THE SMARTS MULTI-EPOCH OPTICAL SPECTROSCOPY ATLAS (SaMOSA): AN ANALYSIS OF EMISSION LINE VARIABILITY IN SOUTHERN HEMISPHERE FERMI BLAZARS

The SMARTS Multi-epoch Optical Spectroscopy Atlas (SaMOSA) was based on the original Fermi-LAT "bright source list" released just before launch in 2008, including those blazars with declination < 20°, given the location of the SMARTS telescopes. The original list included flat-spectrum Radio Quasars (FSRQs) and BL Lac objects (BLLs), which are distinguished by whether broad emission lines are present at levels greater or less than 5 Å, respectively (Angel & Stockman 1980). In subsequent years, the SaMOSA list was expanded to include newly flaring Fermi blazars, defined as having F $_{\gamma }$ (E > 100 MeV) $\geqslant \;1\times {{10}^{-6}}$ photons s⁻¹ cm⁻², and flaring FSRQs publicized on the Astronomers Telegram.⁶ We did not include BLLs in the final analysis because no emission lines were detected in their optical spectra. Since the purpose of the current study is to understand broad-line variability, we only include objects for which at least five epochs of spectroscopy are available, for a total of seven FSRQs. Table 1 lists the SaMOSA sample, Fermi identifier, redshift, number of observations, and emission lines included in the analysis.

The SMARTS OIR photometry is obtained nightly with the 1.3 m + ANDICAM, a dual-channel imager with a dichroic that simultaneously feeds an optical CCD and infrared IR imager, with spectral coverage from 0.4 to 2.2 μm. Data analysis for SMARTS OIR photometry is described in Bonning et al. (2012), so we briefly summarize here. SMARTS differential photometry is obtained by using optical comparison stars calibrated to Landolt standards on photometric nights. The reported SMARTS infrared magnitudes are calibrated using Two Micron All Sky Survey (2MASS) magnitudes (Skrutskie et al. 2006) of at least one secondary star in the same field as the blazar. In both optical and infrared bands, the photometric uncertainties are dominated by the random errors in the comparison star magnitude, and the 1σ uncertainty is reported. In the infrared, the dominant uncertainty is the calibration to the 2MASS magnitude. OIR finding charts and comparison star magnitudes for PKS 0402-362, PKS 0454-234, PKS 2142-75, and PKS 2052-474 are provided in the Appendix; SMARTS finding charts for the remaining sources can be found in Bonning et al. (2012). OIR photometry and finding charts for all SMARTS-monitored sources are publicly available via our website.⁷

SMARTS optical spectroscopy was obtained with the 1.5 m + Cassegrain spectrograph (RCSpec) at an f/7.5 focus with plate scale 181 mm⁻¹ and a LORAL 1 K (1200 × 800) CCD. The primary grating for this study has first-order resolution of 17.2 Å, with spectral coverage of 6600 Å and 2$^{\prime\prime }$ slit width. In the optimal case, spectroscopic data were obtained approximately biweekly, depending on weather conditions and source visibility in Cerro Tololo, although in many cases significantly less data were acquired. No second-order corrections were applied to the spectra, as previous analysis yielded a contamination rate of 8% or less (Isler et al. 2013), which is insignificant compared to the errors introduced by the flux calibration. Nevertheless, the systematic error introduced by the contaminating continuum light is difficult to model, especially in the case when the variability amplitude is larger in the blue than the red part of the continuum. In any case, this contamination applies to emission lines with observed wavelength $\gtrsim$7000 Å, so that Hα in PKS 1510-089 is the only emission line affected by this contamination.

The optical spectroscopic data reduction here is similar to that described by Isler et al. (2013). We use the MPFIT package (Markwardt 2009) to measure the emission line equivalent width by fitting a Gaussian to the emission line above the continuum and minimizing the ${{\chi }^{2}}$ statistic. A linear fit to the continuum was applied on each side of the line over 100 Å. We calculated the noise per pixel by combining the uncertainties from bias subtraction, sky correction, and aperture extraction.

The uncertainty in the equivalent width was estimated by running 500 Monte Carlo simulations of each fitted line, including the measured noise in the count rate of each pixel (as described above). For each Monte Carlo simulation, the emission line was fitted and the values of the equivalent width were calculated. The reported error of the equivalent width of each line is the standard error on the 500 equivalent width simulations.

Emission line flux was derived from the equivalent width and SMARTS B-, V-, or R-band photometry, depending on the observed location of the line in the spectrum. Emission lines with rest-frame line center ${{\lambda }_{{\rm obs}}}$ < 5000 Å were calibrated with the B band (${{\lambda }_{{\rm eff}}}$ = 4400 Å), line centers 4999 < ${{\lambda }_{{\rm obs}}}$ < 5999 Å were calibrated with the V band (${{\lambda }_{{\rm eff}}}$ = 5500 Å), and ${{\lambda }_{{\rm obs}}}$ > 6000 Å were calibrated with the R band (${{\lambda }_{{\rm eff}}}$ = 6600 Å). No observed line was more than 500 Å from the relevant effective wavelength in any case. Table 2 lists the sources, observations, emission line equivalent widths with their associated uncertainties, and B-, V-, R-, and J-band magnitude with associated 1σ uncertainty for the SaMOSA sample.

Table 2.  SaMOSA Optical Photometry and Spectroscopy Log

UTCa	MJD	B	${{\sigma }_{B}}$	V	${{\sigma }_{V}}$	R	${{\sigma }_{R}}$	J	${{\sigma }_{J}}$	W(Mg ii)	${{\sigma }_{W({\rm Mg}\,{\rm II})}}$	W(C iii])	${{\sigma }_{W({\rm C}\,{\rm III}])}}$
PKS 0208-512
20080623	54,640.379	16.472	0.004	⋯	⋯	15.49	0.004	⋯	⋯	4.953	0.467	9.859	1.672
20080805	54,683.331	18.016	0.015	17.497	0.018	17.137	0.018	15.77	0.023	⋯	⋯	13.321	4.131
20080823	54,701.3	18.155	0.024	17.654	0.025	17.282	0.022	⋯	⋯	18.807	2.828	16.451	3.933
20080908	54,717.236	18.372	0.045	17.654	0.058	17.433	0.048	⋯	⋯	23.452	3.344	27.091	6.037
20080926	54,735.151	16.738	0.007	16.283	0.008	15.882	0.007	14.494	0.011	6.5	0.927	8.273	1.567

^aUTC is in YYYYMMDD format. The equivalent widths of the emission lines are reported as W(species), e.g., W(Mg ii), in units of angstroms. SMARTS photometry are given in magnitudes. Table 2 is published in its entirety in the Astrophysical Journal; a portion is shown here for guidance regarding its form and content.

Only a portion of this table is shown here to demonstrate its form and content. A machine-readable version of the full table is available.

Download table as:  DataTypeset image

Since the launch of Fermi, the Steward Observatory of the University of Arizona has carried out regular, publicly available optical spectrophotometry and linear spectropolarimetry of a large sample of γ-ray-bright blazars using the Bok 2.3 m and Kuiper 1.54 m telescopes (Smith et al. 2009). The SPOL spectropolarimeter (Schmidt et al. 1992) is used for this monitoring program. Observations are made in first order using a 600 line mm⁻¹ grating, yielding a spectral range of 4000–7550 Å and resolution of 15–20 Å.

Table 3.  Mean Emission Line and γ-ray Fluxes and Luminosities

Source	Line	$\langle {{{\rm F}}_{{\rm line}}}\rangle$	$\langle {{{\rm L}}_{{\rm line}}}\rangle$	$\langle {{{\rm L}}_{\gamma }}\rangle$
PKS 0208-512	Mg ii	−14.17 (0.06)	43.94	51.29 (0.20)
	C iii]	−14.18 (0.18)	43.95	⋯
PKS 0402-362	C iv	−13.90 (0.06)	44.20	51.74 (0.28)
	S iii]	−14.10 (0.12)	44.01	⋯
	Mg ii a	−14.34 (0.08)	43.38	⋯
PKS 0454-234	Mg ii a	−14.64 (0.07)	43.08	51.35 (0.24)
PKS 1510-089	Mg ii	−13.51 (0.10)	43.14	50.62 (0.31)
	H$\gamma$ a	−13.92 (0.19)	42.72	⋯
	H$\beta$ a	−13.65 (0.20)	42.97	⋯
	Hα	−13.20 (0.08)	43.44	⋯
PKS 2052-474	Mg ii	−14.09 (0.09)	43.48	51.75 (0.19)
	C iv	−14.55 (0.20)	44.07	⋯
	C iii]	−14.82 (0.44)	43.67	⋯
PKS 2142-75	Mg ii	−14.18 (0.10)	43.69	51.58 (0.20)
3C 454.3	Mg ii	−13.96 (0.10)	43.60	52.17 (0.39)
	Hβ	−13.93 (0.14)	43.61	⋯
	Hγ	−13.87 (0.14)	43.67	⋯

Note. Mean line flux and line luminosity are given in units of log erg s⁻¹ cm⁻² and log erg s⁻¹, respectively, and include all available data.

^aThe systematic offset between the Steward and SMARTS data has been applied before calculating the mean line fluxes and luminosities. Fermi γ-ray mean luminosity includes all Fermi data with TS $\geqslant$ 16 in the entire observing window for each source and is in units of log erg s⁻¹. The standard deviation is the reported uncertainty. Data for 3C 454.3 were previously reported in Isler et al. (2013) but are reproduced here for comparison.

Download table as:  ASCIITypeset image

Sky-subtracted spectra and broadband (5000–7000 Å) polarization measurements derived from the spectropolarimetry from 2008 to 2013 were obtained from the publicly accessible Steward Observatory website.⁸ The optical spectroscopy was then reduced in a similar fashion to the SMARTS spectroscopy. The noise per pixel array for the Steward spectroscopy was estimated a posteriori using the gain and read noise. This array was then scaled by a factor of 10 to approximate the rms in the observed spectrum. The scaling of the uncertainties provides a very conservative estimate of the error in the lines and likely overestimates the noise in the spectrum. The broadband polarization measurements were used to derive V-band polarized flux densities by multiplying the fractional linear polarization (P) by the V-band flux density (F_V).

There is a small systematic offset, of order 0.1 dex, in the line fluxes reported by Steward and SMARTS, likely due to differences in the comparison stars used for the analysis. The logarithmic offsets from the SMARTS data are applied to the light curves of PKS 0402-362 (Mg ii: −0.1203), PKS 0454-234 (Mg ii: −0.2493), and PKS 1510-089 (Hγ: +0.0481; Hβ: −0.0501); the other four sources included in the SaMOSA sample contain only SMARTS data, so no cross-calibration is necessary. The mean emission line fluxes and luminosities for the sources in our sample are listed in Table 3.

Fermi/LAT data were obtained from the first SMARTS photometric observation for each source through 2013 July 01 (MJD 56,474), via the Fermi Science Support Center website.⁹ Pass 7 reprocessed data (event class 3) were analyzed using Fermi Science Tools (v9r33p0) with scripts that automate the likelihood analysis. Galactic diffuse models (gll_iem_v05_rev1), isotropic diffuse background (iso_p7v6source), and instrument response functions (P7REP_CLEAN_V15) were utilized in the analysis. Data were constrained to time periods where the zenith angle was less than 100° to avoid Earth limb contamination, and photons to within a 10° region centered on the source of interest.

The Fermi γ-ray spectra of each object were modeled as a power law or log-parabola, according to spectral type listed in the 2FGL catalog, with the photon flux and spectral index as free parameters. Fermi light curves were integrated in 1 day bins to match the average SMARTS photometric cadence, and an integral Fermi γ-ray flux above 100 MeV, F $_{\gamma }$, is reported. Fermi fluxes for which TS $\geqslant$ 16 are plotted in subsequent figures, where TS is the Fermi test statistic and $\sqrt{{\rm TS}}$ is roughly equivalent to the detection significance per integrated bin (Mattox et al. 1996; Abdo et al. 2009). When daily binned fluxes fell below the significance threshold, we plot weekly binned flux, also at the TS $\geqslant$ 16 level. In the case of PKS 0454-234, adaptive binning techniques (Lott et al. 2012) were used to determine the Fermi γ-ray flux in the 10 day span around MJD 56,280. The same TS threshold was used to plot significant Fermi γ-ray fluxes during the daily and weekly analysis.

We note that simultaneous measurements of Fermi and Steward data mean data obtained within the same day. The SMARTS optical spectroscopy was flux-calibrated to OIR photometry to within the hour. In no case are any two data sets matched with temporal separation of more than 4 days.

PKS 0208-512 was observed with SMARTS from 2008 August 6 to 2013 February 12 (MJD 54,640–56,339), and the emission line behavior is shown in Figure 1. Although an increase in line flux was detected in C iii] between MJD 54,701 and 54,892, peaking at MJD 54,735, with log F_{C iii]} = −13.97 erg s⁻¹ cm⁻² (2.5σ), it does not meet the empirical emission line flare significance criteria for an empirical emission line flare and is not included in subsequent analysis. Thus, no significant emission line variability was seen in this blazar.

PKS 0402-362 was observed from 2011 October 2 to 2012 December 12 (MJD 55,836–56,285). The source underwent a large, short-duration γ-ray flare from MJD 55,821 to 55,838 that is not well sampled in emission line flux (see Figure 2). No significant emission line variability was detected in this blazar.

Zoom In Zoom Out Reset image size

PKS 0454-234 was observed between 2011 October 16 and 2013 July 27 (MJD 55,850–56,500); the light curve can be seen in Figure 3. We also plot the optical linear polarization. For this object, calibrated V-band photometry from Steward Observatory was not available, so we used differential photometry (ΔF_V) to determine the relative polarized flux. Mg ii underwent a 3.0σ line flare with peak line flux log F_{Mg ii} = −14.23 erg s⁻¹ cm⁻² on MJD 56,285. No accompanying Fermi γ-ray flare was detected in the daily binned fluxes, so the adaptive binning technique (Lott et al. 2012) was utilized during the 10 day period around MJD 56,280 to determine if sub-day variability was present. We plot the results of the adaptively binned Fermi γ-ray fluxes along with the daily binned γ-ray fluxes and the Mg ii emission line flare for comparison in Figure 4. The apparent deviation of γ-ray flux derived from adaptive binning is of the same order as the rms in this region and is thus not significant.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

We first presented the emission line light curve for 3C 454.3 in Isler et al. (2013); one can also be seen in León-Tavares et al. (2013) for this epoch. We extend the previous analyses by applying the criterion for emission line flares used here; both previously reported emission line flares in Hγ and Mg ii meet both variability criterion and are thus statistically significant.

Optical spectra were obtained for PKS 1510-089 between 2008 May 17 and 2012 May 3 (MJD 54,603–56,050); the light curve can be seen in Figure 5. We detected line flares in Hα peaking on MJD 54,934 at log F $_{{\rm H}\alpha }$ = −13.09 erg s⁻¹ cm⁻² with 8.5σ significance. The leading line flux for this Hα flare was also above the mean at MJD 54,913 with 3.3σ significance (log F $_{{\rm H}\alpha }$ = −13.14 erg s⁻¹ cm⁻²), suggesting that the emission line flare may extend past the range observed on the increasing side of the γ-ray flare. During the Fermi γ-ray flare (MJD 54,850–55,000) with peak flux F$_{\gamma }$ = −5.08 photons s⁻¹ cm⁻² on MJD 54,917, H.E.S.S. also detected very high energy photons from MJD 55,910 to 55,952, with highest emission, log F $_{\gt 0.15\,{\rm TeV}}$ $\approx$ −10.4 photons s⁻¹ cm⁻², on MJD 54915 (H.E.S.S. Collaboration et al. 2013).

Zoom In Zoom Out Reset image size

Hγ and Hβ show significant line flares on MJD 55,292 at the 3.3σ (log F $_{{\rm H}\gamma }$ = −13.79 erg s⁻¹ cm⁻²) and 3.7σ (log F $_{{\rm H}\beta }$ = −13.58 erg s⁻¹ cm⁻²) level, respectively. A significant emission line flare was observed in Hβ on MJD 55,622 with log F $_{{\rm H}\beta }$ = −13.57 erg s⁻¹ cm⁻² (4.0σ). Hβ has an emission line flare during the same γ-ray flare, peaking on MJD 56050 at log F $_{{\rm H}\alpha }$ = −13.54 erg s⁻¹ cm⁻² (5.4σ). In Figure 6 we show the regions where emission line flares were detected in at least one emission line, as described above.

Zoom In Zoom Out Reset image size

Statistically significant emission line variability in the Hα, Hγ, and Hβ emission lines is measured, and the ${{\chi }^{2}}$ variability test also indicates variability: $\chi _{{\rm H}\alpha }^{2}$ = 199.4 (14 dof, p $\ll$ 10⁻³), $\chi _{{\rm H}\gamma }^{2}$ = 32.3 (15 dof, p = 0.01), and $\chi _{{\rm H}\beta }^{2}$ = 87.9 (17 dof, p $\ll$ 10⁻³).

PKS 2052-474 was observed from 2011 May 2 to 2012 July 11 (MJD 55,683–56,119) and is the only source in the sample that did not show any significant Fermi γ-ray flux above log F_γ = −6 photons s⁻¹ cm⁻² over the epoch of observation presented here (see Figure 7). No significant emission line variability was observed in this blazar.

Zoom In Zoom Out Reset image size

PKS 2142-75 was observed from 2010 May 5 to 2012 September 15 (MJD 55,321–56,185), seen in Figure 8. No significant emission line flares were observed during the epoch of observation.

Zoom In Zoom Out Reset image size

We find that four of the seven blazars observed in this sample show no evidence of statistically significant emission line variability. The lack of emission line variability in these blazars is consistent with the standard model that the accretion disk is the predominant source of photoionizing flux.

However, PKS 0454-234, 3C 454.3, and PKS 1510-089 all show statistically significant emission line variability, with a mean peak emission line flare significance of 4σ. We are only aware of a few blazars with published simultaneous Fermi γ-ray and optical emission line data. Simultaneous emission line variability studies in Fermi-monitored blazars have generated mixed results. Among a set of similar γ-ray and optically bright, variable quasars, little emission line variability was detected in PKS 1222+216 or 4C 38.51 (Smith et al. 2011; Farina et al. 2012; Raiteri et al. 2012). It was reported that FSRQ PKS 1222+216 did not have significant emission line variability during recent Fermi γ-ray flaring events (Smith et al. 2011; Farina et al. 2012); however, line variability at the 2.6σ level is evident in both data sets. While emission line variability at this level would not have met the criteria set forth in this work, the lines do show some variation based solely on continuum variations. Without reanalyzing the data, we note that when the Hβ line luminosity from Steward Observatory reported in Farina et al. (2012) is compared to the line luminosity obtained from TNG in the same work, there is a 17σ difference in line luminosity. The latter was obtained following a MAGIC triggered observation.

Zoom In Zoom Out Reset image size

Emission line variability has been detected in PKS 0454-234, 3C 454.3, and PKS 1510-089, which are also the three sources with the highest Fermi variability index. This is suggested by the empirical emission line flares and ${{\chi }^{2}}$ variability test; correlation information including possible lags is harder to assess given the limits of the data. Thus, we attempt to relate the degree of non-thermal jet contribution to the photoionizing flux in the BLR to optical–polarization. It has long been known that during periods of increased γ-ray activity, the thermal contribution to the optical–ultraviolet continuum is swamped by non-thermal emission (e.g., Smith et al. 1994). We confirm this result with the optical polarization for 3C 454.3 and PKS 1510-089, where brighter γ-ray (and emission line) fluxes correspond to highly polarized states, presumably due to jet emission. However, we do not find evidence of this behavior in PKS 0454-234. The synchrotron peak has been shown to be well correlated from infrared to ultraviolet for FSRQs (e.g., Bonning et al. 2009), such that optical variations indicate similar variability patterns in the ultraviolet, where the photoionizing flux peaks.

Zoom In Zoom Out Reset image size

The tentative picture that emerges from this work is that the most active γ-ray flaring sources have evidence of emission line variability that is not seen in sources with less active jets. The correlations between γ-ray flux and emission line flux are likely diluted by the presence of lags and/or leads in the peak of either curve. Thus, we expect that correlated line variability could be detected with higher cadence, multi-epoch optical emission line studies of γ-ray active blazars.

Zoom In Zoom Out Reset image size

Next, we turn our attention to the location of the γ-emitting region, which has been the subject of much research. The models fall into two major categories: (1) near-field γ-emitting models suggest that the bulk of the jet dissipation occurs on sub-pc scales, either very deep in the BLR (Poutanen & Stern 2010) or near the edge of the BLR (Böttcher 2007; Kataoka et al. 2008; Ghisellini et al. 2010; Poutanen & Stern 2010; Tavecchio et al. 2010; Abdo et al. 2011; Stern & Poutanen 2011), but in any case at or below canonical distances of 0.1 pc (e.g., Peterson 1993, 2006), and (2) far-field γ-emitting scenarios, which suggest jet dissipation on much larger spacial scales (tens of pc) from the central source (Agudo et al. 2011; Marscher et al. 2011; Jorstad et al. 2012). These studies attempt to constrain the location of the γ-emitting region based on SED modeling, correlation studies, and/or ultra-short γ-ray variability.

More recently, a few coordinated optical spectroscopic variability studies, like the one presented here, have been undertaken in an effort to identify emission line variability of γ-ray bright blazars (Smith et al. 2009, 2011; Benítez et al. 2010; Raiteri et al. 2012; Isler et al. 2013; León-Tavares et al. 2013). In the cases where line variability is found, attempts are sought to directly (and simultaneously) relate the broad-line variability to jet variability via γ-ray, mm, or other non-thermal emission.

The results of both the indirect and direct studies have been mixed, even for the same source and same flare. For example, in 3C 454.3, following the 2009 December and 2010 November flaring periods, Tavecchio et al. (2010) and Abdo et al. (2011) suggest that the γ-emitting region of 3C 454.3 could be located at the outer edges of the BLR $({{{\rm r}}_{{\rm em}}}\sim 0.14\;{\rm pc})$, using γγ-opacity arguments. Isler et al. (2013) also suggested a near-field dissipation mechanism after observing statistically significant emission line variability in both the 2009 and 2010 flares. By contrast, León-Tavares et al. (2013) argue in favor of a potentially far-field dissipation mechanism in the 2010 flare, given statistically significant emission line variability in close temporal proximity to a mm-core ejection. They argue that their lack of detectable emission line variability during the 2009 flare, in combination with the absence of an additional mm-core ejection, suggests that emission line variability may be caused by the radio core ejections. The two spectroscopic studies come to different conclusions likely due to different observation windows around the 2009 flare. The observations presented by León-Tavares et al. (2013) did not extend across the entire γ-ray flare; therefore, a simultaneous comparison of the peak γ-ray to emission line fluxes was not possible. The data collected by Isler et al. (2013) extended across the entire γ-ray flare, albeit with fewer total observations. Thus, the lack of detected emission line variability in the 2009 flare by León-Tavares et al. (2013) is likely due to lack of temporal coverage and not to the absence of line variability itself.

We also consider the mm-core ejections seen in PKS 1510-089 (Marscher et al. 2010) with respect to the emission line variability reported here. We report emission line variability in Hα that peaks on MJD 54,934. A mm-core ejection is not seen in this flaring period until MJD 54,959, suggesting that it is likely not the cause of the emission line variability (on the pc scales on which the core ejection is observed). To our knowledge, no subsequent studies on core ejections in PKS 1510-089 have been published. Therefore, we are unable to compare our reports of emission line variability. While these two examples of non-coincident core ejections at the time of emission line variability do not preclude such occurrences, we suggest that a core ejection is not a necessary condition for emission line variability. However, the near temporal proximity of γ-ray flares to nearly every instance of emission line variability does suggest that jet flares (and the attendant increase in photoionizing flux) are likely required to produce emission line variability on timescales of a few weeks to months. Thus, we consider our results to favor a near-field jet dissipation region.

While the emission line flux variability is evidence of a BLR response, it does not provide a conclusive test of its dynamical structure. Whether the jet emitting region is within the canonical BLR or part of an outflowing wind (either from the disk or entrained in the jet) on larger scales is still unknown. The broad emission line profiles alone are not sufficient to distinguish between BLR dynamical models (Capriotti et al. 1980), although higher line profile moments like asymmetry can distinguish steady-state versus dynamical theories (Capriotti et al. 1981). Thus, whether the BLR is a distribution of gas clouds (or filaments) driven primarily by Keplerian velocities or a disk (or jet) wind cannot be constrained by the current line variability studies; higher resolution (and cadence) coordinated spectroscopic observations are necessary to truly constrain the dynamical nature of the BLR in blazars.

However, if some part of the BLR were located very far from the central source as a result of entrainment in the jet, we would expect to see a non-variable line core with highly variable blueward line wings due to the high velocities of these outflowing and entrained clouds. Conversely, we do not expect the jet emitting region to be located deep within the BLR, as it would require a more isotropic (likely unbeamed) jet contribution that would not be significantly brighter than the accretion disk in the source frame and hence not produce detectable emission line variability. In addition, observations of very high energy TeV emission coincident with emission line variability, like that seen in PKS 1510-089, are hard to reconcile with the increased probability for γγ-absorption deep in the rich BLR photon field.

If, instead, the jet emission were interacting with the BLR, as suggested in the "mirror model" (Ghisellini & Madau 1996), one could still observe jet-augmented photoionization in the broad line emission. According to this model, which estimates the BLR as a thin shell, the energy density increases dramatically at the location of the BLR, where the gas sees strongly beamed flux from the jet (see their Figure 2). Two factors are at play here: (1) the jet emission is beamed and illuminates only a small fraction of the assumed spherical BLR, within a solid angle π/${{{\Gamma }}^{2}}$, though with a strongly enhanced flux, and (2) the disk radiation is assumed to be roughly isotropic. Together with Equation (26) of Ghisellini & Madau (1996), this implies the measured BLR luminosity, L_BLR ∼ L_disk + $\frac{{{L}_{{\rm jet}}}}{{{{\Gamma }}^{2}}}$, where both L_disk and L_jet are observed luminosities. For an ${\Gamma }\sim 10$, the jet contribution to the total (optical–UV) photoionizing luminosity is a factor of ∼100 less than that of the disk. Thus, for the observed variation of a factor of ∼2 in BLR flux, the jet photoionizing flux should increase by a factor of >100.

For 3C 454.3 and PKS 1510-089, this constraint is easily accommodated, as both have shown significant increases (ΔB $\gtrsim$ 3 mag) in optical–UV flux during flaring events and L_jet > L_disk, especially in high flaring states. However, as was shown in Marscher et al. (2010), the emitting region in a given blazar is quite complicated and has been observed to shift from sub-pc to several pc scales in a matter of months (Finke & Dermer 2010; Ghisellini et al. 2013), and thus we do not expect every flare to take place within the BLR.

In the case where the γ-emitting region is located outside the BLR, we would not expect flares in the jet to cause line variability. However, lines can vary because of variable disk emission; this should happen on longer timescales, since the disk varies slowly compared to the jet, and need not in general be associated with an increase in γ-ray emission. The latter kind of event may have taken place in PKS 0454-234, where we see insufficient optical–UV jet photoionizing flux combined with the lack of a strong γ-ray flare near the occurrence of the emission line flare, suggesting that this instance of line variability may be caused by a different physical mechanism. In this case, the accretion disk variability could be caused by hotspots, which can change the shape and behavior of the optical spectrum as described by Ruan et al. (2014). Furthermore, Ghisellini et al. (2010) have already shown that L_disk $\approx$ L_jet in PKS 0454-234, such that the jet does not produce enough photoionizing radiation in the optical–UV regime to produce the necessary photoionization of the BLR. We also note that the optical polarization is not well correlated with γ-ray flux in PKS 0454-234 as in 3C 454.3 and PKS 1510-089, also suggesting that a different physical mechanism causes the line variability in PKS 0454-234.