Long-term optical spectroscopic variations in blazar 3C 454.3
Krzysztof Nalewajko, Alok C. Gupta, Mai Liao, Krzysztof Hryniewicz, Maitrayee Gupta, Minfeng Gu
aa r X i v : . [ a s t r o - ph . H E ] A ug Astronomy & Astrophysics manuscript no. 35904corr˙arxiv c (cid:13)
ESO 2020August 4, 2020
Long-term optical spectroscopic variations in blazar 3C 454.3
Krzysztof Nalewajko , Alok C. Gupta , Mai Liao , , Krzysztof Hryniewicz , Maitrayee Gupta , andMinfeng Gu Nicolaus Copernicus Astronomical Center, Polish Academy of Sciences, Bartycka 18, 00-716 Warsaw, Polande-mail: [email protected] Aryabhatta Research Institute of Observational Sciences (ARIES), Manora Peak, Nainital 263002, India Key Laboratory for Research in Galaxies and Cosmology, Shanghai Astronomical Observatory, Chinese Academy ofSciences, Shanghai 200030, China University of Chinese Academy of Sciences, 19A Yuquanlu, Beijing 100049, ChinaAugust 4, 2020
ABSTRACT
Aims.
Characterisation of the long-term variations in the broad line region in a luminous blazar, where Comptonisationof broad-line emission within a relativistic jet is the standard scenario for production of γ -ray emission that dominatesthe spectral energy distribution. Methods.
We analysed ten years of optical spectroscopic data from the Steward Observatory for the blazar 3C 454.3,as well as γ -ray data from the Fermi Large Area Telescope (LAT). The optical spectra are dominated by a highlyvariable non-thermal synchrotron continuum with a prominent Mg II broad emission line. The line flux was obtainedby spectral decomposition including significant contribution from the Fe II pseudo-continuum. Three methods wereused to characterise variations in the line flux: (1) stacking of the continuum-subtracted spectra, (2) subtracting therunning mean light curves calculated for different timescales, and (3) evaluating potential time delays via the discretecorrelation function (DCF). Results.
Despite very large variations in the γ -ray and optical continua, the line flux changes only moderately ( < . ∼
600 days. If this time delay results from the reverberation ofpoorly constrained accretion disc emission in both the broad-line region (BLR) and the synchrotron emitting blazarzone within a relativistic jet, we would obtain natural estimates for the BLR radius R BLR , MgII & .
28 pc and for thesupermassive black hole mass M SMBH ∼ . × M ⊙ . We did not identify additional examples of short-term ‘flares’ ofthe line flux, in addition to the previously reported case observed in 2010. Key words.
Galaxies: active – quasars: emission lines – quasars: individual: 3C 454.3
1. Introduction
The cores of active galaxies exhibit a complex mix-ture of energetic physical phenomena (for review seeMadejski & Sikora 2016, and references therein). Super-massive black holes ( M SMBH ∼ M ⊙ ) accrete surround-ing gas, often forming radiatively efficient accretion discsthat produce intense radiation fields (quasar emission; L disc ∼ erg s − ) that ionise the surrounding gas,resulting in emission lines broadened by rapid gas mo-tions ( v > − ). In some cases, magnetic fluxaccumulated on the black hole produces a pair of rela-tivistic collimated outflows called jets that produce non-thermal radiation beams (blazar emission), the luminosityof which can be greatly enhanced (by a factor of ∼ to L jet ∼ erg s − ) by special-relativistic effects forobservers located along one of the jets. A class of ac-tive galaxies called flat spectrum radio quasars (FSRQs)combines the characteristics of blazars, where broad non-thermal spectral components dominate the radio, infrared,optical, X-ray, and gamma-ray bands, and those of quasars,where thermal emission of accretion disc peak in the rest-frame UV, and broad emission lines (BELs) produced in so-called broad line region (BLR) contribute to the rest-frame optical and/or UV. The BELs constitute an impor-tant source of soft radiation that strongly interacts withrelativistic jets, with the potential of producing the bulk ofobserved γ -ray emission by inverse Compton (IC) scatter-ing off ultra-relativistic electrons (Sikora et al. 1994), andof absorbing a part of the γ -ray spectrum in the processof photon-photon pair production (Blandford & Levinson1995). The understanding of the BLR properties, espe-cially its geometry and dynamics, while essential for mak-ing a self-consistent picture of FSRQs, is still rather basic(Tavecchio & Ghisellini 2008; Czerny & Hryniewicz 2011;Tavecchio & Ghisellini 2012; Janiak et al. 2015).A primary example of an FSRQ of exceptional gamma-ray luminosity (up to 10 erg s − Abdo et al. 2011)and large-amplitude time variability is 3C 454.3 ( z =0 . Article number, page 1 of 8 &A proofs: manuscript no. 35904corr˙arxiv
Donnarumma et al. 2009; Vercellone et al. 2010, 2011) andby Fermi-LAT (Abdo et al. 2009a; Bonning et al. 2009;Ackermann et al. 2010; Bonnoli et al. 2011; Raiteri et al.2011; Wehrle et al. 2012; Gaur et al. 2012; Jorstad et al.2013; Kushwaha et al. 2017; Gupta et al. 2017).As one of the brightest gamma-ray blazars of the Fermiera, 3C 454.3 was selected for regular optical spectro-scopic monitoring by the Steward Observatory (Smith et al.2009). Based on a part of this dataset (2008-2011), aflare-like variability of the Mg II broad emission lineflux in 2010 November (MJD ∼ σ , in addition tomore significant variations in the H γ line (Isler et al. 2013,2015). Here, we analyse ten years of spectroscopic data fromSteward Observatory comprising 564 observations, an un-precedented dataset in terms of long-term regular and fre-quent monitoring. We find a hint (statistical significanceof ∼ . σ ) of modulation in the Mg II line luminosity on atimescale of about two years in response to large-amplitudevariability of the optical and gamma-ray continua that werenot noticed by previous studies. The line luminosity is ei-ther anti-correlated with the optical continuum luminosityor delayed by ∼
300 source-frame days. This would be onlythe second case of reverberation mapping in a blazar afterthe case of 3C 273 (PG 1226+023; z = 0 .
2. Data
Regular optical photometric and spectropolarimetric obser-vations of a sample of gamma-ray bright blazars are per-formed by the team of the University of Arizona StewardObservatory as support programme for the Fermi Gamma-Ray Space Telescope . Moderate-resolution ( R ∼ − λ obs = (4000 − V magnitudes of nearby standardstars. Technical details of the instrument and data reduc-tion procedures are described in Smith et al. (2009).We analysed all 564 photometrically calibrated spec-troscopic observations over the period of ten years (2008-2018; MJD = 54743 − A ( V ) = 0 .
33 mag and E ( B − V ) = 0 .
105 mag with R V = 3 . λ src = 2800˚A redshifted to λ obs = 5200˚A. It is imposed on a continuum consisting ofnon-thermal synchrotron emission component of the blazarand of thermal accretion disc emission component of thequasar, as well as an Fe II pseudo-continuum. As the pri-mary analysis method, we performed spectral decomposi-tion of the Mg II line complex following the approach ofShen et al. (2011). The Mg II emission was modelled with http://james.as.arizona.edu/~psmith/Fermi up to three Gaussian components, with the Fe II templateadopted after (Vestergaard & Wilkes 2001), and the contin-uum was fitted within the source-frame spectral windowsof (2200 − − obs of theMg II line was calculated from the superposed profiles of thefitted Gaussian components, and it was corrected for instru-mental line broadening using FWHM int = (FWHM − FWHM ) / with the adopted source-frame instrumen-tal resolution of FWHM instr ≃ − . In order to es-timate observational flux uncertainty, we calculated the rmsvalue of the spectra over the source-frame spectral windowof (3000 − − − − − × − erg s − cm − ˚A − . After analysingthe strongest outlier values among estimates of the Mg II+ Fe II line flux, we decided to introduce an upper limitto the rms value: rms max = 7 × − erg s − cm − ˚A − .Applying this criterion, we rejected 107 (19%) of the mostnoisy individual photometric spectra from further analysis.This selection was also applied to the results of spectraldecomposition analysis. Gamma-ray data from the Fermi Large Area Telescope(Atwood et al. 2009) were extracted from the region ofinterest (ROIs) of radius 10 ◦ centred on the position of3C 454.3. We used the ScienceTools software package,version v11r5p3 , to perform the maximum likelihood anal-ysis, fitting a
PowerLaw2 spectral model over the energy
Article number, page 2 of 8alewajko et al.: Optical spectroscopic variations in blazar 3C 454.3 -7 -6 -5 γ-ray flux at E>100 MeV [ph/s/cm ] -15 -14 flux density f λ at λ_src = 3000Å [erg/s/cm /Å] decompsimplLT135000 5500 6000 6500 7000 7500 8000−2.5−2.0−1.5−1.0−0.50.0 s−ec0.al inde2 α (f λ ∝ λ α ) decompsimpl5000 5500 6000 6500 7000 7500 800010 -14 -13 Mg II line flux [erg/s/cm ] decomp simpl (MgII+FeII) LT13 (MgII+FeII) LT13 I135000 5500 6000 6500 7000 7500 8000200025003000350040004500 Mg II line FWHM [km/s] -17 -16 rms of continuum [erg/s/cm /Å] Fig. 1.
Light curves of blazar 3C 454.3. From the top, the first panel shows the γ -ray photon flux for photon energies above 100 MeVobtained from the Fermi-LAT data in 10-day bins; the second panel shows the optical continuum at source-frame 3000˚ A , obtainedindependently using two analysis methods (spectral decomposition, blue circles; simplified power-law fit, red/grey crosses); the thirdpanel shows the corresponding values of the spectral index α such that f λ ∝ λ α ; the fourth panel shows estimates of the spectrallyintegrated Mg II line flux, either separated from the Fe II continuum (decomposition method) or not (simplified method); the fifthpanel shows the source-frame FWHM width of Mg II line obtained in the decomposition method and corrected for instrumentalbroadening; the sixth panel shows the rms of continuum subtracted spectra calculated over the continuum fitting spectral windows,with the limiting value of 7 × − erg / s / cm / ˚A indicated with the dashed line. Results reported in Le´on-Tavares et al. (2013)(Steward) and Isler et al. (2013) (SMARTS) are indicated in the second and fourth panels. Periods of interest for spectral stackingare marked with vertical stripes or lines. Article number, page 3 of 8 &A proofs: manuscript no. 35904corr˙arxiv range 100 MeV – 100 GeV, using the P8R2 SOURCE V6 response function, maximum zenith angle of 100 ◦ , and asource model including all background sources from the3FGL catalogue (Acero et al. 2015) up to the angular sep-aration of 25 ◦ , with the detection criteria of TS >
10 and N pred > The light curve was obtained over the period ofMJD ≃ −
3. Results
Figure 1 shows γ -ray and optical light curves of 3C 454.3over the period of 10 years (2008 - 2018; MJD 54700 -58300). The γ -ray data show large-amplitude (factor ∼ ≃ f ( E >
100 MeV) ≃ × − ph s − cm − on MJD 55520(Abdo et al. 2011) are not included here. However, takingthem into account, the overall variability amplitude exceedsa factor of 1000.Simultaneous variations in the optical continuum areroughly correlated with the γ -ray signal (cf. Bonning et al.2009; Gaur et al. 2012; Kushwaha et al. 2017; Gupta et al.2017), but with significantly lower amplitude (a factor of ∼ γ -ray flux scalesin at least quadratic relation to the optical continuum(Bonnoli et al. 2011). This is confirmed by the scatter dia-gram of γ -ray photon flux versus optical flux density shownin the top left panel of Figure 2. In the lowest gamma-raystate at MJD ∼ γ -ray fluxes < − ph s − cm − ,the optical λ src = 3000˚ A continuum flux density does notdecrease below 10 − erg s − cm − ˚A, or the level at whichcontribution of the thermal accretion disc emission is ex-pected to become dominant. The optical spectral index α (defined as f λ ∝ λ α ) is generally lower during the low γ -ray and optical state (see the middle left panel of Figure 2),which means a redder-when-brighter trend consistent withthe presence of accretion disc component (also known asthe big blue bump, e.g. Gu et al. 2006; Villata et al. 2006;Raiteri et al. 2007; Bonning et al. 2009), although the valueof α remains highly variable.We present two different estimates of the optical con-tinuum; one is based on the spectral decomposition (bluecircles in Figure 1) including the spectral template for Fe IIemission and the other is a simplified approach (red crossesin Figure 1) in which Fe II emission is not separated fromthe Mg II emission. The normalisations of the optical con-tinuum at λ src = 3000˚ A and the spectral indices α ob-tained with these two methods are consistent. On the otherhand, the fluxes of the broad Mg II emission line integratedover λ obs appear discrepant in the fourth panel of Figure1 because of the inclusion of Fe II in simplified measure-ments. However, for the initial period (MJD < We adopt a less strict detection criterion than the usual TS >
25 in order to cover the period of the lowest gamma-ray flux atMJD = 55800 − -15 -14 -7 -6 -5 γ - r a y p h o t o n f l u x [ p h / s / c m ] -15 -14 −2.5−2.0−1.5−1.0−0.50.0 o p t i c a l s e c t r a l i n d e x −3.0 −2.5 −2.0 −1.510 -7 -6 -5 −3.0 −2.5 −2.0 −1.5−2.5−2.0−1.5−1.0−0.50.010 -15 -14 optical flux density [erg/s/cm /Å]0.50.60.70.80.91.01.1 M g II + F e II li n e f l u x [ − e r g/ s / c m ] −3.0 −2.5 −2.0 −1.5γ-ray photon index0.50.60.70.80.91.01.1 Fig. 2.
Scatter diagrams between γ -ray photon flux, γ -ray pho-ton index, optical continuum flux density, optical spectral index,and integrated Mg II + Fe II line flux measured simultaneouslywithin ± Table 1 in Le´on-Tavares et al. (2013) We find that our de-composition results are consistent with the Mg II fluxes ofLe´on-Tavares et al. (2013), and that our simplified resultsare consistent with their Mg II+Fe II fluxes. On the otherhand, the Mg II flux measurements based on the SMARTSobservations, as reported by Isler et al. (2013, 2015), aresystematically lower (a factor of ∼
3) than the Mg II fluxesobtained from the Steward data. A comparison of these twodatasets is discussed in Isler et al. (2013, 2015); however,the relative spectral calibration between these two instru-ments has not been fully explained.The Mg II line flux light curve reveals several outlierswith respect to the typical values, and occasional differenceswith the results of Le´on-Tavares et al. (2013). Observationat MJD 54771 yields both the line fluxes (either decom-posed Mg II or simplified Mg II+Fe II) and the continuumlevel that are a factor of ∼ The (1 + z ) K-correction was applied to the line fluxes re-ported by Le´on-Tavares et al. (2013); here its effect is removedfor the green symbols shown in Figure 1.Article number, page 4 of 8alewajko et al.: Optical spectroscopic variations in blazar 3C 454.3 tematic difference is also seen in the photometric spectrum.A similar outlier can be seen at MJD 56308, with continuumand line fluxes increased by factor ∼ .
5; unlike the case ofMJD 54771, this spectrum is affected by noise despite thesame exposure time (960 s). Several observations aroundMJD 55295 suggest line fluxes lower than the typical val-ues, and also lower than the results of Le´on-Tavares et al.(2013); these observations are characterised by relativelyshort exposure times (320 s), and hence high statistical fluxuncertainty.
Le´on-Tavares et al. (2013) reported an episode of short-term variation (flare) in the Mg II line flux (and in the Fe IIemission) during an optical flare centred at MJD 55512.This epoch is indicated in Figure 1 by the green dot-ted lines. We confirm that both Mg II and Mg II+Fe IIline fluxes are elevated during the week-long window cen-tred at MJD 55512, compared with neighbouring windows.Because of wide gaps in the observations, the variabilitytimescale is uncertain: it could be closer to weekly thandaily. However, all individual spectra obtained over thatperiod are relatively noisy; their rms values are above ouradopted upper limit rms max (see bottom panel of Fig-ure 1). We searched for more examples of such flares be-yond MJD 55900, and we have not found any. AroundMJD 56832, there was an even brighter flare of the opticalcontinuum (Kushwaha et al. 2017), but it was not accom-panied by any clear increase in the Mg II line flux (and theindividual spectra obtained over that week are even morenoisy than those around MJD 55512).In Figure 3, we present stacked continuum-subtractedspectra evaluated over two short observational windows(without any selection for the rms values): MJD = 55512 ±
10 (green line) and MJD = 56832 ±
10 (magenta line).These line profiles are compared with a ten-year average (thick black line). Differences between short-term stackedspectra and the ten-year average are shown in the lowerpanel of Figure 3. We performed two-sample Kolmogorov-Smirnov (K-S) tests for the difference between stacked spec-tra. This test suggests that the line profile stacked duringthe MJD 55512 flare reported by Le´on-Tavares et al. (2013)is clearly different from the ten-year average ( p ∼ − ;6 σ ), while the line profile stacked during the second opti-cal continuum flare around MJD 56832 is only different atthe level of 2 σ ( p ≃ . The left panel of Figure 4 shows again the light curve ofMg II line luminosity obtained from the spectral decomposi-tion method (corresponding to the blue circles shown in thefourth panel of Figure 1, we take the logarithm and subtractthe mean value). Due to our selection for rms < rms max , theshort-term variations discussed in the previous subsectionare largely excluded. We suggest that this light curve showsa systematic modulation on long timescales ( ∼
100 days).A power spectral density (PSD) analysis showed that varia- Based on 273 individual spectra characterised by rms < × − erg s − cm − ˚A − calculated over the continuum fittingspectral windows. f λ [ − e r g/ s / c m / Å ] stacked {spectra - continuum} All MJD; rms<0.04MJD 55070 - 55315; rms<0.03MJD 55690 - 55980; rms<0.03MJD 55512 +/- 10MJD 56832 +/- 10−0.04 −0.02 0.00 0.02 0.04 0.06 0.08MgII (2798.2Å) −elocit. o set [c]−0.1510.1010.050.000.050.100.15 f λ [ − e r g/ s / c m / Å ] stacked {spectra - continuum} - {All MJD} f λ [ − e r g/ s / c m / Å ] Mg II + Fe II model
All MJD; rms<0.04Mg II (1)Mg II (2)Mg II (3)Fe IIresidual
Fig. 3.
Upper panel: Observed source-frame spectra stackedover different observational periods after subtraction of power-law continuum. Middle panel: Differences between spectrastacked over limited observational periods and the spectrumstacked over all available observations (line styles correspond tothose in the upper panel). Lower panel: model for the spectraldecomposition of the Mg II line and theoretical Fe II pseudo-continuum template. The spectral windows for continuum eval-uation are indicated with grey stripes. tions in the Mg II line luminosity are consistent with the rednoise, with no sign of quasi-periodic oscillations (QPOs).In order to evaluate the significance of this long-termmodulation, we calculate a running mean light curve usinguniform time bins of fixed timescale τ . As an example, inthe left panel of Figure 4, we show the running mean lightcurve for τ = 100 d (solid red line), and in the right panel ofthat figure we show the measured Mg II luminosities aftersubtracting the running mean. We then perform a series oftwo-sample K-S tests between the measured and runningmean subtracted Mg II light curves for a wide range ofrunning mean timescales τ . Figure 5 shows the dependenceof the p -values for the null hypothesis (that subtracting arunning mean does not modify the distribution of Mg II Article number, page 5 of 8 &A proofs: manuscript no. 35904corr˙arxiv
Fig. 4.
Left panel: Light curve of Mg II line luminositylog L MgII obtained with the spectral decomposition method,with the mean value subtracted (blue); running mean calculatedfor the minimum timescale of 100 d (red). Right panel: Residualof the light curve shown in the left panel with the running meansubtracted. Periods of interest for spectral stacking are markedwith vertical stripes or lines (cf. Figure 1). running mean time scale [d]10 -2 -1 K-S test p-valuep-value
Fig. 5.
Probability of rejecting the hypothesis that subtractinga running mean does not affect the distribution of Mg II lineluminosities calculated from two-sample K-S tests between themeasured light curve of log L MgII (left panel of Figure 4) andthe running mean subtracted light curve (right panel of Figure4) as function of the running mean timescale. luminosities) on the running mean timescale τ . The basicresult is that p < . τ <
200 d, hence long-term varia-tions have only moderate statistical significance (1 . σ ).We also calculated the discrete correlation functions(DCF; Edelson & Krolik 1988) in order to estimate thecharacteristic time lags between partially correlated sig-nals. Noting the fundamentally different nature of vari-ations in the Mg II line luminosity (moderate departuresfrom the mean value) and the optical continuum (large-amplitude–order of magnitude variations without a clearmean value), we cross-correlated the logarithms of mea-sured flux values using the time lag bins of 40 days. As areference, we first calculated the DCF between the γ -rayand optical continua (left panel of Figure 6). This DCF hasa clear positive peak of DCF ≃ .
73 at ∆ t ≃
0. In order to Very similar results were obtained with the ZDCF algorithmof Alexander (1997). −1000 −500 0 500 1000t(γ) - t(3000Å) [d]−1.0−0.50.00.51.0
DCF (log f_γ vs. log f_3000 Å) −1000 −500 0 500 1000t(MgII) - t(3000Å) [d]−1.0−0.50.00.51.0 DCF (log f_MgII vs. log f_3000 Å) Fig. 6.
Discrete correlation functions calculated between thelogarithms of the ( >
100 MeV) γ -ray photon flux and the opticalcontinuum flux density (left panel), and between the logarithmsof the Mg II line flux and the optical continuum (right panel). estimate the most likely value of time lag, we fitted a Gaus-sian function over a limited range of time lags (red dottedlines), finding a centroid of ∆ t = 60 ±
41 d and dispersionof σ (∆ t ) = 178 ±
42 d. The right panel of Figure 6 shows theDCF calculated between the Mg II line luminosity and theoptical λ src = 3000˚ A continuum. This DCF shows broadnegative and positive peaks of DCF ≃ − .
76 and 0 .
68 at∆ t ≃ t ≃
600 d, respectively. A Gaussian fit to thenegative peak yields the centroid of ∆ t = − ±
44 d anddispersion of σ (∆ t ) = 222 ±
48 d; a fit to the positive peakyields the centroid of ∆ t = 621 ±
45 d and dispersion of σ (∆ t ) = 221 ±
50 d. This allows for two alternative inter-pretations: (1) that the Mg II luminosity is anti-correlatedwith the optical continuum (as can be seen in the bottomleft panel of Figure 2), or (2) that the Mg II luminosity lagsbehind the optical continuum by ∆ t ≃
600 days. The twointerpretations are discussed in Section 4.Looking at the Mg II light curve more subjectively, wecan distinguish two periods of interest: a period of reducedMg II luminosity at MJD ≃ − ≃ − p ≃ × − (confi-dence level of 3 . σ ) that these spectra are consistent witheach other. In Figure 3, we show an example of an accurate modelfor the decomposition of the stacked continuum-subtractedspectrum into an Mg II line represented by three Gaussiancomponents (e.g. Adhikari et al. 2018) with the FWHMof 1607 , − , and a theoretical Fe IIpseudo-continuum with independent multiplets of 60-63,78, and I Zw 1 (Kovaˇcevi´c-Dojˇcinovi´c & Popovi´c 2015;Popovi´c et al. 2019) with kinematic broadenings in therange 2400 − − . Article number, page 6 of 8alewajko et al.: Optical spectroscopic variations in blazar 3C 454.3
4. Discussion
In many non-blazar AGNs, variations in the BEL flux areobserved to be correlated with the optical continuum (typ-ically interpreted as accretion disc emission) with time de-lays of ∆ t ∼ (10 − R BLR of the BLR; this technique is known as reverberationmapping (Blandford, & McKee 1982).In the case of 3C 454.3, which is both a blazar anda quasar, we have a more complex situation (see Figure7). The optical continuum is a combination of highly vari-able synchrotron emission produced in a blazar zone lo-cated in a relativistic jet up to a parsec in front of the nu-cleus (SMBH), and of thermal emission produced in theaccretion disc. It is not possible to confidently separatethese two components. In the traditional picture of BLRbeing concentrated and slightly flattened around the equa-torial plane of the AGN (e.g. Tavecchio & Ghisellini 2012;Nalewajko et al. 2014; Janiak et al. 2015; Sturm et al.2018), it is this unknown accretion disc emission (ratherthan the jet emission that is relativistically beamed awayfrom the equatorial plane) that is reprocessed into theBEL. Observation of a correlation between synchrotronlight curve produced in a foreground jet and BEL lightcurve produced in the background BLR requires that bothare correlated with the underlying poorly constrained accre-tion disc emission. While the reverberation of accretion discemission in the BLR can be explained purely by light-traveleffects, correlation of the synchrotron emission requires con-sideration of perturbations originating in the inner accre-tion disc and propagating along the jet to the blazar zone.We should note that the DCF presented in the rightpanel of Figure 6, between the Mg II line luminosity andthe optical continuum, shows a broad negative peak aroundzero time lag, suggesting that these two light curves couldbe anti-correlated. This anti-correlation would be difficultto explain in the traditional models of BLR geometry, suchas the one depicted in Figure 7. It could be possibly accom-modated in a variant of the extended BLR scenario, withthe BLR clouds concentrated along the relativistic jets andionised by their relativistically boosted non-thermal radia-tion, as advocated by Le´on-Tavares et al. (2013). Instead ofhaving short-scale flares of line luminosity, positively cor-related with the optical continuum flares, increasing lumi-nosity of jet emission could have a destructive effect onsuch extended BLRs by inducing too strong ionisation ofthe surrounding gas. The main difficulty of such a scenariowould be to explain a much longer variability timescale ( ∼ t obs ≃ (621 ±
45) days between the optical continuum light curveand the Mg II light curve is due to reverberation of thethermal accretion disc emission, the corresponding addi-tional path distance that should be travelled by reverber-ated light in the source frame is ∆ r = c ∆ t obs / (1 + z ) ≃ (0 . ± .
02) pc. For a blazar observer located at small view-ing angle θ obs ∼ Γ − measured from the jet axis, the relativedistance measured along the line of sight between the blazarzone and the BLR zone is r blazar cos θ obs , where r blazar ∼ . − t = 0, a perturbation in the inner accretion disc pro- SYN BLRDISKR
BLR r b l a z a r BLAZAR MgIIUASAR t o o b s e r v e r Fig. 7.
Schematic illustration of an AGN being both a quasar(luminous thermal emission from the accretion disc) and a blazar(even more luminous non-thermal emission from the relativisticjet), i.e. an FSRQ like 3C 454.3. It is assumed here that the BLRis concentrated along the accretion disc plane at characteristicdistance R BLR from the SMBH. The optical continuum is dom-inated by the synchrotron radiation produced in a blazar zonelocated within the jet at characteristic distance r blazar , and theMg II BEL is produced in the BLR. duces a flare in the quasar continuum emission, and triggersa perturbation (e.g. internal shock) propagating along thejet with velocity h β j i = h v j i /c . The continuum flare illu-minates the BLR zone at ct BLR ≃ R BLR , while the jet per-turbation reaches the blazar zone at ct blazar ≃ r blazar / h β j i .We can thus express the additional reverberation path dis-tance as ∆ r = r blazar cos θ obs + ct BLR − ct blazar = R BLR − r blazar (cid:16) h β j i − − cos θ obs (cid:17) . With this, we can attempt toconstrain the BLR radius: R BLR ≃ ∆ r + (cid:16) h β j i − − cos θ obs (cid:17) r blazar & ∆ r . (1)Since h β j i − & & cos θ obs and r blazar ∼ ∆ r , the secondterm in the above equation is not important. This corre-sponds to an independent estimate of the black hole mass, M SMBH M ⊙ = f (cid:18) R BLR R g , ⊙ (cid:19) (cid:16) v rms c (cid:17) & (8 . ± . f . × , (2)where v rms ≃ (3170 ± − is measured from themean FWHM of the Mg II line and f ∼ . − × M ⊙ , based mostly on theempirical relation between line widths and continuum lu-minosities (Gu et al. 2001; Woo & Urry 2002; Liu et al.2006; Bonnoli et al. 2011; Sbarrato et al. 2012; Gupta et al.2017). In the case of a blazar with the jet oriented close tothe line of sight, a BLR that is concentrated along the AGNequatorial plane should be oriented close to the plane of the Article number, page 7 of 8 &A proofs: manuscript no. 35904corr˙arxiv sky, largely eliminating the uncertainty on the BLR incli-nation angle.Our claim for delayed correlation between optical con-tinuum and broad line luminosity is based in the first placeon the modulation of the line luminosity as presented in Fig-ure 4. While we identified two periods of interest in the lineluminosity light curve, their connection to the structure ofthe continuum light curve is not obvious. The period of en-hanced line luminosity at MJD ≃ − ≃ ∼
690 days) or to the followingcontinuum flare peaking at MJD ≃ ∼
320 days). On the other hand, the period of reduced lineluminosity at MJD ≃ − ≃ ∼
290 days),while a longer delay of 600 −
700 days would point to theperiod of MJD ≃ − ∼ ∼ − Acknowledgements.
We thank the referee for helpful suggestions,and Marek Sikora and Bo˙zena Czerny for discussions. We useddata from the Steward Observatory spectropolarimetric monitoringproject that has been supported by Fermi Guest Investigator grantsNNX08AW56G, NNX09AU10G, NNX12AO93G, and NNX15AU81G.We also used public data acquired by the Fermi Large Area Tele-scope created by NASA and DoE (USA) in collaboration with insti-tutions from France, Italy, Japan, and Sweden. KN was supportedby the Polish National Science Centre grant 2015/18/E/ST9/00580.The work of ACG is partially supported by Indo-Poland project no.DST/INT/POL/P19/2016 funded by the Department of Science andTechnology (DST) of the Government of India, and by the ChineseAcademy of Sciences (CAS) President’s International Fellowship Ini-tiative (PIFI) grant no. 2016VMB073. KH acknowledges the supportof the Polish National Science Centre grant 2015/17/B/ST9/03422.MFG was supported by the National Science Foundation of China(grants 11873073 and U1531245). ML was supported by the NationalScience Foundation of China (grants 11773056 and U1831138).
References