243 12 MJD = 52096 10 8 6 4 2 0 3500 4000 4500 5000 5500 6000 6500 7000 12 MJD = 55449 10 8 6 4 2 0 3500 4000 4500 5000 5500 6000 6500 7000
H
rest-frame wavelength [ ]

F [1 0

-17

erg cm-2 s-1

-1

]

8 MJD = 52096 6 4 2 0 4750 4800 4850 4900 4950 5000 5050 8 MJD = 55449velength [ ] wa 6 4 2 0 4750 4800 4850 4900 4950 5000 5050
rest-frame wavelength [ ]

7 6 MJD = 52096 5 4 3 2 1 0 7 6450 6500 6550 6600 6650 6700 6750 wa 6 MJD = 55449 velength [ ] 5 4 3 2 1 0 6450 6500 6550 6600 6650 6700 6750
rest-frame wavelength [ ]

H

]

-1

erg cm-2 s-1

7

-1

F [1 0

Figure 3. Spectral decomp osition for the two epo chs of sp ectra of SDSS J233602.98+001728.7 in the rest-frame, same format as Figure 1. We note that the earlier ep och of spectra at MJD 52096 is a mean stack of four epochs of SDSS sp ectra taken within a 2-year p eriod (MJD of 51783, 51877, 52199, 52525), during which no strong sp ectral changes were observed.

composed quasar spectrum and fit the local continuum emission in the continuum wavelength windows of [6400, 6500]° and [6800, 7000]° to a power-law. In the H line A A wavelength window of [6500, 6800]° we fit for the narA, row H component, the [N II] 6548,6584 doublet, and the [S II] 6717,6731 doublet, using a single Gaussian for each emission line. The redshifts of the narrow lines are constrained to be the same, and their widths are con-

F [1 0

-1

7

erg cm-2 s-1

-1

]

strained to be <1200 km s-1 . The broad component of the H emission is fit with a Gaussian with width constrained to be >1200 km s-1 , with its central wavelength as a free parameter. Similarly, for the H region, we fit a local power-law to the continuum wavelength windows of [4435, 4700]° and A [5100, 5535]° and we fit emission lines in the H line A, wavelength window of [4700, 5007]° In the continuum A.


8

Ruan et al.

wavelength window, we include the optical Fe II template of Boroson & Green (1992) in the fit. However, for each of our three CL quasars in the dimmer galaxy-like epoch and two in the brighter quasar-like epoch (J012648.08083948.0 and J233602.98+001728.7) , the Fe II emission is weak and the template fit is poorly constrained. Thus, the Fe II template is included in the continuum fit only for J0159+0033 in its brighter quasar-like spectral epoch. For the narrow lines, we fit single Gaussians for the narrow component of H and the [O III] 4959,5007 doublet, with widths constrained to be <1200 km s-1 and redshifts constrained to be the same. The broad H component is fit with a single Gaussian with width >1200 km s-1 , and its central wavelength is a free parameter. The above spectral fitting produced the FWHM and luminosities of the broad H and H components, as well as the quasar continuum luminosity at 5100° L5100 , for A each epoch of our CL quasars. These measured properties of the broad emission lines are tabulated in Table 1, and the spectral fits are also presented in Figures 1-3. All uncertainties are calculated through 103 Monte Carlo resamplings of the flux density values in each spectrum from a Gaussian based on the 1 flux density uncertainties, and refitting the parameters on each resampled spectrum. The 1 uncertainties quoted for each parameter are thus the 1 spread in the resulting distributions of resampled parameters. In the latest galaxy-like epoch, the broad H components for all three of our CL quasars and the broad H component in J01264-0839 are not detected in our fitting above the uncertainties, and are thus excluded from the final fits. Using the H broad emission line FWHM and luminosities from our fits, single-epoch black hole masses MBH,H are estimated using the relation from Greene et al. (2010) of M
BH,H

14
-1

12 10 8 6 4 2 0 6400

F [1 0

-1

7

erg cm-2 s-1

J233602.98+001728.7, H best-fit dereddened spectrum MJD = 55449 decomposed quasar spectrum MJD = 52096

]

6500

6600 6700 wavelength [ ]

6800

Figure 4. Decomp osed quasar sp ectrum of J2336+0017 in the H region from the earlier, quasar-like epo ch (red), compared to the dereddened decomp osed quasar sp ectrum from the later, galaxylike epo ch (blue). Although changes in dust extinction can reasonably explain the dimming of the continuum emission in this changing-lo ok quasar, the extinction required cannot explain the strong changes in the broad emission line (see discussion in Section 4.1), disfavoring an extinction origin for changing-lo ok quasars.

within the uncertainties, validating our broad emission line fits.
4. DISCUSSION Using our sample of CL quasars, we investigate the origin of this phenomenon. In particular, we will focus on evidence for and against variable dust obscuration, tidal disruption events, and intrinsic dimming of quasar emission, below.

= 9.7 â 10 â

6

FWHM(H) 1000 km s-1
0.519 1

2.06

L5100 44 erg s- 10

M,

(1)

based on the radius-luminosity relation of Bentz et al. (2009). To verify these H-based estimates, we also estimate single-epoch black hole masses based on our H broad emission line fits MBH,H using the relation from Vestergaard & Peterson (2006) of M
BH,H

= 106 â

.91

FWHM(H ) 1000 km s-1
0.5 1

2

L5100 1044 erg s-

M.

(2)

We estimate the bolometric luminosity Lbol of our quasars by multiplying the continuum luminosity measured from the decomposed quasar spectra at 5100° A, L5100 , by the bolometric correction factor of 8.1 from Runnoe et al. (2012). The Eddington ratio is then Lbol /LEdd = Lbol /(1.3 â 1038 MBH ), for MBH in units of M , and Lbol in units of erg s-1 . These inferred quantities are tabulated for both epochs of our CL quasars in Table 2. The MBH and Lbol /LEdd values derived for each epoch of spectra from H and H are consistent to

4.1. A Dust Extinction Origin? Previous investigations of Seyfert 1.8/1.9 to Seyfert 2 transitions in lower luminosity CL AGNs (and vice versa) observed in repeat spectroscopy by Goodrich (1995) showed that the spectral changes in at least some cases were consistent with strong changes in extinction. This result suggests that extinction by an intervening dust cloud outside the broad line region could cause the strong dimming of the quasar continuum and broad emission lines we observed in our current sample of CL quasars. However, LaMassa et al. (2015) reported that while the dimming of the continuum emission in J0159+0033 (recovered in our search) can reasonably be modeled as due to an increase in dust extinction, the changes in the broad emission lines are poorly fit by the same extinction model. We perform a similar analysis on the CL quasar J2336+0017 to investigate whether dust extinction is consistent with its observed spectral changes. Since the quasar emission in the third CL quasar in our sample (J0126-0839) dimmed completely and is not observed in the second spectral epoch, our extinction analysis cannot be performed on this ob ject. We deredden the decomposed quasar spectrum from the later, galaxy-like epoch of J2336+0017, and fit a E (B - V ) value for which the continuum of the dereddened spectrum best matches the decomposed quasar spectrum from the earlier, quasar-like epoch of this ob-


Archival Search for Changing-Look Quasars ject. This fitted value of E (B - V ) = 0.43 is determined by minimizing the 2 between the two spectra (incorporating all uncertainties) in the wavelength regions outside the H and H wavelength windows discussed in Section 3.2. A Cardelli et al. (1989) reddening law for the dust extinction in the host galaxy is assumed, with RV = 3.1. Figure 4 compares the best-fit dereddened quasar spectrum from the later epoch in the H region to that from the decomposed quasar spectrum from the earlier epoch. It is clear that if the dimming of the continuum emission between the two spectral epochs is caused purely by dust extinction, the change in extinction required is not consistent with the observed changes in the H emission. Furthermore, the profile of the broad H component in Figure 4 broadens between the two epochs of spectra; in an extinction scenario, broadening of the H profile implies that the emission from the outer lowervelocity regions of the broad-lien region is attenuated more than emission from the inner portions. However, given that the quasar continuum from the central accretion disk is also obscured, such a configuration of the obscuring material is unlikely. These results thus disfavor a dust extinction origin for the CL quasar behavior in J2336+0017, similar to the results of LaMassa et al. (2015) for J0159+0033. 4.2. Narrow Emission Lines As noted in Section 1, Merloni et al. (2015) demonstrated that the long-term light curve of nuclear emission in the LaMassa et al. (2015) CL quasar recovered in our search appears to show temporal evolution consistent with TDEs. However, they also present several issues with this interpretation stemming from the strong broad and narrow emission lines observed in the SDSS spectrum. Specifically, the mass of gas in the broad line region inferred by Merloni et al. (2015) from the spectrum (of order 100 M ) is significantly more than could be provided from TDE debris. Furthermore, the possible TDE in this galaxy is unlikely to have ionized the gas producing the observed narrow lines, since the light travel time to the narrow line region (distances of kpc scales, e.g. Liu et al. 2013; Hainline et al. 2013) is 103-4 years. Interestingly, the relative intensities of the narrow emission lines of this CL quasar are not consistent with stellar photoionization, leaving a long-lived AGN as the most plausible power source. Merloni et al. (2015) suggest that these issues with the broad and narrow emission lines for a TDE interpretation of J0159+0033 can be resolved if the galaxy was in a recent active quasar phase, 104 years prior to the TDE. Here, we examine the narrow line properties for the three CL quasars in our sample, and compare them to those from AGN and TDE spectra. Our eigenspectra-based spectral decomposition method in Section 3.1 does not allow a clean separation of the narrow emission lines from the underlying stellar and quasar continuum emission. We therefore utilize the stellar population and emission line fits to SDSS DR12 spectra by Thomas et al. (2013)12 to examine our CL quasars on a classic BPT (Baldwin et al. 1981) diagram. These fits are performed using the Gas and Absorption
12

9

1.0

log([O III] 5007 / H 4861)

0.5 0.0

AGN

Star-Forming
0.5 1.0

SF+AGN

log([N II] 6583 / H 6563)

0.8

0.6

0.4

0.2 0.0

0.2

Figure 5. BPT diagram of the three changing-lo ok quasars in our sample, based on the emission line ratios measured in their latest SDSS spectrum. The line ratios for SDSS J015957.64+003310.5 (red circle), SDSS J012648.08-083948.0 (green star), SDSS J233602.98+001728.7 (blue diamond), and their 1 uncertainties are shown along with all emission line galaxies in SDSS-II I DR12 (blue contours) for comparison. The BPT diagram classification schemes of Kauffmann et al. (2003) (dashed line) and Kewley et al. (2001) (dotted line) are shown. The changing-lo ok quasars app ear to exhibit emission line ratios that are consistent with AGN-like or composite AGN and stellar ionizing continuum emission rather than p owered purely by star formation alone.

www.sdss.org/dr12/spectro/galaxy portsmouth/#kinematics

Line Fitting (GANDALF) software described in Sarzi et al. (2006), and include Gaussian fits to a variety of emission lines to obtain line fluxes. We specifically use the [N II] 6583 to H and [O III] 5007 to H narrow line ratios measured for the latest SDSS spectrum of our CL quasars (during the fainter, galaxy-like state), as shown in the BPT diagram in Figure 5. Figure 5 also indicates these line ratios for all galaxies in SDSS-III DR12 with all four emission lines detected at >3 significance, as well as the BPT classification scheme of Kauffmann et al. (2003) and Kewley et al. (2001), which distinguishes galaxies with emission lines ionized by AGN-like, stellar-like, and composite AGN/stellar continuum. These line ratios for J0159+0033 published by Thomas et al. (2013) are consistent with those independently determined by LaMassa et al. (2015) and Merloni et al. (2015) for the latest SDSS spectrum. Figure 4 clearly demonstrates that the emission lines in our sample of CL quasars are ionized at least in part by an AGN-like continuum; this AGN-like continuum could be provided by either a TDE or quasar disk emission. We examine the fluxes of the narrow emission line, focusing on the strongest narrow line, [O III] 5007. Lines fluxes reported for [O III] 5007 from Thomas et al. (2013) for the three CL quasars in our sample range from (1.2 - 1.7) â 10-15 erg s-1 cm-2 , which correspond to luminosities of (1.7 - 5.3) â 1041 erg s-1 in the restframe. This is in contrast to narrow line emission observed in spectra of UV/optical TDEs, which show no or significantly fainter [O III] 5007 emission (e.g. Gezari et al. 2006, 2009, 2012; Holoien et al. 2014; Chornock et al. 2014). The faint [O III] 5007 lines detected in these TDEs also have ratios relative to other lines that


10

Ruan et al.

5100

[erg s-1 ]

10

43

best fit, t-1.13 ±0.20 TDE model, t-1.67

the power-law index is the 1 spread in the resampled light curves. The light curve's best-fit power-law index of -1.13 ± 0.20 is statically shallower than the -5/3 model predicted for TDEs, although this only weakly disfavors the TDE model due to the sparse light curve sampling and uncertainties in the models for temporal evolution of TDE luminosity. 4.4. Transition Types Our search criteria for CL quasars should allow us to find cases in which ob jects spectroscopically transformed from quasar-like to galaxy-like, as well as the reverse. However, in our visual inspection of the CL quasar candidates, we find no confident cases of galaxy-like to quasar-like CL quasars. Although this reverse transition (galaxy-like to quasar-like) has been observed previously in Seyfert galaxies (e.g., Goodrich 1995; Shappee et al. 2014), these Seyferts are at lower redshifts and lower luminosities than our CL quasars. It is thus unclear whether the lack of reverse transitions in our sample is due to selection effects, a random result stemming from our small sample size, or a true feature of CL quasars. We first consider potential selection effects that may cause us to select CL quasars of only the quasar-like to galaxy-like variety. For example, if SDSS systematically targeted quasars for repeat spectroscopy in greater numbers than galaxies, then naturally we would find more quasar-like to galaxy-like transitions. To test this possibility, we identify the number of repeat spectra in SDSS DR12 of ob jects with CLASS = `GALAXY' and ob jects with CLASS = `QSO', using the same search criteria described in Section 2.1. There are 11,438 repeat spectra of 8,865 unique galaxies, and 7,109 repeat spectra of 5,990 quasars (some ob jects had more than two epochs of spectra). This test shows that our sample of only quasar-like to galaxy-like CL quasars is not simply due to a lack of repeat galaxy spectra in SDSS. However, the myriad other potential selection effects, such as those stemming from the disparate SDSS galaxy and quasar completeness magnitude limits, the quasar and galaxy targeting methods in SDSS, and differences in the galaxy populations between those that host quasars and those targeted by SDSS spectroscopy make it difficult to draw robust conclusions on selection effects without more careful considerations. We next consider whether our yield of three CL quasars of the quasar-like to galaxy-like variety is simply a random result due to the small number of CL quasars thus far discovered. Under the simple assumption that transitions of either variety are equally likely both intrinsically and observationally (i.e. neglecting selection effects), the probability of observing three quasar-like to galaxy-like transition and no cases of the reverse is approximately 0.53 = 12.5% from the binomial distribution. Extending our search to all SDSS-IV spectra (i.e. post-DR12) available as of 13 March 2015, we find one additional CL quasar (Runnoe et al., in preparation) also of the quasar-like to galaxy-like variety; including this additional CL quasar lowers the probability of our results to 0.54 = 6.25%. In the absence of selection effects, it thus appears unlikely for our search to randomly yield only quasar-like to galaxy-like CL quasars if the reverse transition is equally likely, though the statistics are not yet compelling.

L

10

42

10

3

MJD - 50000 [days]

10

4

Figure 6. The light curve of the continuum luminosity in the decomp osed quasar sp ectra of SDSS J233602.98+001728.7. The b est-fit p ower-law mo del is shown as a black solid line, and the b est-fit power-law mo del with spectral index fixed to the -5/3 value expected from tidal disruption events is shown as a black dashed line.

are consistent with star formation rather than AGN photoionization, which may not be surprising given the long light-travel times (102-3 years) for the ionizing TDE continuum to reach the narrow line region gas. Thus, the combination of timescales, narrow emission line luminosities, and emission line ratios in our sample of CL quasars all suggest that CL quasars are linked to quasar activity rather than TDEs. 4.3. The Nuclear Light Curve of SDSS J233602.98+001728.7 The CL quasar J2336+00172 in our sample has a total of five epochs of spectra, including four early epochs in its quasar-like state over approximately 2 years in the observed frame, and one later epoch in its galaxy-like state. Although we stacked the four early epochs of spectra in our earlier spectral analysis since the broad emission lines did not show noticeable evolution, here we decompose separately all five epochs of spectra to study the time-evolution of the quasar continuum emission. A t-5/3 temporal evolution of the decaying continuum luminosity is often taken to be an observational signature of TDEs, since it is the theoretical rate at which the tidal debris is expected to fall back towards the SMBH (Rees 1988; Lodato et al. 2009; Strubbe & Quataert 2009), although the luminosity evolution may not necessarily follow this rate (Lodato & Rossi 2011). We compare the quasar light curve generated from our spectral decomposition of the five spectroscopic epochs to this expected TDE luminosity evolution. Using the same decomposition method as described in Section 3.1, we measure the continuum luminosity at 5100° L5100 of the decomposed quasar spectrum for A the five spectroscopic epochs of J2336+00172. Figure 5 presents the light curve, along with the best-fit powerlaw of t-1.13±0.20 , and the best-fit TDE t-5/3 model. The uncertainty on this power-law index is calculated by 103 Monte Carlo resamplings of each point on the light curve from a Gaussian based on their 1 uncertainties, to produce 103 resampled light curves. Each resampled light curve is refit, and the quoted 1 uncertainty on


Archival Search for Changing-Look Quasars 4.5. Infal l Timescales The observed 10 year timescales we observe for dimming in our sample of CL quasars is significantly shorter than the 104 years expect for this transition to occur from scaling the observed timescales for spectral state transitions in X-ray binaries to 108 M BHs. Here, we assess whether the transition timescales we observe for CL quasars is consistent with the infall timescale of gas in the radiation-pressure dominated inner regions of Shakura-Sunyaev (Shakura & Sunyaev 1973) thin accretion disks, which is also the timescale on which changes in the accretion rate are reflected in changes in the continuum luminosity. Using Equation 5 from LaMassa et al. (2015), we find that the infall timescale for our sample of changing look quasars are approximately 42, 38, and 868 years for J0159+0033, J0126-0839, and J2336+0017, respectively. Although these infall timescales are longer than the transition timescales we observed, we note that the transition timescales we use for our CL quasars are lower limits since the the two epochs of spectra do not fully encompass the full transition (e.g., it is likely that the CL quasars have begun dimming before the first spectral epoch). Furthermore, LaMassa et al. (2015) also point out that magneto-hydrodynamic simulations of quasar accretion flows have suggested that the infall timescale may be a factor of a few shorter than these analytical estimates (Krolik et al. 2005), and much closer to the observed 10 year timescales. However, the 868 year infall timescale estimated for J2336+0017 is problematic in this interpretation, and may indicate that other processes such as thermal or dynamical instabilities may be present in the accretion flows.
5. CONCLUSIONS The discovery of CL quasars presents a new opportunity to study the nuclear environment and structure of quasars, once the origin of this phenomenon is understood. To provide a substantial sample of these ob jects, we performed an archival spectroscopic search in SDSS, yielding three CL quasars, including two new cases. Using this sample, we investigate the detailed properties of their quasar continuum emission and broad and narrow emission lines, with the goal of attempting to discriminate between various possibilities for the origin of this phenomenon. The primary results of our investigation can be summarized as below:

11

uum emission and broad emission lines have disappeared below the SDSS detection limit. · Changes in dust extinction required to match the dimming in the quasar continuum cannot account for the changes in the broad emission lines in either of our two CL quasars for which this analysis was possible, disfavoring an extinction origin for this phenomenon. Narrow emission line diagnostics show that our CL quasars all have luminous narrow lines with line ratios consistent with at least partially AGN-like ionizing emission. We argue that these narrow line properties favors a scenario in which the quasar continuum dims intrinsically over a TDE origin for this phenomenon. If the intrinsic dimming of the quasar emission favored by our analysis is due to draining of the underlying quasar accretion disk, the CL quasar phenomenon will provide a unique new laboratory to study the accretion flow and nuclear environment in luminous AGNs. Longterm spectroscopic and multi-wavelength monitoring of the currently-known changing-look quasars can help further elucidate the origin of CL quasar transitions. For example, observations of the return of any CL quasar to a bright quasar-like state would provide additional constrains on physical mechanism of this phenomenon as well as estimates of its duty cycle. Current multiob ject spectroscopic programs and time-domain imaging surveys are well-poised to serendipitously discover many more CL quasars. For example, the Time-Domain Spectroscopic Survey (Morganson et al. 2015) in SDSS-IV will provide repeat spectroscopy of several thousand lowredshift quasars, while the Pan-STARRS 3 survey has repeated imaged 30,000 deg2 of sky (including the SDSS imaging footprint). With future instruments and surveys such as the Dark Energy Spectroscopic Instrument (DESI, Levi et al. 2013) and the Large Synoptic Survey Telescope (LSST Ivezic et al. 2008), discovery of such rare phenomena will become more routine. JJR thanks James R.A. Davenport for helpful discussions. Support for JJR was provided by NASA through Fermi Guest Investigator grant NNX14AQ23G. Funding for SDSS-III has been provided by the Alfred P. Sloan Foundation, the Participating Institutions, the National Science Foundation, and the U.S. Department of Energy Office of Science. The SDSS-III web site is http://www.sdss3.org/. SDSS-III is managed by the Astrophysical Research Consortium for the Participating Institutions of the SDSS-III Collaboration including the University of Arizona, the Brazilian Participation Group, Brookhaven National Laboratory, Carnegie Mellon University, University of Florida, the French Participation Group, the German Participation Group, Harvard University, the Instituto de Astrofisica de Canarias, the Michigan State/Notre Dame/JINA Participation Group, Johns Hopkins University, Lawrence Berkeley National Laboratory, Max Planck Institute for Astrophysics, Max Planck Institute for Extraterrestrial Physics, New Mexico State University, New York University, Ohio State University, Pennsylvania State University, University of Portsmouth, Princeton University, the Spanish Participation Group,

· The three CL quasars in our sample appear to show similar properties: they are luminous (Lbol 1044.0-44.5 erg s-1 ) quasars at relatively low redshifts (z 0.2 - 0.3) that display strong dimming of the quasar continuum and the broad H Balmer emission lines over timescales of approximately 5 to 7 years in the rest-frame. · Analysis of the variability in the broad emission lines demonstrates that their decrease in luminosity occurs such that the line widths simultaneously broaden, preserving the derived black hole masses. Their Eddington ratios decrease from Lbol /LEdd 0.03 - 0.005 until the broad H Balmer lines have dimmed significantly or disappeared in some cases. In one CL quasar (J0126-0839), all quasar contin-


12

Ruan et al.
Kewley, L. J., Dopita, M. A., Sutherland, R. S., Heisler, C. A., & Trevena, J. 2001, ApJ, 556, 121 Krolik, J. H., Hawley, J. F., & Hirose, S. 2005, ApJ, 622, 1008 LaMassa, S. M., Cales, S., Moran, E. C., et al. 2015, ApJ, 800, 144 Levi, M., Bebek, C., Beers, T., et al. 2013, arXiv:1308.0847 Lin, D. N. C., & Shields, G. A. 1986, ApJ, 305, 28 Lintott, C. J., Schawinski, K., Keel, W., et al. 2009, MNRAS, 399, 129 Liu, G., Zakamska, N. L., Greene, J. E., Nesvadba, N. P. H., & Liu, X. 2013, MNRAS, 430, 2327 Lodato, G., King, A. R., & Pringle, J. E. 2009, MNRAS, 392, 332 Lodato, G., & Rossi, E. M. 2011, MNRAS, 410, 359 Martini, P., & Weinb erg, D. H. 2001, ApJ, 547, 12 Merloni, A., Dwelly, T., Salvato, M., et al. 2015, arXiv:1503.04870 Morganson, E., Green, P. J., Anderson, S. F., et al. 2015, arXiv:1505.00760 Narayan, R., & Yi, I. 1994, ApJ, 428, L13 Netzer, H., & Laor, A. 1993, ApJ, 404, L51 Planck Collaboration, Ade, P. A. R., Aghanim, N., et al. 2015, arXiv:1502.01589 Plotkin, R. M., Anderson, S. F., Brandt, W. N., et al. 2010, AJ, 139, 390 Rees, M. J., Begelman, M. C., Blandford, R. D., & Phinney, E. S. 1982, Nature, 295, 17 Rees, M. J. 1988, Nature, 333, 523 Runnoe, J. C., Brotherton, M. S., & Shang, Z. 2012, MNRAS, 422, 478 Sarzi, M., Falc´ on-Barroso, J., Davies, R. L., et al. 2006, MNRAS, 366, 1151 Schawinski, K., Evans, D. A., Virani, S., et al. 2010, ApJ, 724, L30 Schawinski, K., Koss, M., Berney, S., & Sartori, L. F. 2015, MNRAS, 451, 2517 Schlafly, E. F., & Finkbeiner, D. P. 2011, ApJ, 737, 103 Siemiginowska, A., Czerny, B., & Kostyunin, V. 1996, ApJ, 458, 491 Shakura, N. I., & Sunyaev, R. A. 1973, A&A, 24, 337 Shappee, B. J., Prieto, J. L., Grupe, D., et al. 2014, ApJ, 788, 48 Shen, Y., Richards, G. T., Strauss, M. A., et al. 2011, ApJS, 194, 45 Shen, Y., Greene, J. E., Ho, L. C., et al. 2015, arXiv:1502.01034 Smee, S. A., Gunn, J. E., Uomoto, A., et al. 2013, AJ, 146, 32 Sobolewska, M. A., Siemiginowska, A., & Gierlinski, M. 2011, ´ MNRAS, 413, 2259 Springel, V., Di Matteo, T., & Hernquist, L. 2005, MNRAS, 361, 776 Strubbe, L. E., & Quataert, E. 2009, MNRAS, 400, 2070 Thomas, D., Steele, O., Maraston, C., et al. 2013, MNRAS, 431, 1383 Vanden Berk, D. E., Shen, J., Yip, C.-W., et al. 2006, AJ, 131, 84 Vestergaard, M., & Peterson, B. M. 2006, ApJ, 641, 689 Yip, C. W., Connolly, A. J., Vanden Berk, D. E., et al. 2004a, AJ, 128, 2603 Yip, C. W., Connolly, A. J., Szalay, A. S., et al. 2004b, AJ, 128, 585 York, D. G., Adelman, J., Anderson, J. E., Jr., et al. 2000, AJ, 120, 1579

University of Tokyo, University of Utah, Vanderbilt University, University of Virginia, University of Washington, and Yale University. Based on observations obtained with the Apache Point Observatory 3.5-meter telescope, which is owned and operated by the Astrophysical Research Consortium.
REFERENCES

Alam, S., Albareti, F. D., Allende Prieto, C., et al. 2015, arXiv:1501.00963 Arcavi, I., Gal-Yam, A., Sullivan, M., et al. 2014, ApJ, 793, 38 Baldwin, J. A., Phillips, M. M., & Terlevich, R. 1981, PASP, 93, 5 Bentz, M. C., Peterson, B. M., Netzer, H., Pogge, R. W., & Vestergaard, M. 2009, ApJ, 697, 160 ´ Bolton, A. S., Schlegel, D. J., Aubourg, E., et al. 2012, AJ, 144, 144 Boroson, T. A., & Green, R. F. 1992, ApJS, 80, 109 Cardelli, J. A., Clayton, G. C., & Mathis, J. S. 1989, ApJ, 345, 245 Chornock, R., Berger, E., Gezari, S., et al. 2014, ApJ, 780, 44 Churazov, E., Sazonov, S., Sunyaev, R., et al. 2005, MNRAS, 363, L91 Dawson, K. S., Schlegel, D. J., Ahn, C. P., et al. 2013, AJ, 145, 10 Denney, K. D., De Rosa, G., Croxall, K., et al. 2014, ApJ, 796, 134 Di Matteo, T., Springel, V., & Hernquist, L. 2005, Nature, 433, 604 Eisenstein, D. J., Weinberg, D. H., Agol, E., et al. 2011, AJ, 142, 72 Eracleous, M., Livio, M., & Binette, L. 1995, ApJ, 445, L1 Gezari, S., Martin, D. C., Milliard, B., et al. 2006, ApJ, 653, L25 Gezari, S., Heckman, T., Cenko, S. B., et al. 2009, ApJ, 698, 1367 Gezari, S., Chorno ck, R., Rest, A., et al. 2012, Nature, 485, 217 Go odrich, R. W. 1995, ApJ, 440, 141 Greene, J. E., Peng, C. Y., & Ludwig, R. R. 2010, ApJ, 709, 937 Hainline, K. N., Hickox, R., Greene, J. E., Myers, A. D., & Zakamska, N. L. 2013, ApJ, 774, 145 Holoien, T. W.-S., Prieto, J. L., Bersier, D., et al. 2014, MNRAS, 445, 3263 Homan, J., & Belloni, T. 2005, Ap&SS, 300, 107 Hopkins, P. F., Hernquist, L., Martini, P., et al. 2005, ApJ, 625, L71 Ichimaru, S. 1977, ApJ, 214, 840 Ivezic, Z., Tyson, J. A., Ab el, B., et al. 2008, arXiv:0805.2366 Kauffmann, G., Heckman, T. M., Tremonti, C., et al. 2003, MNRAS, 346, 1055 Keel, W. C., Lintott, C. J., Schawinski, K., et al. 2012a, AJ, 144, 66 Keel, W. C., Cho jnowski, S. D., Bennert, V. N., et al. 2012b, MNRAS, 420, 878 Keel, W. C., Maksym, W. P., Bennert, V. N., et al. 2014, arXiv:1408.5159 Kelly, B. C., Vestergaard, M., Fan, X., et al. 2010, ApJ, 719, 1315

APPENDIX In addition to the three CL quasars yielded by our systematic search, we identified one additional possible CL quasar in our visual inspection (SDSS J132457.29+480241.2) that is classified as CLASS = `QSO' in its earlier SDSS spectrum, and CLASS = `GALAXY' in its latest SDSS spectrum (shown in Figure 6). However, the red side of the latest SDSS spectrum is corrupted beyond 5000° although the blue side appears to show the disappearance of broad H and H A, emission, along with dimming of the quasar continuum, similar to CL quasars. The fiber (527) of the latest SDSS spectrum is known to be affected by columns of bad pixels in the red camera of the BOSS spectrograph, which is likely to be the cause of the corrupted spectrum (and affects spectra from neighboring fibers on this and other plates). Nevertheless, to verify whether this ob ject is indeed a CL quasar, we obtained additional optical long-slit spectra using the Dual Imaging Spectrograph on the Astrophysical Research Consortium 3.5m telescope at Apache Point Observatory, with wavelength coverage of 3400 - 9200 ° Three 15 minute exposures where taken on 14 January A. 2015 UT, at a spectral resolution of R800 using the B400/R300 grating settings, and a 1.5 slit. The seeing was 1.8 on this night, and the observations were obtained at airmass of approximately 1.04. Spectra of the spectrophotometric standard star Feige 34 were also obtained for flux-calibration and removal of atmospheric absorption, and HeNeAr lamps were used to obtain a wavelength solution. These spectra were bias and flat-field corrected, wavelength- and flux-calibrated, and corrected for atmospheric extinction using standard IRAF procedures. The calibrated APO 3.5m spectrum is displayed in the lower panel of Figure 6, which shows that although the quasar continuum, broad H, and broad H emission has dimmed, they remain prominent, and thus we do not include this


Archival Search for Changing-Look Quasars

13

SDSS J132457.29+480241.2, z = 0.272 40 30 20 10 0 40 30 20 10 0 3500 4000 4500
wavelength [ ] erg cm-2 s-1
-1

SDSS spectrum, MJD = 52759

]

3500 spectrum, MJD = 56805 SDSS 4000 4500 5000

F [1 0

-1

w 57036 APO 3.5m spectrum, MJD = avelength [ ]

5500

6000

6500

7000

7

5000

5500

6000

6500

7000

Figure 7. SDSS and APO 3.5m rest-frame spectra of SDSS J132457.29+480241.2.

ob ject in our CL quasar sample. Although the disagreement between the APO 3.5m spectrum and the latest SDSS spectrum is likely due to the known column of bad pixels, a scenario in which this ob ject is a CL quasar that dimmed in the latest SDSS spectrum to a galaxy-like state, then rebrightened back to a quasar-like state in the APO 3.5m spectrum cannot be ruled out. Given the short period between the latest SDSS spectrum and APO spectrum (182 days in the rest-frame), and the known issues with the latest SDSS spectrum in the red side, rebrightening of a CL quasar appears unlikely, but additional spectral monitoring may be warranted.