Spectral Variability of a Sample of Extreme Variability Quasars and Implications for the MgII Broad-line Region

TL;DR

This study investigates the spectral variability of extreme variability quasars over several years, focusing on the Mg II broad-line region, revealing that Mg II flux correlates with continuum changes but its width does not, affecting black hole mass estimates.

Contribution

It provides the first detailed analysis of Mg II line variability in EVQs, highlighting the limitations of using Mg II width for black hole mass estimation due to luminosity-dependent biases.

Findings

Mg II flux varies with continuum, but line width remains mostly unchanged.

Mg II variability indicates reverberation to continuum changes.

Using Mg II width for black hole mass estimates introduces bias.

Abstract

We present new Gemini/GMOS optical spectroscopy of 16 extreme variability quasars (EVQs) that dimmed by more than 1.5 mag in the $g$ band between the Sloan Digital Sky Survey (SDSS) and the Dark Energy Survey (DES) epochs (separated by a few years in the quasar rest frame). The quasar sample covers a redshift range of $0.5 < z < 2.1$. Nearly half of these EVQs brightened significantly (by more than 0.5 mag in the $g$ band) in a few years after reaching their previous faintest state, and some EVQs showed rapid (non-blazar) variations of greater than 1-2 mag on timescales of only months. Leveraging on the large dynamic range in continuum variability between the earlier SDSS and the new GMOS spectra, we explore the associated variations in the broad Mg II,$\lambda2798$ line, whose variability properties have not been well studied before. The broad Mg II flux varies in the same direction as the continuum flux, albeit with a smaller amplitude, which indicates at least some portion of Mg II is reverberating to continuum changes. However, the width (FWHM) of Mg II does not vary accordingly as continuum changes for most objects in the sample, in contrast to the case of the broad Balmer lines. Using the width of broad Mg II to estimate the black hole mass therefore introduces a luminosity-dependent bias.

Figures

Tables

Table 1. Table 1: Variability of EVQs

Name	R.A.	Decl.	Redshift	SDSS	Gemini	ExpTime	$\Delta g_{\max }$	$\Delta g_{\mathrm{phot}}$	$\Delta g_{\mathrm{spec}}$	$\Delta g_{\mathrm{back}}$
					2018	(s)	(mag)	(mag)	(mag)	(mag)
J2159$+$0005	21:59:44.32	$+$00:05:27.8	0.936	2001-09-21	05-12	675$\times$4	2.590$\pm$0.068	2.207$\pm$0.058	2.442$\pm$0.041	$-$0.050$\pm$0.088
J0140$+$0052	01:40:27.89	$+$00:52:12.5	1.020	2004-11-06	09-05	675$\times$4	2.272$\pm$0.067	1.647$\pm$0.061	0.312$\pm$0.013	$-$0.078$\pm$0.086
J2213$-$0037	22:13:12.08	$-$00:37:25.6	2.063	2003-09-30	07-10	675$\times$4	2.258$\pm$0.096	1.585$\pm$0.086	1.148$\pm$0.063	$-$0.176$\pm$0.123
J2252$+$0004	22:52:50.73	$+$00:04:18.3	1.001	2001-09-26	07-10	675$\times$4	2.137$\pm$0.042	1.517$\pm$0.050	0.708$\pm$0.021	$-$0.121$\pm$0.062
J2217$+$0029	22:17:39.23	$+$00:29:04.4	1.643	2003-08-22	06-17	450$\times$2	1.953$\pm$0.038	1.388$\pm$0.037	0.974$\pm$0.027	$-$0.297$\pm$0.048
J2249$+$0047	22:49:24.01	$+$00:47:50.4	1.360	2001-09-26	06-17	675$\times$4	2.419$\pm$0.090	1.249$\pm$0.072	1.202$\pm$0.043	$-$0.344$\pm$0.110
J2228$-$0032	22:28:36.23	$-$00:32:02.9	1.032	2001-08-22	07-10	900$\times$2	1.800$\pm$0.033	1.188$\pm$0.029	0.871$\pm$0.039	$-$0.216$\pm$0.039
J2343$+$0038	23:43:07.38	$+$00:38:54.7	0.667	2002-09-08	07-14	450$\times$1	1.841$\pm$0.042	1.104$\pm$0.047	0.840$\pm$0.037	0.000$\pm$0.056
J2328$-$0053	23:28:38.04	$-$00:53:13.6	1.551	2001-10-17	07-10	675$\times$4	1.971$\pm$0.048	1.020$\pm$0.034	0.623$\pm$0.019	$-$0.859$\pm$0.054
J2338$-$0101	23:38:53.44	$-$01:01:19.4	1.483	2002-09-08	07-12	900$\times$2	2.024$\pm$0.056	0.949$\pm$0.052	0.635$\pm$0.025	$-$0.682$\pm$0.069
J2335$-$0049	23:35:17.82	$-$00:49:27.0	0.671	2001-10-17	07-12	450$\times$2	1.753$\pm$0.082	0.949$\pm$0.032	0.459$\pm$0.096	$-$0.319$\pm$0.081
J0140$-$0035	01:40:48.62	$-$00:35:00.9	1.383	2003-01-05	09-05	675$\times$4	2.255$\pm$0.068	0.863$\pm$0.045	0.115$\pm$0.014	$-$1.067$\pm$0.076
J2141$-$0016	21:41:30.61	$-$00:16:49.1	1.282	2004-08-10	05-12	675$\times$4	1.768$\pm$0.039	0.622$\pm$0.083	1.074$\pm$0.025	$-$1.035$\pm$0.088
J0048$-$0113	00:48:26.09	$-$01:13:10.7	1.034	2003-09-01	07-09	675$\times$4	1.695$\pm$0.052	0.479$\pm$0.040	0.453$\pm$0.018	$-$0.836$\pm$0.052
J2209$-$0055	22:09:08.24	$-$00:55:58.8	0.528	2000-09-01	05-12	450$\times$2	1.966$\pm$0.075	0.251$\pm$0.017	1.065$\pm$0.007	$-$1.715$\pm$0.076
J2350$+$0025	23:50:40.09	$+$00:25:58.9	1.062	2002-09-06	07-14	900$\times$2	2.101$\pm$0.078	0.141$\pm$0.018	$-$0.217$\pm$0.005	$-$1.890$\pm$0.078

Table 2. Table 2:  Spectral Fitting to the Two-Epoch Spectra

Name	MJD	logL(3000 Å)	logL(Mg II)	FWHM(Mg II)	logWλ(Mg II)	$M_{\mathrm{BH}}$
	(days)	(erg s-1)	(erg s-1)	(km s-1)	(Å)	(${10}^8{M}_{\odot }$)
J2159$+$0005	52173	45.09 $\pm$ 0.01	43.43 $\pm$ 0.02	9847 $\pm$ 590	1.76 $\pm$ 0.03	14.76 $\pm$ 1.74
	58250	44.21 $\pm$ 0.01	43.09 $\pm$ 0.02	10826 $\pm$ 1334	2.33 $\pm$ 0.02	5.03 $\pm$ 1.18
J0140$+$0052	53315	45.12 $\pm$ 0.01	43.18 $\pm$ 0.03	4384 $\pm$ 914	1.49 $\pm$ 0.03	3.05 $\pm$ 1.25
	58366	44.90 $\pm$ 0.00	43.35 $\pm$ 0.02	6104 $\pm$ 245	1.88 $\pm$ 0.02	4.33 $\pm$ 0.35
J2213$-$0037	52912	45.61 $\pm$ 0.04	43.91 $\pm$ 0.11	9633 $\pm$ 5678	1.77 $\pm$ 0.12	29.74 $\pm$ 31.98
	58309	44.99 $\pm$ 0.02	43.55 $\pm$ 0.02	7630 $\pm$ 671	1.96 $\pm$ 0.03	7.63 $\pm$ 1.30
J2252$+$0004	52178	45.24 $\pm$ 0.01	43.37 $\pm$ 0.03	7918 $\pm$ 1256	1.55 $\pm$ 0.03	11.85 $\pm$ 3.76
	58309	44.80 $\pm$ 0.00	43.25 $\pm$ 0.01	7979 $\pm$ 506	1.86 $\pm$ 0.01	6.38 $\pm$ 0.77
J2217$+$0029	52873	45.63 $\pm$ 0.01	43.86 $\pm$ 0.05	6500 $\pm$ 983	1.69 $\pm$ 0.05	13.92 $\pm$ 4.08
	58286	45.25 $\pm$ 0.00	43.55 $\pm$ 0.04	6431 $\pm$ 710	1.71 $\pm$ 0.03	7.90 $\pm$ 1.77
J2249$+$0047	52178	45.09 $\pm$ 0.01	43.45 $\pm$ 0.04	12230 $\pm$ 992	1.79 $\pm$ 0.04	22.73 $\pm$ 3.59
	58286	44.72 $\pm$ 0.00	43.39 $\pm$ 0.01	10863 $\pm$ 362	2.09 $\pm$ 0.01	10.51 $\pm$ 0.68
J2228$-$0032	52143	45.36 $\pm$ 0.00	43.69 $\pm$ 0.02	13727 $\pm$ 943	1.76 $\pm$ 0.02	42.08 $\pm$ 5.60
	58309	44.90 $\pm$ 0.01	43.59 $\pm$ 0.01	14021 $\pm$ 775	2.11 $\pm$ 0.01	22.81 $\pm$ 2.74
J2343$+$0038	52525	44.38 $\pm$ 0.01	43.13 $\pm$ 0.08	4907 $\pm$ 433	2.20 $\pm$ 0.07	1.32 $\pm$ 0.21
	58313	44.13 $\pm$ 0.01	42.72 $\pm$ 0.03	6009 $\pm$ 666	2.06 $\pm$ 0.05	1.40 $\pm$ 0.31
J2328$-$0053	52199	45.35 $\pm$ 0.02	43.74 $\pm$ 0.04	7468 $\pm$ 829	1.79 $\pm$ 0.05	12.28 $\pm$ 2.68
	58309	45.01 $\pm$ 0.00	43.56 $\pm$ 0.01	6230 $\pm$ 466	1.97 $\pm$ 0.01	5.28 $\pm$ 0.82
J2338$-$0101	52525	45.38 $\pm$ 0.01	43.51 $\pm$ 0.04	5090 $\pm$ 1117	1.56 $\pm$ 0.05	5.96 $\pm$ 2.49
	58311	45.19 $\pm$ 0.00	43.49 $\pm$ 0.03	6781 $\pm$ 929	1.73 $\pm$ 0.03	8.05 $\pm$ 2.09
J2335$-$0049	52199	44.53 $\pm$ 0.01	43.18 $\pm$ 0.05	6360 $\pm$ 919	2.10 $\pm$ 0.07	2.76 $\pm$ 0.81
	58311	44.39 $\pm$ 0.02	42.85 $\pm$ 0.05	5079 $\pm$ 715	1.87 $\pm$ 0.05	1.43 $\pm$ 0.40
J0140$-$0035	52644	45.22 $\pm$ 0.01	43.51 $\pm$ 0.03	4004 $\pm$ 935	1.71 $\pm$ 0.04	2.91 $\pm$ 1.44
	58366	45.12 $\pm$ 0.00	43.55 $\pm$ 0.04	4978 $\pm$ 1034	1.85 $\pm$ 0.04	3.94 $\pm$ 1.46
J2141$-$0016	53227	45.19 $\pm$ 0.01	43.62 $\pm$ 0.03	4844 $\pm$ 439	1.86 $\pm$ 0.03	4.08 $\pm$ 0.73
	58250	45.02 $\pm$ 0.00	43.36 $\pm$ 0.01	4643 $\pm$ 267	1.76 $\pm$ 0.01	2.98 $\pm$ 0.36
J0048$-$0113	52883	44.86 $\pm$ 0.01	43.26 $\pm$ 0.07	5000 $\pm$ 821	1.83 $\pm$ 0.08	2.75 $\pm$ 0.87
	58308	44.74 $\pm$ 0.00	43.18 $\pm$ 0.02	4671 $\pm$ 516	1.89 $\pm$ 0.02	2.01 $\pm$ 0.45
J2209$-$0055	51788	45.02 $\pm$ 0.00	43.37 $\pm$ 0.02	6184 $\pm$ 393	1.78 $\pm$ 0.02	5.24 $\pm$ 0.68
	58250	44.53 $\pm$ 0.01	43.00 $\pm$ 0.03	5643 $\pm$ 787	1.89 $\pm$ 0.03	2.16 $\pm$ 0.64
J2350$+$0025	52523	45.31 $\pm$ 0.01	43.62 $\pm$ 0.02	4968 $\pm$ 478	1.73 $\pm$ 0.02	5.11 $\pm$ 0.98
	58313	45.38 $\pm$ 0.00	43.69 $\pm$ 0.01	4521 $\pm$ 138	1.72 $\pm$ 0.01	4.71 $\pm$ 0.29

Equations

$$\Delta \mathrm{logL}(\mathrm{MgII})=(0.39\pm 0.07)\times \Delta \mathrm{logL}(3000\text{A˚}).$$

$$\Delta \mathrm{logL}(\mathrm{MgII})=(0.47\pm 0.05)\times \Delta \mathrm{logL}(3000\text{A˚}).$$

$$\log {\mathrm{W}}_{\lambda }(\mathrm{MgII})=(-0.47\pm 0.07)\times \mathrm{logL}(3000\text{A˚})+(23.01\pm 2.97),$$

$$\log {\mathrm{W}}_{\lambda }(\mathrm{MgII})=(-0.33\pm 0.03)\times \mathrm{logL}(3000\text{A˚})+(16.91\pm 1.31),$$

$$M_{\mathrm{BH}}=\frac{V^{2}R}{G}=f\frac{W^{2}R}{G},$$

$$\Delta \mathrm{logFWHM}(\mathrm{MgII})=(-0.013\pm 0.030)\Delta \mathrm{logL}(3000\text{A˚}).$$

$$\Delta \mathrm{logFWHM}(\mathrm{MgII})=(0.012\pm 0.012)\Delta \mathrm{logL}(3000\text{A˚}).$$

$$\Delta \mathrm{logFWHM}(\mathrm{MgII})=(-0.016\pm 0.030)\Delta \mathrm{logL}(3000\text{A˚})$$

$$\Delta \mathrm{logFWHM}(\mathrm{MgII})=(0.019\pm 0.012)\Delta \mathrm{logL}(3000\text{A˚})$$

Full text

††pubyear: 2019††pagerange: Spectral Variability of a Sample of Extreme Variability Quasars and Implications for the Mg ii Broad-line Region–Spectral Variability of a Sample of Extreme Variability Quasars and Implications for the Mg ii Broad-line Region

1 Introduction

The canonical unification model of AGN dictates that Type 2 objects (with only narrow emission lines) are drawn from the same underlying population as Type 1 objects (with both broad and narrow emission lines), but the AGN continuum and broad-line emission is obscured by a dust torus (Antonucci, 1993; Urry & Padovani, 1995). However, this static classification scheme is challenged by an increasingly large body of discoveries of quasars that apparently change spectral types on multi-year timescales (e.g., Denney et al., 2014; LaMassa et al., 2015; MacLeod et al., 2016; McElroy et al., 2016; Ruan et al., 2016; Runnoe et al., 2016; Gezari et al., 2017; Yang et al., 2018; Stern et al., 2018; MacLeod et al., 2019), mostly from recent multi-epoch imaging and spectroscopic surveys of quasars. The broad Balmer emission lines, including H$\alpha$, H$\beta$, and H$\gamma$, and broad helium lines were observed to have dramatically changed between the dim and bright epochs, even completely disappearing or emerging, following large-amplitude variability in the continuum. This population of changing-look quasars (CLQs) challenges the unified model of AGN, and is difficult to understand in the standard accretion disk theory given the observed short timescale of the changes. They may have profound implications for accretion physics (Lawrence, 2018).

The continuum radiation of quasars is observed to vary typically by 0.2 mag on timescales of months to years (e.g., Vanden Berk et al., 2004; Wilhite et al., 2005; Sesar et al., 2007; Schmidt et al., 2010; MacLeod et al., 2012; Morganson et al., 2014). Rumbaugh et al. (2018) found that approximately 10% of quasars can vary by $>1$ mag, dubbed extreme variability quasars (EVQs), over an observed baseline of $\sim 15$ years. Because the broad emission lines are presumably powered by the ionizing continuum, they will reverberate to the continuum variability on the BLR light-travel timescales of days to weeks, which is much shorter compared to the BLR dynamical timescale of the order of a few years. Thus the study of the EVQ population is not only useful to understand accretion physics in the context of extreme variability, but also important to characterize the response of the broad lines to the extreme continuum variability, which in turn will shed light on the kinematics and structure of the BLR.

While there have been numerous variability studies on the broad H$\beta$ line with, e.g., reverberation mapping (RM) data, similar studies on the broad Mg ii line have been sparse (e.g., Trevese et al., 2007; Woo, 2008; Hryniewicz et al., 2014; Cackett et al., 2015; Sun et al., 2015; Zhu et al., 2017). However, the broad Mg ii line is an important line, as it is used for RM (Clavel et al., 1991; Reichert et al., 1994; Metzroth et al., 2006; Shen et al., 2016) and single-epoch black hole mass estimation of quasars at redshift $z>1$ (e.g., McLure & Jarvis, 2002; McLure & Dunlop, 2004; Shen et al., 2008, 2011; Wang et al., 2009), and thus it is important to understand the variability properties of broad Mg ii as the continuum varies.

Assuming that the BLR is virialized, the BH mass is determined by the BLR size and the virial velocity (e.g., Wandel et al., 1999). Practically, the width of broad emission line is used as an indicator of the virial velocity assuming that the line is Doppler broadened. For individual object, the inner gas orbits faster in the virial theorem. Under these assumptions, the broad line width should increase (or decrease) as the BLR contracts (or expands) in respond to a decrease (or increase) in the continuum (i.e., the “breathing” of the BLR). An anti-correlation between the broad line width and the BLR size (proportional to the continuum luminosity) is expected, and has been seen in broad H$\beta$ (e.g., Park et al., 2012; Shen, 2013). However, the situation is much less clear for Mg ii and C iv (Shen, 2013). The full width at half maximum (FWHM) of broad Mg ii is strongly correlated with that of broad H$\beta$ (e.g., Shen et al., 2008, 2011; Wang et al., 2019), but the range of dispersion seen in the width of Mg ii is smaller than that for H$\beta$ for the same objects (Shen et al., 2011), indicating that Mg ii is less variable (e.g., Sun et al., 2015) and possibly has slightly different kinematic structure than broad H$\beta$. In addition, there is a population of galaxy spectra with broad Mg ii but no broad H$\beta$ (e.g., Roig et al., 2014). All these observations suggest that broad Mg ii is somewhat different, and a systematic study of Mg ii is required to understand its phenomenology and physics.

Motivated by the questions about how EVQs vary and how their broad emission lines vary accordingly, we select a sample of EVQs that vary by more than 1.5 mag in the $g$ band, using data from DES and SDSS. We obtain new Gemini/GMOS optical spectroscopy of 16 EVQs, and compare with earlier SDSS spectra. Our EVQ sample covers a redshift range of $0.5<z<2.1$ with Mg ii coverage in both SDSS and GMOS spectra for all our targets, providing a unique opportunity to study Mg ii variability leveraged on the extreme continuum variability. Compared with previous works on CLQs, most of our targets are at higher redshifts and probe higher quasar luminosities.

In this work we present our new GMOS spectroscopy and study the spectral variability of these 16 EVQs between the SDSS epoch and the more recent GMOS epoch. The paper is organized as follows. Section 2 describes the imaging data from multiple sources and SDSS spectroscopic data. Section 3 outlines the target selection, spectroscopic observations using Gemini/GMOS, and spectral fitting procedures. We present our results in Section 4 and discuss the implications in Section 5. We summarize in Section 6. We use a $\Lambda$CDM cosmology with parameters $\Omega_{\Lambda}=0.7$, $\Omega_{\rm m}=0.3$, and $H_{0}=70$ km s*-1* Mpc*-1*.

2 Data

2.1 Imaging Data

The Dark Energy Survey (DES) observed 5000 deg2 of the sky in five filters ($grizY_{\rm DES}$), using a wide-field camera (DECam) on the 4-m Blanco Telescope (Flaugher et al., 2015). The 10$\sigma$ single-epoch PSF magnitude limit in the five $grizY$ bands are 23.57, 23.34, 22.78, 22.10, and 20.69, respectively (Abbott et al., 2018). We used five years DES data starting from 2013 (Abbott et al., 2018; Diehl et al., 2018). For the preliminary year 5 data, we applied the zero-point calibration with an accuracy of $\sim 0.01-0.02$ mag.

SDSS mapped the sky in five filters ($ugriz_{\rm SDSS}$) using a 2.5 m telescope (Gunn et al., 2006) at the Apache Point Observatory (Abazajian et al., 2009), covering 11,663 deg2 of the sky. SDSS also repeatedly imaged a $120^{\circ}\times 2.5^{\circ}$ stripe along the celestial equator centered at zero declination in the Southern Galactic Cap (“Stripe 82") from 1998 to 2007. The frequency was increased from 2005 to 2007 for the supernova survey (Frieman et al., 2008). Over the 10 year duration of the program, there are, on average, more than 60 epochs in the Stripe 82 region.

The Pan-STARRS1 (PS1; Chambers et al., 2016) survey mapped three-quarters of the sky in five broadband filters ($grizy_{\rm PS1}$), using a 1.8-m telescope with a 1.4 Gigapixel camera. PS1 data were obtained from 2011 to 2014, filling the gap between the SDSS and DES. We used multi-epoch photometry from the detection catalog in the PS1 Data Release 2 (DR2). For quasars, the magnitude offset between the SDSS and PS1 (DES) $g$ bands is negligible, between $-0.053$($-$0.065) and 0.005(0.008) mag at redshift $z<2$ (Yang et al., 2018).

2.2 SDSS Spectroscopic Data

We used the SDSS spectra as the earlier epoch comparison spectra. The SDSS spectroscopy from the SDSS-I/II covers a wavelength range from 3800 to 9200 Å at an average resolution of $R=\lambda/\Delta\lambda\sim 2000$ (Abazajian et al., 2009), ranging from $R\sim$ 1500 at 3800 Å to $R\sim$ 2500 at 9000 Å.

3 Methods

3.1 Target Selection

Given the large overlap between DES and SDSS within the Stripe 82 region, we carried out a systematic search for EVQs from 9,258 quasars in the Stripe 82 region (MacLeod et al., 2012), with improved calibration (Ivezić et al., 2007; Sesar et al., 2007) and detailed spectral measurements from their SDSS spectra (Shen et al., 2011). Combining the SDSS Stripe 82 and DES light curves, we selected quasars with $g$-band variability larger than 1.5 mag. For the GMOS followup, we only used DES data up to Y3 (2016) at the time of the proposal. However, in this work we use all the available DES data up to Y5 to construct photometric light curves. In selecting EVQs from the combined SDSS and DES photometric light curves, we rejected points that deviate from the running median of a $\pm$100-day window by more than 0.5 mag, and noisy points with magnitude uncertainties larger than 0.15 mag. This search resulted in 146 quasars. We imposed additional criteria: (1) the quasars were restricted to be those that were in the faint state during the DES epochs; (2) the latest DES epoch (as in Y3 data at the time of the Gemini proposal) is brighter than 22 mag in $i$-band for reasonably good Gemini spectroscopy; (3) the object did not show frequent large-amplitude ($>0.5$ mag) and rapid (shorter than a month) optical variability, which may be due to blazar activity; (4) it is not a broad absorption line quasar according to the SDSS spectrum, thus excluding objects in which the variability is caused by outflows; (5) we rejected objects with poor or problematic SDSS spectra after visual inspection. We finally selected 27 EVQs, all of which have relatively smooth light curves.

3.2 Spectroscopic Observations

We observed 16 of the 27 EVQs during the Gemini 2018A run from May to September, using the Gemini GMOS-South spectrograph (summarized in Table 1). The targets were observed with the R150 grating and a 0.5$\arcsec$ slit. We choose this configuration to balance the needs for wavelength coverage, spectral resolution and sensitivity. To mitigate the effect of CCD gaps on coverage, we split the exposures (ExpTime in Table 1) into pairs with central wavelengths of 5800 Å and 6000 Å, respectively. The coadded spectra cover a wavelength range of 4000 – 10200 Å with a spectral resolution of $R\sim 630$. The impact of the spectral resolution difference between the GMOS and SDSS spectra is negligible, leading to less than 0.7% difference in line width measurements when ${\rm FWHM}>4000\ {\rm km\,s^{-1}}$ as in the case for the quasars we observed.

The spectra were reduced using standard IRAF111 IRAF is the Image Reduction and Analysis Facility, written and supported by the National Optical Astronomy Observatories (NOAO) operated by the Association of Universities for Research in Astronomy (AURA) under cooperative agreement with the National Science Foundation. routines (Tody, 1986, 1993). We used two standard stars (LTT9239 or LTT7987) for our flux calibration. For the 12 objects observed before July 14, the standard stars were observed during the same night of our science observations. For the 4 objects observed on July 14 and September 5, including J2343$+$0038, J2350$+$0025, J0140$-$0035, and J0140$+$0052, we use the standard star LTT9239 observed on July 12. These four objects, as well as the standard star, were observed on photometric nights.

3.3 Spectral Fitting

We fit the spectra using the spectral fitting code from Shen et al. (2018), which models the quasar continuum, broad Fe ii emission, and emission line components. The spectra were fit in the rest-frame of the quasar after correcting for Galactic reddening using the dust map in Schlegel et al. (1998) and the extinction curve from Cardelli et al. (1989). The continuum includes a power-law continuum and a positive 3rd-order polynomial accounting for dust reddening internal to the quasar. We used empirical UV Fe ii emission templates from the literature (Vestergaard & Wilkes, 2001; Tsuzuki et al., 2006; Salviander et al., 2007) covering from 1000 Å to 3500 Å, and an optical Fe ii template (3686-7484 Å) from Boroson & Green (1992). We fitted the Fe ii emission with three free parameters: normalization, broadening, and wavelength shift. We first fit the continuum and Fe ii emission together, choosing a few continuum windows. For instance, to avoid the influence of Mg ii emission, we masked the region between 2675 Å and 2925 Å. Then, we subtracted the continuum and Fe ii emission components, and fitted the other emission lines. Following Shen et al. (2018), we used three (one) Gaussian components for the broad (narrow) H$\beta$ line, and two Gaussian components for broad Mg ii. In addition, we used two Gaussian components for the narrow Mg ii $\lambda\lambda$2796, 2803 Å doublet (Wang et al., 2009). To quantify the measurement uncertainties, we used a Monte Carlo approach by adding a random Gaussian deviate to the flux at each pixel, with the Gaussian $\sigma$ equal to the spectral error at that pixel. We measured the continuum/line properties and their uncertainties using the median value and the semi-amplitude of the range enclosing the 16th and 84th percentiles of the distribution from 50 trials. Our fitting results show that the narrow emission line luminosity is consistent between the SDSS and GMOS spectra, as the narrow line flux is expected to remain constant on multi-year timescales for our quasars. For example, the narrow [O II] $\lambda$3728 luminosity for J2159$+$0005 is $10^{42.34\pm 0.05}$ and $10^{42.30\pm 0.01}$ erg s*-1* from the SDSS and GMOS spectra, indicating that our flux calibration is reasonably accurate.

4 Results

We summarize the photometric variability of these EVQs in Table 1. $\Delta g_{\rm max}$ is the maximum variability in the $g$ band from the SDSS, PS1, and DES observations. All of them have $\Delta g_{\rm max}$ larger than 1.5 mag. $\Delta g_{\rm phot}$ is the $g$-band magnitude difference between imaging photometry closest to the Gemini and SDSS spectra epochs. The majority of them rebrightened since they were intially selected in the first 3 years of DES (see their light curves in the left panels of Figure 1 and Figure A.1). There are only four with $\Delta g_{\rm phot}>1.5$ mag, and 9 with $\Delta g_{\rm phot}>1.0$ mag. $\Delta g_{\rm spec}$ is the magnitude difference between the Gemini and SDSS spectra convolved with the SDSS $g$-band filter curve. $\Delta g_{\rm back}$ is the $g$-band magnitude difference between the latest DES imaging epoch and the faintest epoch in earlier DES data, describing how much they have brightened since. Among them, 9 EVQs did not vary much with $-0.3\leq\Delta g_{\rm back}\leq 0$, and 7 EVQs brightened with $-1.7<\Delta g_{\rm back}<-0.6$. Hence nearly half of them have significantly brightened (by more than 0.5 mag) since the faintest epoch, which is consistent with the findings in MacLeod et al. (2019). The light curves show that the extreme variability mostly occurs over a few years, but can also occur on much shorter timescale. For example, J0140$+$0052 brightened by $\sim 2$ mag from the GMOS spectral epoch to the last DES epoch (less than one year in the observed frame), and J2209$-$0055 brightened by $\sim 1$ mag and then again dimmed by $\sim 1$ mag from 2016 to 2018. This is consistent with the type transition timescale of CLQs spanning from 0.9 to 13 years in Yang et al. (2018).

J2159$+$0005 is an EVQ that has been dimming continuously since 2001 as seen from its light curves in the $g$ and $r$ bands (shown in Figure 1). The latest DES photometry of J2159$+$0005 is $>2$ mag fainter than the earlier SDSS photometry in 2001. The right panel in Figure 1 shows that its continuum became much fainter and redder. Figure 2 shows our fitting results around Mg ii and H$\beta$. While there is no detectable broad H$\beta$ flux in the GMOS spectrum, broad Mg ii is still visible (albeit at reduced flux; see below). The narrow Mg ii emission on top of the broad-line emission is more distinctive in the faint state spectrum. We witness a normal blue quasar changing to a quasar with a broad Mg ii emission line but no broad H$\beta$ emission line, which bears some similarities to the objects presented in Roig et al. (2014), and similar to those CLQs presented in MacLeod et al. (2019). Although the SDSS spectrum does not cover H$\beta$, this quasar is possibly a CLQ with the highest redshift of $z=0.936$ known to date. It is plausible that some broad-line quasars at redshift $z>1$, where H$\beta$ moves out of the optical window, with a red continuum are similar to J2159$+$0005 seen at redshift $z<1$.

EVQ J2343$+$0038 is a CLQ at $z=0.667$, based on the behaviour of H$\beta$. It has broad H$\beta$ emission with ${\rm L_{H\beta}}=10^{42.32\pm 0.09}$ erg s*-1* and ${\rm FWHM_{H\beta}=4068\pm 1980}$ km s*-1* in the bright state from the SDSS spectrum. In the dim state, in the GMOS spectrum, the broad H$\beta$ emission has disappeared and we can clearly see H$\beta$ in absorption instead.

The light curves and spectra of the other 14 EVQs are listed in the Appendix. H$\beta$ is available in eight GMOS spectra, in four of them H$\beta$ is broad in the faint state (J0140$+$0052, J0048$-$0113, J2209$-$0055, and J2350$+$0025), in one it is narrow (J2159$+$0005), in one it is absorption (J2343$+$0038), and in other two H$\beta$ is absent (J2252$+$0004 and J2228$-$0032). The H$\beta$ and [O III] lines in J2252$+$0004 and J2228$-$0032 are marginally present probably because these lines lie beyond $10000$ Å, where the signal-to-noise ratio (S/N) is too low to verify the existence of these lines. H$\beta$ is only available in three of the SDSS spectra, including J2343$+$0038, J2335$-$0049, and J2209$-$0055. Unfortunately, H$\beta$ of J2335$-$0049 lands within a CCD gap, and we accidentally failed to obtain one exposure for it with the central wavelength at 6000 Å. The H$\beta$ emission of J2209$-$0055 is broad in both epochs, and its light curves show that it brightened before its GMOS spectrum was taken. The continuum of J2141$-$0016, at $\lambda<2200$ Å in the rest-frame, becomes much redder when it fades, similar to cases in Guo & Gu (2016); Ross et al. (2018). But the S/N of the GMOS spectra near 4000 Å is low, and a higher quality spectrum is needed to confirm this result.

Broad Mg ii line remains visible in all 16 EVQs in both bright and faint epochs. We fit the SDSS and GMOS spectra in detail to analyse their continuum and emission line properties, and explore Mg ii variability in §4.2 and §4.3.

4.1 A Supplemental Sample of EVQs

To increase sample statistics, we add a supplemental sample of EVQs with multi-epoch spectra covering Mg ii from the SDSS quasar catalog222https://data.sdss.org/sas/dr14/eboss/qso/DR14Q/DR14Q_v4_4.fits in the 14th Data Release (DR14Q, Pâris et al., 2018). The selection criteria are (1) the quality flag “zWarning" in the quasar catalog is 0; (2) the median spectral S/N per pixel is higher than 2; (3) the spectrum covers Mg ii; (4) the median spectral S/N around Mg ii and the continuum at rest-frame 3000 Å, specifically from 2700 to 3100 Å, is higher than 5 to ensure the good quality of our spectral fitting; (5) multi-epoch spectra of the same quasar must all satisfy the criteria from (1) to (4); and (6) for the same quasar, the continuum flux ratio at rest-frame 3000 Å between the brightest and faintest epochs from SDSS is larger than 3, which corresponds to a magnitude difference of 1.2 mag. These criteria result in 33 EVQs at $0.44<z<2.33$ selected from SDSS spectroscopy alone, nearly triple our EVQ sample with multi-epoch spectra.

4.2 Mg ii Line Variation

Our EVQ sample reveals that the broad Mg ii line does vary in the same direction as the continuum. We summarize the spectral fitting of the main EVQs sample in Table 2. Figure 3 shows the variability of broad Mg ii line luminosity between the two spectroscopic epochs, $\Delta{\rm logL(MgII)}$, as a function of the 3000 Å continuum luminosity variability between the same two epochs, $\Delta{\rm logL(3000\text{\AA})}$ (in the left panel). The filled black dots are the 16 EVQs selected from SDSS and DES and followed-up with Gemini, and the open gray circles represent the additional EVQs selected from SDSS. The broad Mg ii line luminosity increases when the continuum luminosity increases, as naively expected from photoionization. We perform least-square fits to the main EVQ sample of 16 objects, forcing the line to cross $[0,0]$. We obtain

[TABLE]

The result for the combined sample, including the main sample and the supplemental sample, is

[TABLE]

The variation of broad Mg ii is indeed smaller than that in the continuum, but still echos the continuum variation to some extent. This result suggests that at least some part of broad Mg ii reverberates to continuum changes, or that the continuum flux that ionizes broad Mg ii varies less than the continuum at rest-frame 3000Å. It is interesting to note that Bruce et al. (2017) found that Mg ii flux does not seem to respond to continuum variations in two extreme variability quasars, where the extreme variability can be caused by rare, high-amplitude microlensing events.

The weaker Mg ii variations relative to the nearby continuum variations also manifests as an anti-correlation between the equivalent width ($W_{\lambda}$ hereafter) of Mg ii, ${\rm logW_{\lambda}(MgII)}$, with continuum luminosity, ${\rm logL(3000\text{\AA})}$, which is the well-known Baldwin effect (e.g., Baldwin, 1977; Green et al., 2001). In the right panel of Figure 3, we fit linear regressions and obtain

[TABLE]

for the main sample, and

[TABLE]

for the combined sample. For the same object, the $W_{\lambda}$ increases when the continuum becomes fainter (listed in Table 2). For the most variable EVQ, J2159$+$0005, in this sample, the log$W_{\lambda}{\rm(MgII)}$ increases from $10^{1.76\pm 0.03}$ to $10^{2.33\pm 0.02}$ Å when the continuum ${\rm L(3000\text{\AA})}$ dims from $10^{45.9\pm 0.01}$ to $10^{44.21\pm 0.01}$ erg s*-1*. The variable $W_{\lambda}$ indicates that the dramatic decrease in continuum and Mg ii is not caused by variable dust reddening.

4.3 Mg ii Line Width

In contrast to the flux, we found that the width (in terms of FWHM) of the broad Mg ii line barely changes with luminosity for the bulk of the sample. For J2159$+$0005, which has the most dramatic change in continuum flux, the broad Mg ii component has a FWHM of $9881\pm 693$ km s*-1* for the bright state and a FWHM of $10858\pm 579$ km s*-1* for the faint state, resulting in a change of FWHM of $977\pm 903$ km s*-1*. Assuming that the gas clouds emitting broad Mg ii are virialized, the BH mass is

[TABLE]

where $V$ is the virial velocity, $R$ is the BLR radius, $f$ is a scale factor that accounts for the orientation, kinematics, and structure of the BLR, and $W$ is the width of the broad line assuming that the broad line is Doppler broadened by virial motion. If the BLR size increases as continuum luminosity increases, as observed for broad H$\beta$ as ${R\propto L^{0.5}}$ as found in local RM studies or based on the simple photoionization argument, then a relation between the EW and luminosity of the form $\Delta{\rm logW}=-0.25\Delta{\rm logL}$ is expected. Figure 4 shows the changes of broad Mg ii FWHM between the bright and faint epochs, $\Delta{\rm logFWHM(MgII)}$, as a function of $\Delta{\rm logL(3000\text{\AA})}$, the corresponding changes of the continuum luminosity. Our results show that the broad Mg ii width does not vary as expected from the simple virial relation above, in sharp contrast to the behaviour of broad H$\beta$ (e.g., Park et al., 2012; Shen, 2013). Fitting a linear regression model to our main EVQ sample, we obtain

[TABLE]

For the combined EVQ sample we obtain

[TABLE]

In both cases the broad Mg ii FWHM remains more or less constant despite the large changes in the continuum luminosity, which is consistent with the findings in Shen (2013).

To eliminate a possibility that variable fitting of Fe ii emission compromises the line width variation of Mg ii, we do a test by fixing the Fe ii parameters for the two epochs of the same object. We get

[TABLE]

for the main EVQ sample, and

[TABLE]

for the combined sample. The result shows that the line width of Mg ii barely changes, irrespective of our approach.

We discuss the potential causes of the non-variable broad Mg ii FWHM and its implications for virial BH mass estimation using Mg ii in §5.

5 Discussion

The coordinated variations in the continuum flux and in the broad Mg ii flux suggest that at least part of the Mg ii is photoionized by the continuum. This result offers some support to use the RM technique to infer the distances of the Mg ii broad-line clouds. However, the associated Mg ii variability is reduced compared to the level of continuum changes, indicating some part of the Mg ii line might not be photoionized, thus complicating the interpretation of the Mg ii RM results. In addition, the weaker Mg ii variability, as observed for a large sample of quasars in a dedicated RM program (Sun et al., 2015), makes it generally more difficult to detect a RM lag for the broad Mg ii line.

On the other hand, for the bulk of EVQs in our sample, the FWHM of broad Mg ii does not vary with luminosity as expected from the simple virial assumption, in contrast to the case of broad H$\beta$. This result suggests that the single-epoch virial BH mass based on Mg ii, which depends both on the continuum luminosity and single-epoch Mg ii width, will have an additional scatter in individual objects when luminosity varies while Mg ii width remains the same. To demonstrate this point, we estimate single-epoch Mg ii-based black hole masses using the bright and faint-state spectra and the recipe from McLure & Dunlop (2004, listed in Table 2). Apart from some quasars with small continuum changes (since some EVQs brightened again in our late GMOS spectroscopy) and some quasars with bad spectral quality, the black hole mass estimates in the bright and faint states are inconsistent (at $>1\sigma$ of measurement errors) for 8 of the EVQs in our main sample, including J2159$+$0005, J2252$+$0004, J2249$+$0047, J2228$-$0032, J2328$-$0053, J2335$-$0049, J2141$-$0016, J2209$-$0055. For example, the black hole mass estimates for J2159$+$0005 are $(1.48\pm 0.17)\times 10^{9}M_{\odot}$ in the bright state and $(5.03\pm 1.18)\times 10^{8}M_{\odot}$ in the dim state. These two black hole masses are inconsistent at $>4\sigma$ confidence level. Using the width of broad Mg ii to estimate the black hole mass therefore introduces a luminosity-dependent bias (see detailed discussion in Shen, 2013) with the single-epoch mass technique that is most severe for quasars that undergo significant luminosity changes. For individual quasars, clearly the $R-L$ relation is different or even absent for broad Mg ii, although there may still be a global $R-L$ relation for quasars over a broad range of luminosity. The latter case will justify the use of Mg ii as an single-epoch mass estimator for the general population of quasars (albeit still suffering from the luminosity-dependent scatter in individual objects), and the existence of such a global $R-L$ relation for broad Mg ii can be tested with upcoming RM results on large quasar samples (e.g., Shen et al., 2015; Hoormann et al., 2019).

The so-called “breathing" effect of the BLR, an increase in ionizing continuum luminosity leads to an increase in BLR radius and hence a decrease in line width, is observed conclusively for H$\beta$ (Korista & Goad, 2004; Cackett & Horne, 2006; Denney et al., 2009; Park et al., 2012; Barth et al., 2015; Runco et al., 2016). Different from H$\beta$ as a recombination line, Mg ii are resonance lines. The greater optical depth for Mg ii, which results in a large number of scatterings before escape from the BLR, may cause the Mg ii photons to be emitted over a larger radius than H$\beta$ (Korista & Goad, 2004). Thus the variation in ionizing continuum may be diluted at a larger radius, leading to a smaller amplitude of Mg ii variability. The lack of a clear intrinsic $R-L$ relation for broad Mg ii could mean that the broad Mg ii emission is confined in a narrower range of radii than the broader distribution of gas that can potentially produce H$\beta$ emission. The transition layer between the H II zone and the H I zone, where most of the continuum photons between 13.6 and 55 eV are absorbed and a large portion of the line emission is produced, is geometrically thin (Krolik, 1999). This layer could emit a larger fraction of the total Mg ii emission than it does for the total Balmer-line emission (e.g., Kwan & Krolik, 1981). Thus when the ionizing continuum changes, the flux of Mg ii echos, but the radius of the gas emitting Mg ii barely changes, leading to more or less constant broad Mg ii width. Another possibility is that part of gas emitting Mg ii is virialized and photoionized, while the other part is not. It has been shown that Mg ii profile is complex (Kovačević-Dojčinović & Popović, 2015; Jonić et al., 2016), and can be affected by non-virial effects, such as outflows (León-Tavares et al., 2013; Popović et al., 2019), especially for very broad Mg ii (${\rm FWHM}>6000\ {\rm km\ s^{-1}}$). Potential detailed shape change of Mg ii profile can be analysed with dedicated Mg ii RM data and used to test these scenarios.

6 Summary

We have presented Gemini/GMOS spectroscopy of 16 EVQs, with $>1.5$ mag maximum variability in the $g$ band from the combined DES and SDSS photometric light curves (Rumbaugh et al., 2018), and studied their spectral changes between the GMOS spectra and the earlier SDSS spectra.

Our main results are:

(1) About half of the EVQs brightened again a few years since reaching the minimum flux. This is consistent with the conclusions in Rumbaugh et al. (2018) that extreme variability events are common among the general population of quasars on multi-year timescales, as well as the findings in the recent spectroscopic work by MacLeod et al. (2019). The extreme variability can also occur, albeit much less frequently, on much shorter timescales of $<1$ year, consistent with the findings in previous studies (Gezari et al., 2017; Yang et al., 2018; Trakhtenbrot et al., 2019).

(2) We identified two quasars that apparently changed their spectroscopic appearance. J2159$+$0005 is a possible CLQ. If confirmed, it would be the most distant CLQ so far discovered at $z=0.936$. J2343$+$0038 is a CLQ, at $z=0.667$, based on the disappearance of of H$\beta$.

(3) The Mg ii broad line flux varies accordingly as the continuum changes, albeit with a smaller amplitude than that for the continuum, $\Delta{\rm logL(MgII)}=(0.39\pm 0.07)\times\Delta{\rm logL(3000\text{\AA})}$. This suggests that at least part of the broad Mg ii emission is photoionized, or the continuum flux that ionizes broad Mg ii varies less than the continuum at rest-frame 3000Å.

(4) The Mg ii broad line FWHM remains roughly constant for the bulk of the objects, even though the continuum luminosity changed significantly, in contrast to the properties of broad H$\beta$. We discussed potential causes of the different kinematic properties of the broad Mg ii line, but to confirm these speculations require future dedicated Mg ii RM data. Nevertheless, the different variability properties of broad Mg ii compared to broad H$\beta$ provide some cautionary notes on the use of Mg ii as a single-epoch BH mass estimator. In particular, the absence of a clear intrinsic $R-L$ relation for Mg ii will inevitably lead to a luminosity-dependent bias in the BH mass estimates in individual systems as the continuum luminosity varies significantly.

Acknowledgements

QY and YS acknowledge support from an Alfred P. Sloan Research Fellowship (YS) and NSF grant AST-1715579. We thank Patrick Hall, Tamara Davis, Shu Wang, and Hengxiao Guo for useful discussions and suggestions.

Funding for the DES Projects has been provided by the U.S. Department of Energy, the U.S. National Science Foundation, the Ministry of Science and Education of Spain, the Science and Technology Facilities Council of the United Kingdom, the Higher Education Funding Council for England, the National Center for Supercomputing Applications at the University of Illinois at Urbana-Champaign, the Kavli Institute of Cosmological Physics at the University of Chicago, the Center for Cosmology and Astro-Particle Physics at the Ohio State University, the Mitchell Institute for Fundamental Physics and Astronomy at Texas A&M University, Financiadora de Estudos e Projetos, Fundação Carlos Chagas Filho de Amparo à Pesquisa do Estado do Rio de Janeiro, Conselho Nacional de Desenvolvimento Científico e Tecnológico and the Ministério da Ciência, Tecnologia e Inovação, the Deutsche Forschungsgemeinschaft and the Collaborating Institutions in the Dark Energy Survey.

The Collaborating Institutions are Argonne National Laboratory, the University of California at Santa Cruz, the University of Cambridge, Centro de Investigaciones Energéticas, Medioambientales y Tecnológicas-Madrid, the University of Chicago, University College London, the DES-Brazil Consortium, the University of Edinburgh, the Eidgenössische Technische Hochschule (ETH) Zürich, Fermi National Accelerator Laboratory, the University of Illinois at Urbana-Champaign, the Institut de Ciències de l’Espai (IEEC/CSIC), the Institut de Física d’Altes Energies, Lawrence Berkeley National Laboratory, the Ludwig-Maximilians Universität München and the associated Excellence Cluster Universe, the University of Michigan, the National Optical Astronomy Observatory, the University of Nottingham, The Ohio State University, the University of Pennsylvania, the University of Portsmouth, SLAC National Accelerator Laboratory, Stanford University, the University of Sussex, Texas A&M University, and the OzDES Membership Consortium.

Based in part on observations at Cerro Tololo Inter-American Observatory, National Optical Astronomy Observatory, which is operated by the Association of Universities for Research in Astronomy (AURA) under a cooperative agreement with the National Science Foundation.

The DES data management system is supported by the National Science Foundation under Grant Numbers AST-1138766 and AST-1536171. The DES participants from Spanish institutions are partially supported by MINECO under grants AYA2015-71825, ESP2015-66861, FPA2015-68048, SEV-2016-0588, SEV-2016-0597, and MDM-2015-0509, some of which include ERDF funds from the European Union. IFAE is partially funded by the CERCA program of the Generalitat de Catalunya. Research leading to these results has received funding from the European Research Council under the European Union’s Seventh Framework Program (FP7/2007-2013) including ERC grant agreements 240672, 291329, and 306478. We acknowledge support from the Brazilian Instituto Nacional de Ciênciae Tecnologia (INCT) e-Universe (CNPq grant 465376/2014-2).

This manuscript has been authored by Fermi Research Alliance, LLC under Contract No. DE-AC02-07CH11359 with the U.S. Department of Energy, Office of Science, Office of High Energy Physics. The United States Government retains and the publisher, by accepting the article for publication, acknowledges that the United States Government retains a non-exclusive, paid-up, irrevocable, world-wide license to publish or reproduce the published form of this manuscript, or allow others to do so, for United States Government purposes.

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 website 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, University of Tokyo, University of Utah, Vanderbilt University, University of Virginia, University of Washington, and Yale University.

The PS1 has been made possible through contributions by the Institute for Astronomy, the University of Hawaii, the Pan-STARRS Project Office, the Max-Planck Society and its participating institutes, the Max Planck Institute for Astronomy, Heidelberg and the Max Planck Institute for Extraterrestrial Physics, Garching, The Johns Hopkins University, Durham University, the University of Edinburgh, Queen’s University Belfast, the Harvard-Smithsonian Center for Astrophysics, the Las Cumbres Observatory Global Telescope Network Incorporated, the National Central University of Taiwan, the Space Telescope Science Institute, the National Aeronautics and Space Administration under Grant No. NNX08AR22G issued through the Planetary Science Division of the NASA Science Mission Directorate, the National Science Foundation under Grant No. AST-1238877, the University of Maryland, and Eotvos Lorand University (ELTE).

The light curves and spectra of the remaining 14 EVQs, in addition to the two shown in Figure 1, are provided in Figure A.1, in the same order as that in Table 1 in terms of the level of dimming.