The 'Big Dipper': The nature of the extreme variability of the AGN SDSS J2232-0806
Daniel Kynoch, Martin J. Ward, Andy Lawrence, Alastair G. Bruce, Hermine Landt, Chelsea L. MacLeod
MMNRAS , 1–16 (2019) Preprint 20 February 2019 Compiled using MNRAS L A TEX style file v3.0
The ‘Big Dipper’: The nature of the extreme variability ofthe AGN SDSS J2232 − Daniel Kynoch (cid:63) , Martin J. Ward , Andy Lawrence , Alastair G. Bruce ,Hermine Landt and Chelsea L. MacLeod Centre for Extragalactic Astronomy, Department of Physics, Durham University, South Road, Durham, DH1 3LE, UK Institute for Astronomy, SUPA (Scottish Universities Physics Alliance), University of Edinburgh, Royal Observatory, Blackford Hill,Edinburgh EH9 3HJ, UK Harvard-Smithsonian Center for Astrophysics, 60 Garden Street, Cambridge, MA 02138, USA
Accepted XXX. Received YYY; in original form ZZZ
ABSTRACT
SDSS J2232 − Key words: galaxies: active – black hole physics – accretion, accretion discs – quasars:emission lines – galaxies: individual: SDSS J223210.52 − Active galactic nuclei (AGN) are powered by the gravi-tational energy reprocessed as matter spirals inward andis finally accreted by the central supermassive black hole(BH). Two of the defining characteristics of AGN are theirvery high bolometric luminosities, and in the case of thosethat are not obscured, by their significant multi-frequencyvariability on many timescales. Numerous variability stud-ies have been conducted, both on large samples of AGN(e.g. Stripe 82, MacLeod et al. 2012, Schmidt et al. 2012,and Zuo et al. 2012) and detailed studies of individual cases(e.g. NGC 4593 by McHardy et al. 2018 and NGC 5548 byPei et al. 2017 and references therein). In addition to thesestudies some cases of extreme variability have been iden-tified in the form of the so-called ‘changing-look’ quasars(CLQs: e.g. MacLeod et al. 2018, Yang et al. 2018, Rum-baugh et al. 2018 and LaMassa et al. 2015) which are AGN (cid:63)
E-mail: [email protected] with (dis)appearing broad emission lines as well as strongcontinuum changes. It is very probable that more than onephysical mechanism is responsible for the variations seenacross all samples. Changes in the dust extinction in someAGN were proposed in early studies (e.g. Goodrich 1995),but this explanation is not generally preferred in the caseof changing-look AGN. In recent studies, often the mostfavoured cause is a change in the emission from the accretiondisc or its associated Comptonisation regions (e.g. Katebiet al. 2018, Noda & Done 2018, Stern et al. 2018, Ross et al.2018, Wang et al. 2018, Sheng et al. 2017, Gezari et al. 2017,Parker et al. 2016, Ruan et al. 2016, MacLeod et al. 2016,Runnoe et al. 2016 and LaMassa et al. 2015). Other, rarerevents, such as stellar tidal disruption, supernovae in the nu-clear regions, and gravitational microlensing, have also beenproposed (e.g. Lawrence et al. 2016, Bruce et al. 2017 andreferences therein). To make further progress it is importantto better characterise the properties of variability to helpdistinguish between the various mechanisms responsible. © a r X i v : . [ a s t r o - ph . GA ] F e b D. Kynoch et al. − SDSS J223210.51 − − z = . (Collinson et al. 2018). It wasidentified as a ‘slow-blue nuclear hypervariable’ object byLawrence et al. (2016) on the basis that it showed large-amplitude optical brightness variability ( | ∆ g | (cid:62) . ) and thechange was slow and blue (occurring over several years, incontrast to the fast and red transients which are likely asso-ciated with supernovae).Our photometric monitoring of this source with the Liv-erpool Telescope since 2013 has captured one substantialdimming event, and there is sparsely sampled archival pho-tometry that is consistent with similar past events. We aim to investigate whether the variability behaviour ofthis source is best explained by either obscuration of the nu-cleus, or by some intrinsic change in the emission from thecentral engine. The optical spectroscopic monitoring cam-paign conducted with the William Herschel Telescope allowsus to investigate changes in both the AGN continuum andline emission from the broad line region (BLR).Throughout this paper, we assume a flat Λ CDM cosmol-ogy with H = km s − Mpc − , Ω m = . and Ω Λ = . . Forthe redshift z = . this cosmology implies a luminositydistance of 1410.8 Mpc and a flux-to-luminosity conversionfactor of . × cm . Lawrence et al. (2016) found that in 2012 the PanSTARRS-1 (PS1) 3 π Survey g band photometry of SDSS J2232 − was 1.8 magnitudes brighter than it was in a SDSS photo-metric observation made in 2000. To further investigate thisinteresting source, a photometric monitoring campaign be-gan in 2012 using the Liverpool Telescope and is ongoing.Optical spectroscopic monitoring commenced in 2013, pri-marily using the William Herschel Telescope, with an addi-tional two spectra taken in late 2017 with the MMT. The ob-serving campaign has revealed a dip in brightness of arounda factor three in flux and shows a recovery in our most re-cent observations. In this section we present our analysis ofthe optical data. The Liverpool Telescope (LT) is a fully-robotic, remotelycontrolled 2 m telescope that observes autonomously fromLa Palma in the Canary Islands. Photometric observationswere taken in the r , g and u bands. Forty-four independentphotometric observations were obtained using the g filter( λ eff . = ˚A) between 2012 September and 2018 July areshown in Figure 1. The g and r bands are much more fre-quently sampled than the u band, for which we have onlytwenty-one photometry points. The observed variability am-plitude in the g band ( ∆ g ≈ . ) is greater than that of r band( ∆ r ≈ . ) although we note that the r band ( λ eff = ˚A)is subject to increasing contamination from the host galaxy as the AGN contribution diminishes. In addition, the u band( λ eff = ˚A) covers the strong, broad Mg ii emission line(observed at 3573 ˚A) and so it is not a clean measure ofthe AGN continuum. For these reasons, in this study we useonly photometry obtained in the g band. The 4.2 m William Herschel Telescope (WHT) is also situ-ated on the island of La Palma. SDSS J2232 − × binning in thespatial direction was used to improve the signal-to-noise ra-tio (SNR). This set-up gave a spectral resolution of R ≈ at 7200 ˚A in the red and R ≈ at 5200 ˚A in the blue,for a slit width of 1 arcsecond. The total wavelength cover-age was ≈ –10600 ˚A, this window includes the principalemission lines Mg ii λ , H β λ , [O iii ] λλ , andH α λ .The data reduction was performed with a pipeline usingcustom pyraf scripts and standard techniques. The pipelineis described in detail in Bruce et al. (2017) (Section 2.3.3 inthat paper).Unfortunately, we do not have a spectrum contempora-neous with the nadir of the LT lightcurve, which occurredaround 2014 September 17. The spectra obtained on 2014July 23 and December 16 were recorded 56 days before and90 days after the photometric minimum and sample thefalling and rising side of the dip in the lightcurve, respec-tively (see Figure 1). The MMT is a single 6.5 m mirror telescope on MountHopkins, Arizona. Two optical spectra of SDSS J2232 − − grating and a 1 arcsecond slit. This set-up gives a spectralresolution of R ≈ at 4800 ˚A, lower than that we obtainedwith the WHT. The target spectra that we use here are theco-added medians of three 10 minute exposures. Optical and ultraviolet fluxes are affected by reddeningcaused by dust in the Milky Way. The Galactic neutralhydrogen column density towards SDSS J2232 − N H = . × cm − (Dickey & Lockman 1990), implies a colourexcess E ( B − V ) = . mag based on the relation derived byBohlin et al. (1978). Here, and in Section 3, we correct ourdata for Galactic reddening using this value of E ( B − V ) andthe Milky Way reddening curve of Cardelli et al. (1989). MNRAS , 1–16 (2019) xtreme variability of SDSS J2232 − LT gSDSS gWHT ‘g’MMT ‘g’PS1 g
XMM
CSS
WISE
W2 (4.6 µ m) WISE
W1 (3.4 µ m) M ag n i t ud e Figure 1.
Top : the optical lightcurve of SDSS J2232 − ∆ g = . magnitudes between theSloan Digital Sky Survey (SDSS) observation made in 2000 and the PanSTARRS-1 (PS1) observation of 2012. We show our follow-upoptical photometric monitoring with the Liverpool Telescope (LT) and archival data from the Catalina Sky Survey (CSS). As well asthe direct photometric points, we show the equivalent g magnitudes derived from spectroscopic observations made with the WilliamHerschel Telescope (WHT) and the MMT. A global greyscale flux correction of − . mag has been applied to the spectral magnitudes(see Section 2.2 in the text). The date of our XMM-Newton
X-ray and optical-UV observations is also indicated.
Bottom : the
WISE infrared lightcurves of SDSS J2232 − Before we perform our spectral analysis, we rescale our spec-tra to account for variations in the absolute flux calibra-tion caused by effects such as seeing (slit losses) and thincloud. Since the strong, narrow [O iii ] λ forbidden emis-sion line originates in a low-density, large-volume gas, itshould not vary during the course of our monitoring pe-riod and is therefore a suitable line to use for internal cross-calibration (provided it it not spatially resolved). Ratherthan simply assuming the flux in the line remains constant(which depends upon an accurate determination of the un-derlying continuum flux level), we assume instead that theline profile is constant and determine the appropriate fluxscaling factors using the python package mapspec devel-oped by Fausnaugh (2017). This package is an implementa-tion of, and improvement on, the method of van Groningen& Wanders (1992). As noted by them this method shouldproduce a more accurate internal flux scaling than the stan-dard method of simply scaling each spectrum so that theintegrated [O iii ] λ line flux is equal to a chosen refer-ence value. From our internally-scaled optical spectra, we calculated theequivalent LT g magnitude. The LT optical CCD camera was changed from the RATcam to the IO:O at the end of2014 February, so in our calculations we use the filter spec-ifications appropriate to the LT instruments in use at thetime the spectrum was recorded, although the resultant dif-ference in magnitude is very minor. For each spectrum wemeasured the mean flux (cid:104) ν F ν , g (cid:105) in the LT g band (RAT-cam 3945–5532 ˚A, IO:O 3933–5630 ˚A) then calculated the g magnitude equivalent g = − . (cid:18) (cid:104) ν F ν , g (cid:105) × ν eff ZP (cid:19) mag , (1)where ZP is the zero point magnitude of the filter (RAT-cam 3940.5 Jy; IO:O 3936.7 Jy) and ν eff is the frequencyequivalent to the filter’s effective wavelength λ eff (RATcam4730 ˚A; IO:O 4696 ˚A).By comparison with the LT g magnitudes, we foundthat the equivalent magnitudes appeared systematically off-set by ≈ . mag. This slight discrepancy is likely due toslit losses, resulting in a lower flux in our narrow-slit spec-tra compared with the large-aperture photometry. Adjustingthe magnitudes by − . mag (an increase of ≈ per centin flux) the equivalent magnitudes replicate both the shapeand level of the LT lightcurve, as can be seen in Figure 1.In the following, all of the measurements that we make fromthe spectra include the internal and absolute flux scalingsdescribed here. MNRAS000
WISE infrared lightcurves of SDSS J2232 − Before we perform our spectral analysis, we rescale our spec-tra to account for variations in the absolute flux calibra-tion caused by effects such as seeing (slit losses) and thincloud. Since the strong, narrow [O iii ] λ forbidden emis-sion line originates in a low-density, large-volume gas, itshould not vary during the course of our monitoring pe-riod and is therefore a suitable line to use for internal cross-calibration (provided it it not spatially resolved). Ratherthan simply assuming the flux in the line remains constant(which depends upon an accurate determination of the un-derlying continuum flux level), we assume instead that theline profile is constant and determine the appropriate fluxscaling factors using the python package mapspec devel-oped by Fausnaugh (2017). This package is an implementa-tion of, and improvement on, the method of van Groningen& Wanders (1992). As noted by them this method shouldproduce a more accurate internal flux scaling than the stan-dard method of simply scaling each spectrum so that theintegrated [O iii ] λ line flux is equal to a chosen refer-ence value. From our internally-scaled optical spectra, we calculated theequivalent LT g magnitude. The LT optical CCD camera was changed from the RATcam to the IO:O at the end of2014 February, so in our calculations we use the filter spec-ifications appropriate to the LT instruments in use at thetime the spectrum was recorded, although the resultant dif-ference in magnitude is very minor. For each spectrum wemeasured the mean flux (cid:104) ν F ν , g (cid:105) in the LT g band (RAT-cam 3945–5532 ˚A, IO:O 3933–5630 ˚A) then calculated the g magnitude equivalent g = − . (cid:18) (cid:104) ν F ν , g (cid:105) × ν eff ZP (cid:19) mag , (1)where ZP is the zero point magnitude of the filter (RAT-cam 3940.5 Jy; IO:O 3936.7 Jy) and ν eff is the frequencyequivalent to the filter’s effective wavelength λ eff (RATcam4730 ˚A; IO:O 4696 ˚A).By comparison with the LT g magnitudes, we foundthat the equivalent magnitudes appeared systematically off-set by ≈ . mag. This slight discrepancy is likely due toslit losses, resulting in a lower flux in our narrow-slit spec-tra compared with the large-aperture photometry. Adjustingthe magnitudes by − . mag (an increase of ≈ per centin flux) the equivalent magnitudes replicate both the shapeand level of the LT lightcurve, as can be seen in Figure 1.In the following, all of the measurements that we make fromthe spectra include the internal and absolute flux scalingsdescribed here. MNRAS000 , 1–16 (2019)
D. Kynoch et al. β and [O iii ]4700 4800 4900 5000 5100Rest wavelength [˚A]0123 123456789 H α and [N ii ]6400 6600 68000123123456789 F λ [ − e r g s − c m − ˚A − ] Mg ii R a t i o F λ [ − e r g s − c m − ˚A − ] M g ii [ N e v ] [ N e iii ] H (cid:15) H δ H γ H e ii H β } [ O iii ] H e i H α Brightest: 2013 June 10Faintest: 2014 December 16246 D i ff e r e n ce Brightest − Faintest3000 4000 5000 6000 7000Rest wavelength [˚A]123 R a t i o Brightest ÷ Faintest
Figure 2.
Top : All eleven optical spectra of SDSS J2232 − iii ] λ emission line profile and corrected forGalactic reddening ( A V = . ). The brightest spectrum (2013 June) is shown in blue and the faintest spectrum (2014 December) isshown in red; the other nine spectra are shown in grey. Prominent emission lines are labelled. In the lower panels the difference spectrumis shown in green and the ratio spectrum in purple. Bottom : Continuum-subtracted regions containing key emission lines. Spectra arecolour-coded as in the top plot. In the lower panels, the ratios between the brightest and faintest spectra is shown in purple.
All eleven optical spectra are shown in the top panel of Fig-ure 2. To highlight the spectral variability we have colouredthe brightest and faintest spectra in blue and red, respec-tively, and plotted both their difference and ratio in greenand purple, respectively, in the panels below. The ratio be-tween the brightest and faintest spectrum shows the frac-tional variability at each wavelength. The fractional variabil-ity at longer wavelengths is diluted by emission from the host galaxy and we see in the ratio spectrum that the fractionalvariability is greater in the blue end. Taking the differenceremoves the constant components including the host galaxy.In the difference spectrum it can be seen that the absoluteflux variation in the blue continuum ( λ (cid:46) ˚A) is greaterthan in the red. The [O iii ] λλ , lines, which we as-sumed to be non-variable, are absent in the difference spec-trum which gives us confidence that the flux scaling methodwe have adopted works well. Whereas differences in the H α MNRAS , 1–16 (2019) xtreme variability of SDSS J2232 − and H β lines between bright and faint spectra are clear, theMg ii line appears to be less variable. This is obvious in theratios of the continuum-subtracted lines (shown in purplein the lower panels of the bottom three plots of Figure 2)where the core of Mg ii changes very little and no substantialchange is apparent in the broad wings. The change in theBalmer lines is most apparent on the blue side of the lines,which seem to have a slight ‘red shoulder’ in the fainter spec-tra. A similar skewness of the H α profile in the faint state ofthe CLQ J0159 + iii ] λ lines (as determined from our model fits) are aligned.The two MMT spectra are noisier than the nine obtained atthe WHT and have less wavelength coverage (particularlyredward of H α ). We confirmed that the shapes and generalfeatures of our mean and RMS spectra are (broadly) un-changed if we exclude the MMT spectra. Having done so,we proceeded with the mean and RMS spectra determinedfrom just the WHT observations, so as to extend our resultsinto the red.The resultant spectra are shown in Figure 3. As inthe difference spectrum, the [O iii ] lines are removed in theRMS spectrum whereas the Balmer lines, Balmer continuum( ≈ –4000 ˚A) and He ii λ are all visible. We also notethat the Mg ii emission line is absent from the RMS spec-trum, which we discuss later. Comparing the shapes of themean and RMS spectra, we see that the RMS spectrum isbluer since the non-variable host galaxy component has beenremoved: we discuss this in Section 3.3.1. The shape of theRMS spectrum is very similar to that of an accretion disc;in Figure 3 we show a standard disc spectrum for compari-son, calculated for a BH mass of × M (cid:12) , L / L Edd = . and outer radius of 100 gravitational radii (equal to thatdetermined in our SED model in Section 3.3.2). The continuum and emission line fitting was performed us-ing a custom python script employing the lmfit package which employs a Levenberg-Marquardt algorithm for non-linear, least-squares minimisation. The fitting routine ap-peared to underestimate the errors on the returned param-eters, so rather than quoting the error on a single fit, aniterative approach was taken. Each spectrum was fitted 100times: on each iteration Gaussian noise was added to the fluxdensity with the amplitude of the noise determined by themeasurement error. The final model parameters and errorsare the mean and standard deviation calculated from the100 iterations. The standard deviation quantifies the spreadof parameter values that can reasonably fit the data. Theerrors on the physical quantities derived from the model pa-rameters (e.g. the line flux, equivalent width etc.) have beenpropagated using standard methods. The results of our iter-ative fitting procedure are tabulated in Tables A1, A2 andA3 in the Appendix. https://lmfit.github.io/lmfit-py/ F λ [ − e r g s − c m − ˚A − ] MeanRMS × Figure 3.
The mean spectrum of SDSS J2232 − L / L Edd = per cent and R out = R g . For the WHT spectra, the (rest frame) 3900–7800 ˚A con-tinuum is estimated from five emission line free windows ofwidth 50 ˚A; these are centred on the wavelengths 4240, 5100,6205, 7050 and 7700 ˚A. Because of the narrower wavelengthcoverage of the two spectra obtained using the MMT, onlythe first three of these windows are available. We fit a power-law continuum of the form F λ = C (cid:0) λ / ˚A (cid:1) − α through thesepoints to determine the global continuum, allowing the slope α and normalisation C to be free parameters in the fit. To model the Balmer lines, the red continuum is subtractedfrom two wavelength windows containing the emission linesof interest (rest frame 4740–5100 ˚A for H β and [O iii ]; 6380–6800 ˚A for H α and [N ii ]). The permitted lines were initiallyfit with a sum of two Gaussians (one broad and one nar-row) with the same central wavelength. However, there wereclearly substantial residuals in the line profiles, particularlyprominent in the red wing of H α . We therefore added a thirdGaussian component to the Balmer lines, modelling a verybroad base, and allowed this to be offset from the centralwavelength of the narrower components. The two [N ii ] for-bidden lines were each fit with a single, narrow Gaussian. Aswell as a strong, narrow Gaussian, a weak, broad Gaussianbase was added to the [O iii ] λ and λ lines. In allfits we include the permitted Fe ii emission line template ofBruhweiler & Verner (2008), with its normalisation left as afree parameter in the fits. The model was refined to includethe following constraints:i) all narrow, broad and very broad lines have the samevelocity width (with the exception of the broad basesof the [O iii ] forbidden lines: these had equal width but MNRAS000
The mean spectrum of SDSS J2232 − L / L Edd = per cent and R out = R g . For the WHT spectra, the (rest frame) 3900–7800 ˚A con-tinuum is estimated from five emission line free windows ofwidth 50 ˚A; these are centred on the wavelengths 4240, 5100,6205, 7050 and 7700 ˚A. Because of the narrower wavelengthcoverage of the two spectra obtained using the MMT, onlythe first three of these windows are available. We fit a power-law continuum of the form F λ = C (cid:0) λ / ˚A (cid:1) − α through thesepoints to determine the global continuum, allowing the slope α and normalisation C to be free parameters in the fit. To model the Balmer lines, the red continuum is subtractedfrom two wavelength windows containing the emission linesof interest (rest frame 4740–5100 ˚A for H β and [O iii ]; 6380–6800 ˚A for H α and [N ii ]). The permitted lines were initiallyfit with a sum of two Gaussians (one broad and one nar-row) with the same central wavelength. However, there wereclearly substantial residuals in the line profiles, particularlyprominent in the red wing of H α . We therefore added a thirdGaussian component to the Balmer lines, modelling a verybroad base, and allowed this to be offset from the centralwavelength of the narrower components. The two [N ii ] for-bidden lines were each fit with a single, narrow Gaussian. Aswell as a strong, narrow Gaussian, a weak, broad Gaussianbase was added to the [O iii ] λ and λ lines. In allfits we include the permitted Fe ii emission line template ofBruhweiler & Verner (2008), with its normalisation left as afree parameter in the fits. The model was refined to includethe following constraints:i) all narrow, broad and very broad lines have the samevelocity width (with the exception of the broad basesof the [O iii ] forbidden lines: these had equal width but MNRAS000 , 1–16 (2019)
D. Kynoch et al. this was not tied to the width of the broad permittedlines);ii) the very broad lines in the H α and H β profiles havethe same velocity offset;iii) it proved impossible to reliably fit both the widthand offset of the very broad lines simultaneously sowe fixed the velocity width of these components to ≈ km s − and placed the limit ∆ v vb (cid:46) + km s − on the offset ;iv) the [O iii ] λ and λ lines have a fixed flux ratioof 1:3;v) the [N ii ] λ and λ lines have a fixed ratio of1:3;vi) the stronger [N ii ] λ line has its amplitude fixed tothe mean value determined in the WHT spectra;vii) the narrow lines ought not to vary significantly overthe monitoring period, therefore the H α narrow line wasfixed to 0.67 of the [O iii ] λ flux, the error-weightedmean value determined from all of the WHT spectra;viii) the Balmer decrement of the narrow lines was a chal-lenge to determine so was fixed at 6.7, again the error-weighted mean value determined from the WHT spec-tra.The narrow-line Balmer decrement adopted here is high, al-though it is within the range ≈ –12 found by Jin et al.(2012) for a sample of fifty-one type 1 AGN and at the upperend of the range found by Lu et al. (2019) for 554 SDSS DR7quasars. If the intrinsic narrow line region (NLR) Balmerdecrement is 2.9 (Osterbrock & Ferland 2006) the measuredvalue implies an NLR reddening of A V ≈ . mag. However,our aim is to investigate relative changes in the broad linedecrement so as long as the subtraction of the narrow linecomponents is consistent, its precise value will have littleeffect on our results.In calculating the Balmer and [O iii ] emission line EWswe have subtracted the host galaxy contribution to the fluxbeneath the line (these are determined in § . The Balmer, [O iii ] and [N ii ] lineproperties derived from the best fit model parameters arequoted in Tables A1 and A2. Examples of our Balmer, [O iii ]and Mg ii emission line fits are shown in Figure 4.Since there are no emission line free regions in the vicin-ity of the Mg ii line, we do not subtract the continuum beforefitting the line. Instead we fit the line, Fe ii template and apower-law continuum simultaneously in the wavelength win-dow 2650–2950 ˚A. The Mg ii λλ , doublet was notresolved in the composite spectrum produced by stackingthe WHT spectra; we therefore fit a single Mg ii λ pro-file. This emission line was fitted with two Gaussians, onebroad and one very broad for the base. As well as measur-ing the FWHM of the two components separately, we alsocalculate the FWHM of the total line profile. The quantities The line width is approximately equal to the mean FWHM ofthe very broad Balmer line components of the broad line AGNmodelled by Jin et al. (2012). The offset was limited to keep thecentre of the very broad component within the core of the line. The host galaxy makes a negligible contribution at the wave-length of Mg ii . Table 1.
Black hole mass estimates from optical spectra
Relation log ( K ) α FWHM
L M BH FWHM(H α ), λ L ˚A . FWHM(H α ), L H α . FWHM(H β ), λ L ˚A . FWHM(Mg ii ), λ L ˚A . The broad line FWHMs are in km s − , the luminosities L in erg s − and the calculated black hole masses M BH = K L α × FWHM in M (cid:12) . derived from the best fit model parameters are quoted inTable A3. To calculate the mass of the black hole from our emissionline and continuum measurements we use the relation M BH = K × L α × FWHM (2)of Mej´ıa-Restrepo et al. (2016) with the appropriate valuesof K and α taken from their Table 7 (the local calibrationcorrected for small systematic offsets) for the relevant com-binations of the emission line FWHM and continuum or lineluminosity L . In Table 1 we quote the K and α values used foreach relation along with the line and continuum parametersdetermined as the error-weighted means of values obtainedin the four brightest spectra . We find that the mass is inthe range M BH = . – . × M (cid:12) (see Table 1), marginallygreater than the . – . × M (cid:12) determined by Collinsonet al. (2018). There are considerable uncertainties on themasses estimated by virial methods, which are due to thescatter on the scaling relations. For relations based on H α and H β the 1 σ scatter is in the range 0.13–0.18 dex; theMg ii relation has a greater scatter of 0.25 dex. We adopt amass of × M (cid:12) in the following. XMM-Newton
A 30 ks
XMM-Newton observation of SDSS J2232 − XMM
Science AnalysisSoftware ( sas , v16.0.0) and the latest calibration files avail-able at the time. The X-ray observation suffered from sub-stantial particle background flaring such that, after filtering,the remaining good time intervals were 8.6, 8.5 and 8.2 ksfor the pn, MOS1 and MOS2 detectors, respectively. Thesource spectra were extracted from 47 arcsec radius circu-lar regions centred on the source. The background spectrum Those recorded on 2013 June 10 and August 7 and 2016 July 9and October 22. MNRAS , 1–16 (2019) xtreme variability of SDSS J2232 − β and [O iii ]4700 4800 4900 5000 51000 . . . α and [N ii ]6400 6600 68000 . . . F λ [ − e r g s − c m − ˚A − ] Mg ii . . . R a t i o β and [O iii ]4700 4800 4900 5000 51000 . . . α and [N ii ]6400 6600 68000 . . . F λ [ − e r g s − c m − ˚A − ] Mg ii . . . R a t i o Figure 4.
Examples of emission line fits to continuum-subtracted spectral windows. In the upper panels, the solid black lines show thewavelength regions of the spectra that were fit, and the solid grey area indicates the error on the flux density; the green short-dashed,dashed and long-dashed lines show the modelled narrow, broad and very broad components of the permitted lines, respectively; themagenta short-dashed lines show the modelled [O iii ] and [N ii ] forbidden lines; an Fe ii emission template is shown by the purple dottedline and the total model is shown by the solid blue line. The lower panels show the data / model ratios in the fitted regions. was extracted from larger 94 arcsec radius circular regionsoffset from the source on a blank area of sky. The spectrawere regrouped so as not to oversample the detectors’ in-trinsic energy resolution by a factor of more than three andto contain at least 20 counts per energy bin, so that theyare suitable for a χ analysis.The OM photometry in the three filters were extractedusing the sas tasks omichain and omisource , following thestandard procedures. The OM filter bandpasses cover sev-eral emission lines and so do not accurately represent thecontinuum flux level. Following the method of Elvis et al.(2012), we can ‘correct’ the photometric fluxes to obtain animproved estimate of the continuum level by multiplying themeasured fluxes by the photometric correction factor P c = BWEW rest × ( + z ) + BW (3)where BW is the bandwidth of the photometric filter cover-ing a line of rest-frame equivalent width EW rest . The OM Ufilter ( BW = ˚A) covers the Mg ii emission line, for whichwe estimate EW rest ≈ ˚A → P c = . . Assuming a C iii ] EW rest ≈ ˚A (Vanden Berk et al. 2001), the correction fac-tor in the UVM2 filter is P c = . . We conclude that theUVW1 filter is very weakly affected by line emission, sincethe C iii ] and Mg ii lines only partially appear at the veryends of its bandpass where the sensitivities are lowest. Analysis of the X-ray spectra was performed in xspec (Ar-naud 1996) v12.9.1e. The spectra from the three EPICdetectors were fitted simultaneously, allowing for cross-normalization factors to account for differences in calibra-tion between the detectors; these did not vary by morethan 5 per cent. All models included a Galactic absorp-tion component ( phabs ) with the column density fixed at N GalH = . × cm − . A single power-law was an unsatis-factory fit to the data, giving a reduced χ of 1.27. A broken Table 2.
X-ray spectral modelsModel Parameter Value powerlaw Γ . ± . Norm. ( . ± . ) × − χ /d.o.f. / = . bknpower Γ . + . − . E brk (keV) . + . − . Γ . + . − . Norm. (cid:16) . + . − . (cid:17) × − χ /d.o.f. / = . zphabs × N intH ( cm − ) < bknpower Γ . + . − . E brk (keV) . + . − . Γ . + . − . Norm. (cid:16) . + . − . (cid:17) × − χ /d.o.f. / = . All models included a Galactic absorption component( phabs ) with the column density fixed at N GalH = . × cm − . power-law was a significant improvement, decreasing the χ value by 48 for the introduction of two additional free pa-rameters and we achieve an acceptable fit with a reduced χ of 1.01. The F -test probability of this improved model was > . per cent. We then tested for an intrinsic absorber bythe inclusion of a zphabs component with the redshift fixedto that of the source. This gave no significant improvementin the fit and we determined an upper limit on the intrinsiccolumn density N intH < × cm − . The
Wide-field Infrared Survey Explorer ( WISE , Wrightet al. 2010) telescope observed SDSS J2232 − MNRAS000
Wide-field Infrared Survey Explorer ( WISE , Wrightet al. 2010) telescope observed SDSS J2232 − MNRAS000 , 1–16 (2019)
D. Kynoch et al.
Catalog, hosted by the Infrared Science Archive (IRSA ); inTable 3 we quote the reported instrumental profile-fit mag-nitudes. The photometric quality of these detections wereA (best) for the W1, W2 and W3 filters and B for the W4filter.As well as the catalogue magnitudes, we also obtainedinfrared lightcurves in the W1 and W2 filters from the WISE and
Near-Earth Orbit WISE Reactivation ( NEO-WISE ) archives . In addition to the two visits made dur-ing the WISE mission, SDSS J2232 − NEOWISE mission in December 2013. Typically a dozen exposures aremade on each visit; to construct the lightcurves shown inFigure 1, we have calculated the mean and standard error onthe magnitudes recorded on each visit. We exclude the sevenexposures taken on MJD 57345, because there was a largescatter on these magnitudes and a set of eleven exposureswas taken three days later. This visit on MJD 57348 (2015November 19) corresponds to the minima of the infraredlightcurves and occurs 428 days later than the observed min-imum in the LT optical lightcurve (see Section 2.1.1). Thereis a 0.26 mag peak-to-trough change in W1 and a 0.21 magchange in W2.
SDSS J2232 − J , H and K s profile-fit magnitudes reported in the 2MASSAll-Sky Point Source Catalog (PSC) . The observation wasmade on 1998 October 1 and the photometric quality flag isC for all filters. Although no Sloan Digital Sky Survey (SDSS) spectroscopicdata exists for this source, photometry was obtained on 2000March 9. As can been seen in Figure 1, the source was ina very low state at this time. The object was classified as a(passive) galaxy based on its photometric colours. π Survey
The PanSTARRS-1 (PS1) 3 π Survey was conducted between2009 and 2014, observing the / of the sky north of − ◦ declination multiple times per year in each of five filters (seeMagnier et al. 2013 and Chambers et al. 2016). Originallysearching for tidal disruption events, Lawrence et al. (2016)identified SDSS J2232 − as one of a number of ‘slowblue nuclear hypervariables’: objects with no previouslyknown AGN, blue colours and evolution on timescales ofyears. This particular source was brighter by ∆ g = . ± . in 2012 compared with the SDSS photometry of 2000. http://irsa.ipac.caltech.edu/ Because of the depletion of hydrogen coolant, only the W1 andW2 filters have been operable since the beginning of the
NEO-WISE mission. Also available from IRSA, see earlier note. The common name of the source in this paper is J223210.
We located a record for SDSS J2232 − ). The B j band ( λ = ˚A) ob-servation was made using the UK Schmidt Telescope (Can-non 1975) on Siding Spring Mountain, NSW, Australia, on1986 August 1. Its sCorMag (stellar magnitude in the Vegasystem) is given in the SSA as B j = . mag. Convertingthis to a g band AB magnitude, we estimate g ≈ . ± . ,where the uncertainty is the standard single-passband uncer-tainty on SuperCOSMOS magnitudes (Hambly et al. 2001). Two short-exposure photometric observations were madewith the Wide Field Camera 3 (WFC3) onboard the
HubbleSpace Telescope ( HST ) on 2015 September 18. The exposuretimes were 330 s in the wide IR F125W filter ( λ eff = . µ m, J band) and 1200 s in the extremely wide UVIS F475X filter( λ eff ≈ ˚A, and including the g band). Two epochs of ultraviolet (UV) photometry were found bysearching the
Galaxy Evolution Explorer ( GALEX , Mar-tin et al. 2005) space telescope archive. In both records,the UV source is coincident with the optical coordinates ofSDSS J2232 − ≈ between the two epochs and there is also anapparent colour change, with SDSS J2232 − Infrared and optical emission from the host galaxy bulgemay make a non-negligible contribution to our spectra, par-ticularly in the faint state. It can be seen in our SED (Fig-ure 5) that the bulge component dominates over the AGNcontinuum redward of H β . However, this is not representa-tive of the host galaxy flux in our spectra, since our nar-row 1 arcsec wide slit excludes much of the extended hostgalaxy emission: a typical bulge diameter of 15 kpc wouldbe ≈ . arcsecs across on the sky.We examined the HST images of the source, taken in2015 September (see Section 3.2.6). The high spatial reso-lution of the instrument in principle allows us to separatethe point-like AGN emission from the more extended hostgalaxy. We made a visual inspection of the 1D brightnessprofiles of the source in the two filters. Whereas the sourceemission in the UVIS filter was PSF-like, the J band profilehad a slightly more extended base than the PSF, suggest-ing the presence of some light from the host galaxy. Un-fortunately, however, the snapshot HST exposures are notsufficiently deep to robustly assess the host galaxy emission.Instead, we can estimate the host galaxy luminosity at5100 ˚A in our spectral extraction aperture using the rela-tion of Landt et al. (2011). From a sample of low-redshift http://ssa.roe.ac.uk/ . MNRAS , 1–16 (2019) xtreme variability of SDSS J2232 − Table 3.
The multiwavelength photometric datasetDate Telescope or Filter Measurement Unit log ( ν ) a Flux b Luminosity c survey2010/05/27–28 WISE W4 . ± . Vega mag 13.13 . ± .
16 2 . ± . WISE W3 . ± . Vega mag 13.41 . ± .
44 15 . ± . WISE W2 . ± . Vega mag 13.81 . ± .
21 20 . ± . WISE W1 . ± . Vega mag 13.95 . ± .
21 20 . ± . K s . ± . mag 14.14 . ± .
06 15 . ± . H . ± . mag 14.26 . ± .
21 15 . ± . J . ± . mag 14.39 . ± .
01 13 . ± . z . ± . asinh mag 14.53 . ± . . ± . i . ± . asinh mag 14.61 . ± .
07 10 . ± . r . ± . asinh mag 14.69 . ± .
07 10 . ± . g . ± . asinh mag 14.81 . ± .
06 6 . ± . π g . ± . mag 14.80 . ± . ± g . ± . AB mag 14.81 . ± . . ± . g ∗ . ± . AB mag 14.81 ± ± B j . ± . Vega mag 14.88 ± ± u . ± . asinh mag 14.92 . ± . . ± . XMM-Newton
OM U . ± . cts s − . ± . . ± . XMM-Newton
OM UVW1 . ± . cts s − . ± . . ± . XMM-Newton
OM UVM2 . ± . cts s − . ± . ± GALEX
NUV . ± . AB mag 15.12 . ± . ± GALEX
NUV . ± . AB mag 15.12 . ± . ± GALEX
FUV . ± . AB mag 15.29 . ± . ± GALEX
FUV . ± . AB mag 15.29 . ± . ± Notes : a Logarithm of the observed frequency ν in Hz; b observed flux ν F ν in units of − erg s − cm − ; c intrinsic luminosity ν L ν in units of erg s − , dereddened where appropriate. ∗ Converted from the quoted B j magnitude below. ( z (cid:46) . ), bright, broad emission line AGN, the authors de-termined the host galaxy luminosities enclosed in the aper-tures from stacked HST images (see their Section 3 and Fig-ure 1). When extracting the WHT spectra, we integrated onaverage 4.75 arcsec in the spatial direction; the 4.75 arcsec aperture is therefore equivalent to a spatial size of 20 kpc at the source. From the Landt et al. (2011) relation we thenestimate F ˚A ≈ . × − erg s − cm − ˚A − .The RMS spectrum we constructed in Section 2.2.3largely removes the non-variable host galaxy contribution,whereas the mean spectrum does not. Therefore, if we as-sume that the mean AGN emission has the same spectralshape as the variable component, we can estimate the hostgalaxy contribution by the ‘red excess’ of the mean spectrumin comparison with the RMS. For the host galaxy compo-nent we used the 5 Gyr old elliptical galaxy template ofPolletta et al. (2007). We add the RMS and host galaxyspectra, and rescale the two components until the sum sat-isfactorily matches the shape of the mean spectrum. Fromthe appropriately-scaled galaxy template we determine themean flux densities in several 150 ˚A wide windows. The fluxdensities at 4861, 5007, 5100 and 6563 ˚A are 4.9, 4.8, 4.6and . × − erg s − cm − ˚A − , respectively; the value at5100 ˚A is consistent with the Landt et al. (2011) estimatecalculated above, given the uncertainties. The host galaxycontribution to the fluxes at 2800 ˚A (under Mg ii ) and at3000 ˚A is negligible. In the rest of this study we correct theAGN continuum fluxes (and hence the emission line EWs)using these values. The emission line EWs recorded in theTables in the Appendix reflect this correction. To model the multiwavelength SED, we use the energy-conserving accretion flow model optxagnf of Done et al.(2012). The model has a standard thin accretion disc fromouter radius R out to R cor . Interior to R cor , the accretion poweris divided between soft and hard Comptonisation regions.The hard Comptonisation region receives the fraction f pl ofthe available accretion power and produces power-law emis-sion with photon index Γ . The soft Comptonisation regionupscatters seed photons from the inner edge of the stan-dard thin disc producing soft X-ray emission in excess ofthe hard coronal power-law (this emission is often called the‘soft X-ray excess’: SX). The soft Comptonisation region isparameterised by its optical depth τ and warm electron tem-perature kT e .In addition to the direct accretion flow emission, we in-clude a redshifted blackbody ( zbbody ) modelling the hotdust which is sampled by the WISE
W1 and W2 bands. InFigure 5 we show the W1 and W2 fluxes corresponding tothe earliest
NEOWISE observation (2014 May 31: the clos-est in time to the
XMM-Newton pointing). The downwarderror bars show the extent of the flux diminution over theobserving period. For completeness, the figure also shows the
WISE
W3 and W4 band fluxes, which sample cooler dust.We do not model these data points; the emission may beattributed to AGN- or starlight-heated dust (or some mix-ture of the two). We show our model SED in Figure 5, alongwith the modelled multiwavelength data. Archival data arealso shown for illustrative purposes, including two epochs of
GALEX
UV photometry, 2MASS infrared photometry andthe SDSS optical photometry from 2000 during which theAGN was in a deep flux minimum. In Figure 5 we also show
MNRAS000
MNRAS000 , 1–16 (2019) D. Kynoch et al. the Polletta et al. (2007) 5 Gyr old elliptical host galaxytemplate which is normalised to fit the SDSS photometry.Our SED model has a very prominent soft Comptoni-sation region that emits from the optical/UV into the softX-ray band. The standard disc component is required onlyto provide a source of seed photons for the soft Compton-isation region in the model calculations and not to fit theshape of the SED itself. We note that Collinson et al. (2018)presented an alternative SED model which contained no softComptonisation region and in which the optical/UV emis-sion was attributed to a standard accretion disc, with theX-ray spectrum modelled by a single power-law component.This model cannot replicate the curvature in the X-ray spec-trum which we detected significantly in Section 3.1.1. Addi-tionally, whilst the single power-law of Collinson et al. (2018)has a photon index of Γ = . , a harder index (such as the Γ = . we determine here) would be expected for a sys-tem of this Eddington ratio (e.g. Kubota & Done 2018).However, the Eddington ratio determined in both models, L / L Edd = . , is the same. We now bring together all of these data-sets, and use themto confront two generically distinct scenarios i.e. that theflux changes seen in Figure 2 are due to reddening by dust,or, that they are a result of an intrinsic variation in the con-tinuum emission from the nuclear region, primarily poweredby processes occurring within the accretion disc.
In Figure 6 we show the relative variations of the continuumfluxes and those of the Mg ii and broad Balmer emissionlines. The estimated host galaxy flux at 5100 ˚A has beensubtracted from the red continuum flux (see Section 3.3.1).Shorter wavelengths are more sensitive to reddening thanlonger ones so, under the assumption that the observedchanges are due to reddening, we would expect the 5100 ˚Aflux to have a shallower fractional variability curve than at3000 ˚A. Based on the Cardelli et al. (1989) Milky Way reddening curve, we have calculated the extinction ( A V ) re-quired to cause the observed fractional changes in the bluecontinuum and then predict the fractional change in the redcontinuum for the same A V . We see that the observed AGNflux at 5100 ˚A shows a significantly greater fractional vari-ability than this prediction, and is broadly consistent withthe fractional variations at 3000 ˚A. There is considerableuncertainty in the AGN continuum 5100 ˚A fractional fluxvariations due to the uncertainty in the host galaxy flux sub-traction. However, even in the very conservative case whenwe perform no host galaxy flux subtraction, the 5100 ˚A frac-tional flux variability is still inconsistent with that predictedfrom a reddening law (as indicated by the upper error barsin Figure 6).Additionally, we find that the amplitude of line flux We note the reddening curves for the Small and Large Magel-lanic Clouds are very similar to the Milky Way curve for wave-lengths > ˚A which we consider here. changes are somewhat lower than those in the continuum.The 3000 ˚A flux exhibits variability of more than a factor twowhereas the lines show only ≈ per cent decrease. (Notethat our spectroscopic observations did not cover the deepflux minimum seen in the photometric lightcurve.) There isa trend for the emission line EWs to be anticorrellated withthe continuum fluxes: increasing when the continuum dimsand vice versa. We find that the minimum (maximum) emis-sion line EWs determined over the spectroscopic monitoringperiod are 570 (1200), 110 (250) and 50 (110) ˚A for H α , H β and Mg ii , respectively. In the case of a simple screen ob-scuring both the accretion disc (from which the continuumoriginates) and BLR (from which the broad lines originate),the equivalent widths of the lines ought not to change sinceboth continuum and line flux at any given wavelength willbe suppressed equally. However, if the absorber covers moreof the very compact accretion disc than the larger BLR thenthe EW of the broad lines would be seen to increase.In Figure 7 we show how the continuum colour (the ra-tio of red to blue fluxes) and Balmer decrement have variedtogether. In the simple scenario of a reddening screen of vari-able column density obscuring both the BLR and accretiondisc, there would be a linear relationship between the Balmerdecrement and red/blue continuum flux. We show a redden-ing vector describing the predicted relationship, again basedon the Galactic reddening curve of Cardelli et al. (1989) andpositioned so that the Balmer decrement in the case of zeroreddening is 2.72 (Gaskell 2017). It can be seen that our datado not follow the trend of this reddening vector so reddeningalone cannot explain the observed spectral changes. If the dimming of the AGN continuum and broad emissionline fluxes is due to an obscurer moving across our line ofsight, then we can predict the timescale on which such anoccultation event would occur. We estimate the BLR sizefrom Bentz et al. (2013) using the equation log (cid:18) R BLR (cid:19) = K + α log (cid:18) λ L ˚A erg s − (cid:19) , (4)with values K = . and α = . taken from their‘Clean2 + ExtCor’ calibration. For the range of λ L ˚A ob-served in our monitoring campaign, the BLR size is ≈ –60 light days.Following LaMassa et al. (2015) we calculate the cross-ing time t cross of a cloud occulting the central regions as t cross = . (cid:18) R orb (cid:19) / (cid:32) M (cid:12) M BH (cid:33) / arcsin (cid:18) R src R orb (cid:19) years , (5)where R orb is the orbital radius of the cloud and R src is theradius of the emission source being obscured (here the BLR).As a conservative estimate (minimising the crossing time),we calculate the crossing time for a cloud at the inner edge ofthe BLR, i.e. R src = R orb = R BLR ≈ light days. The cloudcrossing time at this radius is ≈ years, much longer thanthe dip-and-rise event we observe in the lightcurve whichtakes ≈ years in the rest frame. MNRAS , 1–16 (2019) xtreme variability of SDSS J2232 − Table 4.
Multiwavelength SED model parametersModel Parameter Units Description Value zbbody kT dust keV (K) Hot dust temperature . × − ( )B.body norm. Hot dust blackbody normalisation . × − hostpol Gal. norm. Host galaxy template normalisation . × − optxagnf log ( L / L Edd ) Eddington ratio − . kT e keV Electron temperature of soft Comptonisation region . τ Optical depth of soft Comptonisation region . Γ Photon index of power-law coronal emission . f pl Fraction of power below R cor emitted in power-law . R cor R g Inner (standard) accretion disc radius . ( R out ) R g Outer accretion disc radius . F dust erg s − cm − Flux of hot dust blackbody . × − F disc erg s − cm − Flux of (standard) accretion disc . × − F SX erg s − cm − Flux of of soft Compton emission . × − F pl erg s − cm − Flux of coronal power-law emission . × − F UV erg s − cm − AGN flux between 100–4000 ˚A (rest-frame) . × − F AGN erg s − cm − Total AGN flux . × − Note:
Distances are measured in gravitational radii R g = GM BH / c . Rest frequency [Hz]10 L u m i n o s i t y [ e r g s − ] DiscSXCoronaTotal AGNDustGalaxy2MASSSDSSWHT 2013-09-09
WISE
W1 & W2
WISE
W3 & W4
GALEXXMM OM XMM pn Figure 5.
The multiwavelength spectral energy distribution of SDSS J2232 − XMM-Newton
OM and EPIC-pn data of 2013 December 14; WHT spectrum of 2013 September 9 and
WISE
W1 and W2 IR photometry. Additionally,we show other archival data in white:
WISE
W3 and W4 IR photometry from 2010; 2MASS IR photometry from 1998; SDSS photometryfrom 2000 and two epochs of
GALEX
UV photometry from 2003 (faint) and 2004 (bright).
As noted in Section 3.2.1, there is a dip in the infraredlightcurves, delayed with respect to the optical dip byaround 400 days. It can be seen in Figure 5 that there isnegligible host galaxy emission at the wavelengths of the
WISE
W1 and W2 bands (this is true even in the case of astarburst host galaxy, as the IR emission of starlight-heateddust peaks at longer wavelengths). The infrared lightcurves may therefore be evidence of AGN-heated dust reverberatingwith the variable intrinsic AGN continuum. However, whilstthere is a large (factor ≈ ) change in the optical flux, thechange in the near-infrared is much more modest ( ≈ percent). The dust emission ought to be a good bolometer ofthe intrinsic AGN luminosity, so we might expect it to showvariability of the same amplitude as seen in the optical. Ifwe attribute the infrared variability to an echo response tovariations in the central source, we must account for thisdiscrepancy. Here, we assess whether the observed infrared MNRAS000
W1 and W2 bands (this is true even in the case of astarburst host galaxy, as the IR emission of starlight-heateddust peaks at longer wavelengths). The infrared lightcurves may therefore be evidence of AGN-heated dust reverberatingwith the variable intrinsic AGN continuum. However, whilstthere is a large (factor ≈ ) change in the optical flux, thechange in the near-infrared is much more modest ( ≈ percent). The dust emission ought to be a good bolometer ofthe intrinsic AGN luminosity, so we might expect it to showvariability of the same amplitude as seen in the optical. Ifwe attribute the infrared variability to an echo response tovariations in the central source, we must account for thisdiscrepancy. Here, we assess whether the observed infrared MNRAS000 , 1–16 (2019) D. Kynoch et al. F r a c t i o n a l v a r i a t i o n H β broad fluxH α broad fluxMg ii flux H β broad EWH α broad EWMg ii total EW Figure 6.
Fractional variations in the 3000 and 5100 ˚A contin-uum fluxes and emission line fluxes and equivalent widths (EWs)over the monitoring period. In the top panel, as well as the ob-served continuum variations we also show the predicted 5100 ˚Avariations, calculated from the observed 3000 ˚A variations, on theassumption that these are caused by reddening (see text). Themeasured 5100 ˚A fluxes have been corrected for host galaxy con-tamination using our estimate determined in Section 3.3.1 in thetext; the upper error bars indicate the fractional variations calcu-lated with no host galaxy subtraction. For the Balmer lines, valuesare calculated from the sum of very broad and broad components;the Mg ii values are calculated from the whole line profile. R e d / B l u ec o n t i nuu m WHTMMT 18 . . . . . . . . . Figure 7.
The Balmer decrement (the ratio of broad H α to broadH β fluxes) versus continuum colour (the ratio of ˚A to ˚Amonochromatic fluxes) as measured in each of the eleven opticalspectra taken at the WHT and MMT. The colour of the points in-dicates the equivalent g band magnitude of the spectra calculatedin Section 2.2.2: fainter spectra are a darker green. The red lineshows the predicted relation for a Cardelli et al. (1989) Galacticreddening curve (assuming the intrinsic Balmer decrement in thecase of zero reddening is . ). lag and magnitude changes can be plausibly attributed todust reverberation.We can calculate the expected dust reverberation radiusfrom our model SED parameters via R dust , rev = (cid:114) L UV π σ T , (6)where σ is the Stefan-Boltzmann constant and the T is thedust temperature T = K (from Table 4). Since the dustreverberates with the dip in the optical/UV continuum, wetake the UV luminosity in the dip to be a factor 2.5 less thanwe determined at the time of the
XMM-Newton observation: L UV , dip ≈ . × erg s − . We therefore calculate R dust , rev ≈ light days. The observed delay between the minimumof the infrared lightcurve with respect to the optical is ≈ days, equivalent to ≈ days in the rest frame, arounda factor two greater than R dust , rev .We employ the model tori of Almeyda et al. (2017) tosimulate how the dust may respond to a variable, drivingoptical source. The authors consider the cases of a compactand extended torus, in which the ratio of outer to innerdust cloud radii are 2 and 10, respectively and the innerdust radius in their model is set by dust sublimation. Theyconsider the effects of differing illumination of the torus dustclouds. In the case of isotropic illumination, dust sublima-tion surface is spherical. In the anisotropically-illuminatedcase, more ionising flux is emitted in polar directions thanin the equatorial plane; the resultant dust sublimation sur-face is ‘bowl-shaped’ (see e.g. Kawaguchi & Mori 2010) andthe dust near the equatorial plane can survive much closerto the central source than in the isotropic case. The innerdust radius is dependent on the AGN luminosity and thedust sublimation temperature, for which we adopt a valueof 1500 K, close to the mean hot dust temperature found byLandt et al. (2011). For SDSS J2232 − L AGN from the bolometric flux of our modelSED and assume that was 30 per cent greater in the brightstate than observed at the time of the
XMM-Newton obser-vation. We therefore determine that L AGN ≈ × erg s − .For the isotropic case, R in = R sub ≈ . pc ( ≈ light days)whereas for the anisotropic case R in = R sub ( θ = ◦ ) ≈ . pc( ≈ light days).To construct our driving lightcurve, we interpolate be-tween the LT optical photometry points to create a contin-uous lightcurve, which we then smooth to remove the short-term, stochastic variability and retain only the shape of thelonger-term, systematic, large-amplitude changes. Almeydaet al. (2017) provide their impulse response functions at3.6 µ m for a torus viewed at a polar angle of θ obs = ◦ (seetheir Figure 8). We convolve our optical lightcurve with fourresponse functions (for compact/extended, isotropically-/anisotropically-illuminated tori) and compare the simu-lated dust responses with our WISE
W1 data. We find thatthe response functions for the isotropically-illuminated toriproduce much longer lags than is observed. The lags for theanisotropically-illuminated tori are shorter because the ofthe closer proximity of the dust to the optical/UV source,and are in much better agreement with our data. The sim-ulated responses for the compact tori are too deep, an ex-tended distribution is required to smear out the responseand reduce its amplitude. In Figure 8 we show the simulateddust response in the case of an extended, anisotropically-
MNRAS , 1–16 (2019) xtreme variability of SDSS J2232 − Optical dataInput lightcurve
IR dataOutput lightcurve N o r m a li s e d fl u x Figure 8.
The simulated dust response to the variable opti-cal source.
Top:
Our optical data (LT g band photometry) areshown as green circles. We linearly interpolate between these andsmooth the result to create an input optical lightcurve (the blueline). Bottom:
We convolve the input lightcurve with an impulseresponse function to predict the infrared lightcurve (the orangeline). The impulse response function was calculated by Almeydaet al. (2017) for an anisotropically-illuminated, radially extended( R out / R in = ) distribution of dust clouds in a torus of angularwidth σ = ◦ with R dust = light days and a viewed at a po-lar angle of θ obs = ◦ . The WISE
W1 (3.4 µ m) photometry areshown as red squares and the data is normalised such that thefirst point falls on the predicted lightcurve. illuminated torus. In this figure we have slightly decreased R in to 250 light days from the 300 light days calculated from L AGN , to better match the observed lightcurve.
We now assess the predicted timescales for the transmissionof changes through a standard thin accretion disc. In Sec-tion 3.3.2 we determined the outer radius of the accretiondisc to be R ∼ R g . For a disc of this size, the dynamicaltimescale is t dyn ≈ (cid:18) R GM BH (cid:19) / ≈
10 days; (7)the thermal timescale is t therm ≈ t dyn α ≈ , (8)where α ≈ . is the disc viscosity parameter; the viscoustimescale is t visc ≈ t dyn α (cid:18) HR (cid:19) − ≈ (9)where H / R is the ratio of the disc’s thickness to its radius.As Noda & Done (2018) found for Mrk 1018, we find that forSDSS J2232 − The hypothesis of an extrinsic cause of the variability (i.e.variable obscuration) is inconsistent with the observationsin several important respects: • The continuum colour change is inconsistent with red-dening since we see approximately equal fractional fluxchange in the red as in the blue (Figure 6). Even if we per-form no subtraction of host galaxy flux at 5100 ˚A the sourcestill exhibits significantly more variability in the red thanwould be inferred from the blue, assuming that reddeningcauses the variability. We note that the choice of redden-ing curve makes very little difference at the wavelengths westudied. • The Balmer decrements do not change consistently(Figure 7), although this test is less compelling given thesubstantial uncertainties in the measurements. However,since the emission line EWs change, the obscurer cannotbe covering both the accretion disc and all of the BLR. • We are able to place an upper limit of × cm − onthe intrinsic column density from the XMM-Newton
X-rayobservation, although a column of ≈ × cm − wouldbe required to produce the observed 30 per cent drop inthe g band flux. Furthermore, Maiolino et al. (2001) re-ported that the dust reddening of AGN is generally muchlower than one would calculate from the gas column den-sity probed by X-rays, assuming a Galactic dust-to-gas ratioand extinction curve, as we do here. If this were the case forSDSS J2232 − N intH would be predicted,increasing the discrepancy with the X-ray observations. • The timescale for obscuration is far too long. We calcu-late that the crossing time of an obscuring cloud at the innerBLR radius is ≈ years, much longer than the 3 years weobserve. Furthermore, this scenario does not explain how adust cloud could survive relatively near to the central ionis-ing source. • Variable obscuration fails to explain the observed varia-tions in the infrared. Since mid-infrared wavelengths are lesssensitive to reddening than the optical, a 0.26 mag changeat 3.4 µ m would imply a simultaneous . mag change inthe g band which is clearly inconsistent with out data. Ifthe obscurer were exterior to the torus, it would be at anextremely large orbital radius and the crossing time wouldbe even longer than calculated above. If the obscurer wereinterior to the torus it would need to be implausibly close tothe accretion disc (to explain the lag), and implausibly large(to obscure a sufficient fraction of the AGN flux as seen bythe dust). Having ruled out the possibility of an extrinsic change, weconsider that the variability is due to an intrinsic changein the luminosity of the accreting matter. In Section 4.2.1we simulated dust responses to a driving optical continuum.Our intention with this test was not to infer the propertiesof the torus but to examine the plausibility that the in-frared emission reverberates with the optical. Although wehave tested only a few points in the dust response parame-ter space presented by Almeyda et al. (2017), the simulated
MNRAS , 1–16 (2019) D. Kynoch et al.
IR lightcurve shown in Figure 8 captures both the lag andshape of the observed IR variability very well. It is thereforevery plausible that the IR emission exhibits a genuine lightecho of the optical variability.Sheng et al. (2017) studied a sample of changing-lookquasars that exhibited significant, large-amplitude ( | ∆ W1 | or | ∆ W2 | > . mag) mid-infrared variability. Since mid-infrared wavelengths are not strongly affected by dust ex-tinction, the mid-infrared variability would imply muchgreater changes in the optical than observed if both were dueto variable obscuration. They also found that the timescalesfor dust cloud obscuration of the torus were far too longwhereas the observed lags between infrared and optical wereconsistent with those expected for hot dust reverberation.They concluded that in all of the ten objects they investi-gated that the variability was intrinsic in nature. We arguethat SDSS J2232 − ii λ and He i λ emission lines are known to re-spond strongly and rapidly to changes in the continuum, andare also prominent. However, Mg ii almost completely dis-appears in the RMS spectrum, indicating that it has variedvery little over our monitoring campaign. Both Zhu et al.(2017) and Sun et al. (2015) have studied the reverbera-tion of Mg ii in quasars observed multiple times as part ofthe SDSS. Zhu et al. (2017) noted that Mg ii responds rel-atively weakly to changes in the 3000 ˚A continuum. Sunet al. (2015) compared the Mg ii and H β emission line vari-ability and found that Mg ii is ≈ . times less responsiveto changes in the continuum than H β . It is not currentlyknown why this is the case. It may be that Mg ii is emit-ted over a much larger range of radii than H β and so itsresponse is more strongly diluted. Alternatively, differencesin the excitation/de-excitation mechanisms or in the opticaldepths of the two lines are also possible explanations.Whilst we favour an intrinsic cause of the variabilityover an extrinsic cause, our calculations in Section 4.2.2show that the predicted timescales for such changes do notmatch the observations. It has been known for some timethat large-amplitude variability of AGN occurs on timescalesmuch shorter than predicted for thin, viscous accretion discs.Dexter & Begelman (2019) address this so-called ‘quasar vis-cosity crisis’ (Lawrence 2018) and propose that all AGN ac-cretion discs may be ‘magnetically elevated’ and have a muchgreater scale height than is typically assumed, dramaticallyreducing the predicted variability timescales. Alternativemodels have recently been developed to explain the extremevariability seen in individual sources. Ross et al. (2018) pre-sented a scenario for the CLQ SDSS J110057.70 − ∼ R g . Taking a different approach, Noda& Done (2018) determined that Mrk 1018 underwent a spec-tral state transition , similar to those seen in stellar-massblack hole binaries (BHBs). Whilst scaling up to AGN size-scales by BH mass predicts too long variability timescalesin AGN, the authors discuss ways in which scalings betweenBHBs and AGN may break down. − Both MacLeod et al. (2018) and Rumbaugh et al. (2018)have recently presented the results of systematic searches oflong-term extremely variable quasars (EVQs: sources with | ∆ g | > mag) from archival optical data. Rumbaugh et al.(2018) found that EVQs account for ≈ –50 per cent of allquasars and that the EVQs had systematically lower L / L Edd than the parent sample of ‘normal’ quasars. MacLeod et al.(2018) presented follow-up spectroscopic observations of asample of EVQs and were able to confirm that ≈ percent of these were CLQs. The authors compared the CLQswith a luminosity- and redshift-matched, lesser-variable con-trol sample and again found that CLQs on average havelower L / L Edd than their less-variable counterparts. Both stud-ies suggested that EVQs and CLQs represent the extremesof a tail of ‘normal’ quasar variability. At the far range ofthis tail, some sources exhibit nearly an order of magni-tude change in optical flux over a baseline of ∼ years.Compared to many of these changing-look AGN, the con-tinuum flux change we observed during our monitoring ofSDSS J2232 − log ( L / L Edd ) = − is slightlyhigher than the peaks of the distributions of CLQs andEVQs (which occur at log ( L / L Edd ) ≈ − . , see Figure 6 ofMacLeod et al. 2018) although it is consistent with the rangeof values for all of the populations shown (CLQs, EVQs,the less-variable control sample and all 105783 of the SDSSDR7 quasars). Assuming the bolometric flux of the sourcedecreases proportionally to the observed optical, we can es-timate that the accretion rate of SDSS J2232 − ∼ a few per cent of Eddington in the faint state.Elitzur & Ho (2009) proposed a disc wind model ofthe BLR in which AGN with a very low L / L Edd are un-able to support a BLR. After studying a sample of low-luminosity AGN, they determined that the BLR disap-pears when the AGN luminosity drops below a criticalvalue, L AGN (cid:46) × ( M BH / M (cid:12)) / erg s − . MacLeodet al. (2018) found that their CLQs were distributed closeto this critical value and likely dropped below it in theirfaint state, naturally explaining the disappearance of thebroad emission lines. Whilst the broad Balmer emission linesin SDSS J2232 − u . Its UV flux in thisepoch was ≈ times fainter than when it was observedby XMM-Newton . Assuming the bolometric flux was also4 times fainter, its luminosity in 2000 was ≈ × erg s − .For the BH mass of SDSS J2232 − ≈ × erg s − , sothe broad lines ought to have been visible even in this deepminimum. Therefore, we suggest that SDSS J2232 − − MNRAS , 1–16 (2019) xtreme variability of SDSS J2232 − the longer-wavelength fluxes for host galaxy contamination,we show in Figure 6 that the fractional variabilities in theblue and red are similar (i.e. there is no significant colourchange). In Figure 3 we show that the shape of the RMSspectrum is very similar to that of the mean spectrum atthe shorter wavelengths less affected by host galaxy contam-ination. We note that Wilhite et al. (2005) used a sampleof higher-redshift quasars than SDSS J2232 − z > . )so they probed further into the rest frame UV than we do.The authors show that there is a spectral break in the vari-ability of their sample around 2500 ˚A in the rest frame,with wavelengths shorter than this being more strongly vari-able. We may not see evidence of a spectral shape changein SDSS J2232 − − Our observing campaign was fortunate to haverecorded a dramatic dimming and brightening eventof SDSS J2232 − − − Our recent optical photometric and spectroscopic monitor-ing campaign on the hypervariable AGN SDSS J2232 − ≈ flux change over four years. Whilst the observedvariability of the source is modest compared to that seen inchanging-look AGN, it is extreme compared to the broaderAGN population. We have been able to demonstrate thatvariable obscuration does not explain the observed spectralchanges, nor does it fit the observed timescales for variabil-ity in the optical or near-infrared. An intrinsic change in theAGN luminosity is therefore a likelier explanation, although the observed changes are much more rapid than the the-oretical accretion disc viscous timescale. SDSS J2232 − ACKNOWLEDGEMENTS
DK acknowledges the receipt of a UK Science andTechnology Facilities Council (STFC) studentship(ST/N50404X/1). DK and MJW acknowledge supportfrom the STFC grant ST/P000541/1. Thanks to MichaelFausnaugh for his assistance in the use of mapspec . Thanksalso to Ra’ad Mahmoud, Raj Sathyaprakash, Chris Done,David Rosario and Brad Peterson for useful discussions.In this research we have made use of the following: • data from the William Herschel Telescope, operated onthe island of La Palma by the Isaac Newton Group of Tele-scopes in the Spanish Observatorio del Roque de los Mucha-chos of the Instituto de Astrof´ısica de Canarias; • data from the Liverpool Telescope, operated on the is-land of La Palma by Liverpool John Moores University inthe Spanish Observatorio del Roque de los Muchachos of theInstituto de Astrof´ısica de Canarias with financial supportfrom the STFC; • observations obtained at the MMT Observatory, a jointfacility of the Smithsonian Institution and the University ofArizona; • data products from the WISE mission, which is a jointproject of the University of California, Los Angeles, and theJet Propulsion Laboratory/California Institute of Technol-ogy, and
NEOWISE , which is a project of the Jet PropulsionLaboratory/California Institute of Technology.
WISE and
NEOWISE are both funded by the National Aeronauticsand Space Administration (NASA); • data from and software developed for XMM-Newton ,an ESA science mission with instruments and contributionsdirectly funded by ESA Member States and NASA; • observations made with the NASA Galaxy EvolutionExplorer . GALEX is operated for NASA by the CaliforniaInstitute of Technology under NASA contract NAS5-98034; • data products from the Two Micron All Sky Survey(2MASS), which is a joint project of the University of Mas-sachusetts and the Infrared Processing and Analysis Cen-ter/California Institute of Technology, funded by NASA andthe National Science Foundation; • data from SDSS: funding for the SDSS and SDSS-IIhas been provided by the Alfred P. Sloan Foundation, theParticipating Institutions, the National Science Foundation,the U.S. Department of Energy, the National Aeronauticsand Space Administration, the Japanese Monbukagakusho,the Max Planck Society, and the Higher Education FundingCouncil for England. The SDSS Web Site is ; • data from Pan-STARRS-1: the Pan-STARRS-1 Surveys(PS1) have been made possible through contributions of theInstitute for Astronomy, the University of Hawaii, the Pan-STARRS Project Office, the Max-Planck Society and its MNRAS000
NEOWISE are both funded by the National Aeronauticsand Space Administration (NASA); • data from and software developed for XMM-Newton ,an ESA science mission with instruments and contributionsdirectly funded by ESA Member States and NASA; • observations made with the NASA Galaxy EvolutionExplorer . GALEX is operated for NASA by the CaliforniaInstitute of Technology under NASA contract NAS5-98034; • data products from the Two Micron All Sky Survey(2MASS), which is a joint project of the University of Mas-sachusetts and the Infrared Processing and Analysis Cen-ter/California Institute of Technology, funded by NASA andthe National Science Foundation; • data from SDSS: funding for the SDSS and SDSS-IIhas been provided by the Alfred P. Sloan Foundation, theParticipating Institutions, the National Science Foundation,the U.S. Department of Energy, the National Aeronauticsand Space Administration, the Japanese Monbukagakusho,the Max Planck Society, and the Higher Education FundingCouncil for England. The SDSS Web Site is ; • data from Pan-STARRS-1: the Pan-STARRS-1 Surveys(PS1) have been made possible through contributions of theInstitute for Astronomy, the University of Hawaii, the Pan-STARRS Project Office, the Max-Planck Society and its MNRAS000 , 1–16 (2019) D. Kynoch et al. participating institutes, the Max Planck Institute for As-tronomy, Heidelberg and the Max Planck Institute for Ex-traterrestrial Physics, Garching, The Johns Hopkins Uni-versity, Durham University, the University of Edinburgh,Queen’s University Belfast, the Harvard-Smithsonian Cen-ter for Astrophysics, the Las Cumbres Observatory GlobalTelescope Network Incorporated, the National Central Uni-versity of Taiwan, the Space Telescope Science Institute,the National Aeronautics and Space Administration underGrant No. NNX08AR22G issued through the Planetary Sci-ence Division of the NASA Science Mission Directorate, theNational Science Foundation under Grant No. AST-1238877,the University of Maryland, and Eotvos Lorand University(ELTE); • data obtained from the SuperCOSMOS ScienceArchive, prepared and hosted by the Wide Field Astron-omy Unit, Institute for Astronomy, University of Edinburgh,which is funded by the UK STFC; • the SpectRes spectral resampling tool (Carnall 2017); • Doug Welch’s Excellent Absorption Law Calculator( ); • Ned Wright’s Cosmology Calculator (Wright 2006).
REFERENCES
Almeyda T., Robinson A., Richmond M., Vazquez B., NikuttaR., 2017, ApJ, 843, 3Arnaud K. A., 1996, in Jacoby G. H., Barnes J., eds, AstronomicalSociety of the Pacific Conference Series Vol. 101, AstronomicalData Analysis Software and Systems V. p. 17Bentz M. C., et al., 2013, ApJ, 767, 149Bohlin R. C., Savage B. D., Drake J. F., 1978, ApJ, 224, 132Bruce A., et al., 2017, MNRAS, 467, 1259Bruhweiler F., Verner E., 2008, ApJ, 675, 83Cannon R. D., 1975, Proceedings of the Astronomical Society ofAustralia, 2, 323Cardelli J. A., Clayton G. C., Mathis J. S., 1989, ApJ, 345, 245Carnall A. C., 2017, preprint, ( arXiv:1705.05165 )Chambers K. C., et al., 2016, preprint, ( arXiv:1612.05560 )Collinson J. S., et al., 2018, MNRAS, 474, 3565Dexter J., Begelman M. C., 2019, MNRAS, 483, L17Dickey J. M., Lockman F. J., 1990, ARA&A, 28, 215Done C., Davis S. W., Jin C., Blaes O., Ward M., 2012, MNRAS,420, 1848Elitzur M., Ho L. C., 2009, ApJ, 701, L91Elvis M., et al., 2012, ApJ, 759, 6Fausnaugh M. M., 2017, PASP, 129, 024007Gaskell C. M., 2017, MNRAS, 467, 226Gezari S., et al., 2017, ApJ, 835, 144Goodrich R. W., 1995, ApJ, 440, 141Hambly N. C., et al., 2001, MNRAS, 326, 1279Jin C., Ward M., Done C., 2012, MNRAS, 422, 3268Katebi R., et al., 2018, preprint, ( arXiv:1811.03694 )Kawaguchi T., Mori M., 2010, ApJ, 724, L183Kubota A., Done C., 2018, MNRAS, 480, 1247LaMassa S. M., et al., 2015, ApJ, 800, 144Landt H., Elvis M., Ward M. J., Bentz M. C., Korista K. T.,Karovska M., 2011, MNRAS, 414, 218Lawrence A., 2018, Nature Astronomy, 2, 102Lawrence A., et al., 2016, MNRAS, 463, 296Lu K.-X., Zhao Y., Bai J.-M., Fan X.-L., 2019, MNRAS, 483,1722MacLeod C. L., et al., 2012, ApJ, 753, 106MacLeod C. L., et al., 2016, MNRAS, 457, 389MacLeod C. L., et al., 2018, preprint, ( arXiv:1810.00087 ) Magnier E. A., et al., 2013, ApJS, 205, 20Maiolino R., Marconi A., Salvati M., Risaliti G., Severgnini P.,Oliva E., La Franca F., Vanzi L., 2001, A&A, 365, 28Martin D. C., et al., 2005, ApJ, 619, L1McHardy I. M., et al., 2018, MNRAS, 480, 2881Mej´ıa-Restrepo J. E., Trakhtenbrot B., Lira P., Netzer H.,Capellupo D. M., 2016, MNRAS, 460, 187Noda H., Done C., 2018, MNRAS, 480, 3898Osterbrock D. E., Ferland G. J., 2006, Astrophysics of gaseousnebulae and active galactic nuclei, 2nd edn. University ScienceBooks, Mill Valley, CAParker M. L., et al., 2016, MNRAS, 461, 1927Pei L., et al., 2017, ApJ, 837, 131Peterson B. M., et al., 2004, ApJ, 613, 682Polletta M., et al., 2007, ApJ, 663, 81Ross N. P., et al., 2018, MNRAS, 480, 4468Ruan J. J., et al., 2016, ApJ, 826, 188Rumbaugh N., et al., 2018, ApJ, 854, 160Runnoe J. C., et al., 2016, MNRAS, 455, 1691Schmidt K. B., Rix H.-W., Shields J. C., Knecht M., Hogg D. W.,Maoz D., Bovy J., 2012, ApJ, 744, 147Sheng Z., Wang T., Jiang N., Yang C., Yan L., Dou L., Peng B.,2017, ApJ, 846, L7Skrutskie M. F., et al., 2006, AJ, 131, 1163Stern D., et al., 2018, ApJ, 864, 27Sun M., et al., 2015, ApJ, 811, 42Vanden Berk D. E., et al., 2001, AJ, 122, 549Wang J., Xu D. W., Wei J. Y., 2018, ApJ, 858, 49Wilhite B. C., Vanden Berk D. E., Kron R. G., Schneider D. P.,Pereyra N., Brunner R. J., Richards G. T., Brinkmann J. V.,2005, ApJ, 633, 638Wright E. L., 2006, PASP, 118, 1711Wright E. L., et al., 2010, AJ, 140, 1868Yang Q., et al., 2018, ApJ, 862, 109Zhu D., Sun M., Wang T., 2017, ApJ, 843, 30Zuo W., Wu X.-B., Liu Y.-Q., Jiao C.-L., 2012, ApJ, 758, 104van Groningen E., Wanders I., 1992, PASP, 104, 700
APPENDIX A: OPTICAL SPECTRAL FITTINGRESULTS
This paper has been typeset from a TEX/L A TEX file prepared bythe author. MNRAS , 1–16 (2019) xtreme variability of SDSS J2232 − Table A1.
Balmer, [O iii ] and [N ii ] emission line measurements H α (cid:122) (cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32) (cid:125)(cid:124) (cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32) (cid:123) Date Scale ∆ v vb W b W n f vb × − f b × − f n × − f tot × − EW vb + b + ±
700 4460 ±
50 480 ±
10 3 ± . ± .
08 1 . ± . . ± . ± + ±
700 4530 ±
60 550 ±
10 3 ± . ± .
08 1 . ± . . ± . ± + ±
800 4280 ±
50 510 ±
10 9 ± . ± . . ± .
08 6 . ± . ± + ±
700 4200 ±
60 480 ±
10 6 ± . ± .
09 1 . ± .
09 5 . ± . ± + ±
500 4300 ±
50 570 ±
20 5 ± . ± . . ± . . ± . ± (cid:54)
700 4600 ±
100 490 ±
10 3 ± . ± . . ± .
06 5 . ± . ± + ±
500 4510 ±
40 540 ±
10 3 ± . ± .
05 1 . ± .
05 5 . ± . ± + ± ±
100 470 ±
10 4 ± . ± . . ± .
07 5 . ± . ± + ±
90 700 ±
200 1 . ± . ± . ± . ± ± + ±
500 500 ±
100 2 . ± . ± . ± . ± ± + ±
800 4560 ±
80 490 ±
30 4 ± . ± . . ± . . ± . ± Table A2.
Balmer, [O iii ] and [N ii ] emission line measurements (continued) [ N ii ] λ H β [ O iii ] λ (cid:122) (cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32) (cid:125)(cid:124) (cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32) (cid:123) (cid:122) (cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32) (cid:125)(cid:124) (cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32) (cid:123) (cid:122) (cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32) (cid:125)(cid:124) (cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32) (cid:123) Date f × − f vb × − f b × − f n × − f tot × − EW vb + b BD vb + b f × − EW2013-06-10 . ± . ± . ± .
09 2 . ± . . ± . ± . ± . . ± . ± . ± . ± . ± .
04 2 . ± . . ± . ±
10 2 . ± . . ± . ± . ± . ± . ± .
05 2 . ± . . ± . ±
20 3 . ± . . ± . ± . ± . ± . ± .
05 2 . ± . . ± . ±
20 3 . ± . . ± . ± . ± . ± . ± .
03 2 . ± . . ± . ±
20 4 . ± . . ± . ± . ± . ± . ± .
09 2 . ± . . ± . ±
10 3 . ± . . ± . ± . ± . ± . ± .
06 2 . ± .
08 1 . ± . ±
10 3 . ± . . ± . ± . ± . ± . ± .
08 2 . ± . . ± . ±
20 3 . ± . . ± . ± ± ± . ± . . ± . . ± . ±
20 3 . ± . . ± . ± ± ± . ± . . ± . . ± . ±
20 4 . ± . . ± . ± . ± . ± . ± .
09 2 . ± . . ± . ±
10 3 . ± . . ± . ± ‘Scale’ is the flux scaling factor applied to each spectrum, including both internal and absolute scalings (see Section 2.2.2 in thetext). Subscripts ‘vb’, ‘b’ and ‘n’ refer to the very broad, broad and narrow emission line components, respectively and ‘tot’ isthe total. ∆ v vb is the velocity offset (in km s − ) of the very broad emission line components relative to the narrower components;positive values indicate a redward offset. Fluxes f in erg s − cm − ; widths ‘W’ are FWHM in km s − and equivalent widths‘EW’ are in ˚A. ‘BD’ is the Balmer decrement H α / H β . Table A3. Mg ii emission line measurementsDate W vb f vb × − W b f b × − W tot f tot × − EW tot ±
800 2 . ± . ±
300 1 . ± . ±
400 3 . ± . ± ± . ± . ±
500 1 . ± . ±
500 3 . ± . ± ±
600 2 . ± . ±
100 1 . ± . ±
200 4 . ± . ± ±
700 1 . ± . ±
200 2 . ± . ±
200 3 . ± . ± ±
600 2 . ± . ±
200 1 . ± . ±
300 3 . ± . ± ± . ± . ±
400 1 . ± . ±
500 3 . ± . ± ±
500 2 . ± . ±
200 1 . ± . ±
200 4 . ± . ± ±
700 2 . ± . ±
200 2 . ± . ±
200 4 . ± . ± ± . ± . ±
200 2 . ± . ±
300 5 . ± . ± ± . ± . ±
300 1 . ± . ±
500 4 . ± . ± ± . ± . ±
400 1 . ± . ±
500 3 . ± . ± Subscripts ‘vb’ and ‘b’ refer to the very broad and broad emission line components, respectively and ‘tot’ isthe total. Fluxes f in erg s − cm − ; widths ‘W’ are FWHM in km s − and equivalent widths ‘EW’ are in ˚A.MNRAS000