A TIME-RESOLVED STUDY OF THE BROAD-LINE REGION IN BLAZAR 3C 454.3

Active galactic nuclei (AGNs) are actively accreting supermassive black holes at the centers of massive galaxies. They have been well studied since their discovery 50 years ago (e.g., Antonucci 1984; Peterson & Ferland 1986; Tadhunter et al. 1992; Maraschi et al. 1992; Ghisellini et al. 1993; Urry & Padovani 1995; Korista et al. 1997; Fossati et al. 1998; Kaspi et al. 2005; Marscher et al. 2008, etc.) and, in broad terms, their basic characteristics are reasonably well known. An ultraviolet (UV)/X-ray continuum is produced near the black hole, probably in an accretion disk and energetic corona. Optical and near-infrared spectra of AGNs reveal emission lines produced in gas clouds photoionized by this continuum. In approximately 10% of the total AGN population, a jet of radio-emitting plasma moves outward relativistically (Kellermann et al. 1989; Urry & Padovani 1995). We still do not know how jets form and are collimated, what distribution of jet energies nature produces, or how (if) the accretion flow is linked to the jet.

Blazars are AGNs oriented with the jet axis at very small angles to the line of sight. This orientation causes Doppler beaming of the observed jet luminosity, making the jet appear brighter and the variability timescale appear shorter. Beaming makes blazars an ideal laboratory for studying the relativistic jets, because the observed jet emission dominates the thermal emission. The observed jet luminosity, L, is related to the intrinsic jet luminosity, $\mathcal {L}$, by the Doppler beaming factor, δ: $L=\delta ^{4}\mathcal {L}$ (with luminosity integrated over energy). For blazars with orientation angles that are small with respect to the line of sight, δ ∼ Γ at θ = Γ⁻¹ and β_app ∼ Γβ, where Γ is the bulk Lorentz factor, β_app is the observed velocity of the emitter, and β is the true bulk velocity. Typically, Γ ∼ δ ∼ 10 in blazars.

Outflow energy is thought to be extracted from the black hole through the accretion disk or black hole spin (Blandford & Payne 1982; Blandford & Znajek 1977). Thus, the relationship between the accretion disk and relativistic jet is of primary importance to understanding the energy budget of the system. The accretion disk light is often swamped by the non-thermal jet emission, however, even in the optical regime. Still, in some blazars we can see (unbeamed) emission lines and even a hint of a thermal accretion disk spectrum (Pian et al. 1998; Raiteri et al. 2007a; Bonning et al. 2009; D'Ammando et al. 2009; Bonnoli et al. 2011; Raiteri et al. 2011); polarization data provide an independent detection of the Big Blue Bump in blazars (Smith et al. 1986, 1988).

Perhaps the most powerful tool to address the connection between the disk and jet is time variability. By monitoring the different components and emission regimes and tracking the leads and lags between them, one can in principle build up a full geometric picture of the flow of matter and resulting radiation.

The short timescale γ-ray variability of blazars can now be studied extremely well thanks to the launch of the Fermi γ-ray satellite in 2008 June. Since then, several groups have been monitoring γ-ray bright blazars in other wavebands. In particular, the Small and Medium Aperture Research Telescope System (SMARTS) daily optical-infrared photometry and bi-monthly optical spectroscopy was obtained for several southern hemisphere sources (Bonning et al. 2012), including 3C 454.3 (Bonning et al. 2009). This blazar was the brightest γ-ray source in the Fermi sky during and after launch. During the EGRET era, it was much fainter, with a maximum jet flaring flux (E > 100 MeV) of F₁₀₀ ∼ 0.5 × 10⁻⁶ photons cm⁻² s⁻¹, more than a factor of 120 times fainter than the brightest state during the Fermi era.

Optical-infrared photometry of 3C 454.3 during the first year of Fermi operations showed correlated variability with γ-ray and UV light curves (Bonning et al. 2009, Figure 1), a result that was confirmed for subsequent flares in the same source (Pacciani et al. 2010; Raiteri et al. 2011; Vercellone et al. 2011; Jorstad et al. 2010). Five major γ-ray flares have been recorded for 3C 454.3 over the 3 yr period discussed here. High-resolution observations of the radio core of 3C 454.3 have also been undertaken (Jorstad et al. 2005, 2010, 2012). Recent observations show that the 1 mm and 7 mm radio core emission is well correlated with the γ-ray and optical flaring events, with the 7 mm core ejections detected in the 2009 December and 2010 November flaring periods (Jorstad et al. 2012). This correlation suggests that in addition to the synchrotron-emitting electrons and upscattered γ-ray radiation being cospatial (Bonning et al. 2009), the high-frequency radio core emission must also be in close proximity given their simultaneous flaring episodes.

Zoom In Zoom Out Reset image size

Emission-line variability information in blazars comparable to the reverberation mapping studies of moderate-luminosity AGNs is scarce (Peterson & Ferland 1986; Clavel et al. 1991; Netzer & Peterson 1997), although some line variability has been observed on timescales of months (Zheng & Burbidge 1986; Bregman et al. 1986; Ulrich et al. 1997; Perez et al. 1989; Corbett et al. 2000). For two prominent blazars, the continuum variations are a factor of 2 (3C273) to a factor of 50 (3C279) times larger in the optical/UV than the emission line variations, which are less than 25% in amplitude (Ulrich et al. 1993; Falomo et al. 1994; Koratkar et al. 1998). This lack of large-amplitude line variability in blazars implies that the underlying photoionizing flux comes from a relatively constant source, presumably the accretion disk. Historically, no evidence for an increase in line flux related to the jet flares was detected, even when the beamed contribution to the UV continuum swamped the unbeamed thermal disk contribution at periods of high jet activity (e.g., Smith et al. 2011).

However, the lack of long-term simultaneous high-energy data and multi-epoch emission-line studies made it difficult to detect any disk-jet connection. Multi-epoch optical spectroscopy of mostly northern hemisphere sources was carried out in conjunction with Fermi observations, including 3C 454.3 (Smith et al. 2009; Benítez et al. 2010). Recently, a 40% variation in the Mg ii line was reported (León-Tavares et al. 2013). Other contemporaneous, spectroscopic studies have compared single-epoch, Fermi-detected blazar emission lines and the total radiative γ-ray luminosity to derive correlations between the broad-line region (BLR) and γ-ray luminosity (Chen et al. 2009; Ghisellini et al. 2011; Sbarrato et al. 2012; Shaw et al. 2012).

In this paper, we present time-resolved spectroscopy of the bright, superluminal blazar 3C 454.3 obtained with SMARTS, which we supplement with publicly available data. The variability of the strongest features, the Mg ii, Hβ, and Hγ emission lines, is discussed. We then compare with the continuum variability, in particular the optical and γ-ray flux, which comes from the aligned jet. In Section 2, we describe the SMARTS observational program and the Fermi data. In Section 3, we discuss the line variations. We discuss the implications of our findings in Section 4 and our conclusions in Section 5. We use a concordance cosmology to describe our results: H₀ = 70 km s⁻¹ Mpc⁻¹, Ω_m = 0.27, and Ω_Λ = 0.73.

2.1. Photometry and Polarimetry

2.2. SMARTS Spectroscopy

Low-resolution optical spectra (R ∼ 500) of 3C 454.3 (z = 0.859) were obtained from 2008 August 22 through 2011 December 1, approximately twice monthly, as detailed in Table 1, using the SMARTS consortium 1.5 m + RCSpectrograph in queue-scheduled mode. The 1.5 m Cassegrain spectrograph is at an f/7.5 focus with a plate scale of 181 mm⁻¹ and a LORAL 1K (1200 × 800) CCD. The primary grating for this study has a first-order resolution of 17.2 Å, spectral coverage of 3600–9000 Å, and a slit width of 2''. In 2008–2010, the integration time was set to 900 s, with three sequential exposures per night. Beginning in 2011, exposure times were reduced to 600 s to allow observation of more sources per observing night. Two illustrative examples of SMARTS optical spectra, during the two highest γ-ray flux states, can be seen in Figure 2.

Zoom In Zoom Out Reset image size

The spectra were bias- and overscan-subtracted, then flat-fielded using standard IRAF procedures. Wavelength calibration was completed using reference HeAr spectra taken immediately prior to the target integration. The spectra were then interpolated to a linear wavelength scale. Bad pixels were rejected and then spectra were interpolated across them. The three nightly exposures were then averaged together to produce the highest possible signal-to-noise spectrum per night.

The spectra were calibrated using near-simultaneous photometry from SMARTS (see Table 1). The continuum flux near the Mg ii line (λ_obs = 5200 Å) was calibrated using the same SMARTS V-band photometry (effective wavelength 5500 Å) since the observed wavelength is redshifted into this band. Similarly, the Balmer lines were calibrated with the R-band photometry (Hγ, λ_obs = 8071 Å; Hβ, λ_obs = 9040 Å; see Figure 2). The conversion from each waveband magnitude to flux density, in units of erg s⁻¹ cm⁻² Å⁻¹, utilized Bessell et al. (1998) zero points. The equivalent width can be converted to an emission line flux by multiplying by the flux density. This conversion to line flux is the largest error introduced in our calculation. We note that on 2010 May 20, although Steward Observatory had a more simultaneous photometric observation, we used the SMARTS photometry due to calibration mismatches between the two data sets, which caused the R-band magnitudes to differ by more than the stated uncertainties.

No order-blocking filters were used to obtain the spectra presented here. Given the width of the spectral range covered, second-order contamination may contribute to the continuum flux at wavelengths redder than ≈6400 Å, which could affect line flux measurements of the Balmer lines. An estimation of the second-order contamination for this optical setup was obtained using historical observations of the standard star LTT 4364 with and without the order blocking filter GG495 on 2004 April 20. We find the second-order contamination to be ≈5% at 8070 Å and ≈8% at 9040 Å, the observed locations of the Balmer lines in this study. These estimates are in agreement with previous characterizations of second-order contamination found in the literature (Szokoly et al. 2004; Stanishev 2007). The second-order contamination measured in the standard star is an upper limit on the contamination expected in 3C 454.3 given that it is often redder than the standard star. The measured equivalent width would change by, at most, a few percent, which contributes less than other uncertainties in the line flux measurement. Thus, second-order contamination is not considered in the following analysis.

The line equivalent widths were measured by fitting a Gaussian to the emission lines above the continuum, minimizing the χ² statistic, using the MPFIT package (Markwardt 2009). The continuum range used on each side of the line was 50 Å wide. The uncertainty in the equivalent width was determined by running 500 Monte Carlo simulations of the fitted line, including the measured noise in the count rate of each pixel. For each Monte Carlo simulation, the emission lines were fit and the equivalent widths were calculated. The reported error of the equivalent width of each line is the standard error on the 500 measured equivalent widths from the simulation. The low-resolution spectroscopy presented here does not allow the measurement of accurate line shape characteristics.

2.3. Fermi Observations

3.1. Emission Line Variability

Figure 3 shows the equivalent width versus optical magnitude for the Mg ii (purple circles), Hβ (cyan squares), and Hγ (orange stars) emission lines over the total observation period. The data were compared with models of constant line flux in each emission line. The best-fit constant line flux is found by minimizing the χ² statistic for a family of constant emission line fluxes ranging from 7 × 10⁻¹⁵ erg s⁻¹ cm⁻² to 20 × 10⁻¹⁵ erg s⁻¹ cm⁻² in 0.1 × 10⁻¹⁵ increment steps. Best-fit constant lines of 9.3 × 10⁻¹⁵ erg s⁻¹ cm⁻², 1.4 × 10⁻¹⁴ erg s⁻¹ cm⁻², and 9.9 × 10⁻¹⁵ erg s⁻¹ cm⁻² were found for Mg ii, Hβ, and Hγ, respectively. While the continuum flux increases, the emission line fluxes show little significant deviation from the constant line flux approximation.

Zoom In Zoom Out Reset image size

Emission line light curves for Mg ii, Hβ, and Hγ are shown in Figure 4, compared with the Fermi γ-ray light curve simultaneous to within one day. Dashed lines are mean flux values and dot-dashed lines are the ±2σ uncertainties. From 2008 August through 2011 December, the emission line fluxes are largely consistent with constant line flux, except during two epochs of observation, shown in Figure 5.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Figure 5 (left) shows the 2009 December γ-ray flaring period (MJD 55150–55200). During the 2009 December flare (peak γ-ray flare at MJD 55167), we see positive spectral deviations of 1.2σ in Hβ, 1.8σ in the Mg ii line, and 2.8σ in the Hγ line (at MJD 55165), corresponding to a factor of 1.7 from the mean flux for both Mg ii and Hγ. Gray points in the line flux plots mark data obtained from Steward Observatory, as reported by León-Tavares et al. (2013), which extend from 4000 to 7550 Å rest-frame waveband coverage, so that the Mg ii line is the only emission line that can be compared with our findings. The León-Tavares et al. (2013) data appear to have significantly smaller error bars than we report here, which may be due to the fact that the flux calibration errors are not included in their published data tables. We estimate the additional error due to flux calibration in their data, in addition to the reported error, and plot these uncertainties with the light gray dotted error bars. Hγ and Hβ are reported only in the current paper. While no significant deviation in the Mg ii line is reported in the León-Tavares et al. (2013) data for this flaring period, we note that those observations occurred immediately pre-flare (MJD 55155) and so do not characterize the response of the spectrum during the peak flaring period itself.

Figure 5 (right) expands the 2010 November flaring period (MJD 55500–55540) when 3C 454.3 was the brightest source in the γ-ray sky. León-Tavares et al. (2013) reported a 40% increase in the Mg ii line flux, which we see as a 1.5σ deviation from the mean. In addition, we see a factor of 2.5 increase in the line flux of Hγ (corresponding to a 3.7σ deviation) during the same epoch (MJD 55518). We note that while the increase in Mg ii line flux is similar to that in the 2009 December flare, Hγ showed a more significant increase in the line flux during the more powerful γ-ray flaring period in 2010 November.

During both flaring periods, the measured continuum flux was measured to be significantly higher than would be predicted if the line fluxes were assumed to be constant, indicating that the measured line fluxes were larger than the mean line flux in each case. For the 2009 December flaring period, the predicted continuum flux differed from the empirical line flux at the peak line flux by 69% (Mg ii), 77% (Hβ), and 56% (Hγ), corresponding to an average 4σ difference; similarly, for the 2010 November flaring period, the deviations were 32% (Mg ii), 65% (Hβ), and 15% (Hγ), an average 2σ difference between the measured and predicted continuum flux. Thus, the line flux variations are distinguishable from increases in continuum flux alone, as can be seen in Figure 3. We determine the errors on the predicted line flux by summing the uncertainties in the equivalent width and photometry in quadrature (see Table 1).

Figure 6 shows emission line flux versus the γ-ray flux; there is clear evidence that high jet flux states contribute to the photoionizing flux, although the jet cannot contribute the majority of the photoionizing flux. Otherwise, the quiescent jet states would correspond to decreased line flux, which is not seen. Instead, the data remain near the mean emission line flux in all three emission lines during quiescent jet states.

Zoom In Zoom Out Reset image size

3.2. Broad-Line Region Variability

The three emission lines discussed in this work can be combined to estimate the total broad-line and accretion disk luminosities. The total BLR luminosity, L_BLR, can be estimated empirically by extrapolation from a few individual lines. The six strongest broad emission lines in quasars—Lyα (100), C iv (63), Mg ii (34), Hγ(+O iii) (13), Hβ (22), and Hα (77)—represent ∼60% of the total luminosity emitted in broad lines (Celotti et al. 1997), where the numbers in parentheses indicate the observed line strengths relative to Lyα (Francis et al. 1991; Gaskell et al. 1981). Using these line ratios and summing the three lines that are detected in virtually all spectra obtained for this study—Mg ii, Hγ, and Hβ—we estimate the total broad-line luminosity:

$$\begin{equation} L_{{\rm BLR}} \simeq \eta (L_{{\rm Mg\,\scriptsize{II}}} + L_{{\rm H}\beta } + L_{{\rm H}\gamma }), \end{equation} \tag{ 1 }$$

where η = 7.5 when all three lines are used and η = 14.7 when only the Balmer lines are detected. The line luminosities obtained for each spectrum are listed in Table 3 and the mean broad-line luminosity is L_BLR ≃ 10⁴⁵ erg s⁻¹ (see Table 4). Using three lines makes the broad-line luminosity estimate more robust than the usual correction from one line (Celotti et al. 1997; Sbarrato et al. 2012). We note that the line ratios suggested by Celotti et al. (1997) are consistent for the three lines we observed in the SMARTS spectra.

Table 3. Emission Line Luminosities

UTC	MJD	log LMg ii	log LHβ	log LHγ
20080823	54701.2	43.5	43.6	43.4
20081014	54753.1	43.6	43.8	43.5
20081104	54774.0	43.6	43.8	43.7
20090710	55022.4	43.5	43.7	43.7
20090825	55068.2	43.6	43.6	43.8
20090902	55076.2	43.7	43.6	43.8
20090912	55086.1	43.6	43.4	43.6
20090923	55097.1	43.3	43.5	43.6
20091002	55106.2	43.7	43.4	43.8
20091012	55116.1	43.6	43.6	43.7
20091023	55127.1	43.6	43.5	43.7
20091117	55152.1	43.6	43.5	43.6
20091130	55165.0	43.8	43.7	43.9
20100520	55336.4	43.6	43.7	43.7
20100702	55379.4	43.6	43.7	43.6
20100714	55391.4	43.6	43.8	43.5
20100802	55410.4	...	43.7	43.8
20100920	55459.1	43.5	43.3	43.7
20101112	55512.0	43.7	43.6	43.9
20101118	55518.1	43.8	43.7	44.1
20101203	55533.0	43.6	43.5	43.8
20110611	55723.4	43.7	43.9	43.6
20110626	55738.4	43.4	43.6	43.6
20111002	55836.2	...	43.7	43.6
20111017	55851.1	43.6	43.7	43.5
20111102	55867.1	43.6	43.8	43.6
20111202	55897.0	43.7	43.6	43.7

Note. The UTC is in YYYYMMDD format and all luminosities in units of erg s⁻¹.

Download table as:  ASCIITypeset image

Table 4. Derived Physical Quantities

Parameter		Units
〈FMg ii〉	−13.96 (0.10)	erg s−1 cm−2
〈FHβ〉	−13.93 (0.14)	erg s−1 cm−2
〈FHγ〉	−13.87 (0.14)	erg s−1 cm−2
〈LBLR〉	45	erg s−1

Notes. The mean line flux and broad-line luminosity derived for 3C 454.3. All quantities are in rest-frame units. Average fluxes and luminosities are given in logarithmic units.

Download table as:  ASCIITypeset image

Our results show that the BLR in 3C 454.3, as traced by the Mg ii, Hβ, and Hγ emission lines, is not undergoing large variations over 90% of the period of observation. This is consistent with a slowly varying accretion disk providing a large fraction of the photoionizing flux in the BLR. If the accretion disk was undergoing significant changes in accretion rate, the (disk) photoionizing flux would change and in turn the emission line fluxes would also vary, likely with lags on the order of several months. Given the lack of large-amplitude variability in the broad lines and thus in the photoionizing disk emission, the jet does not contribute much photoionizing flux to the BLR in general. This is consistent with the 3C279 result (Koratkar et al. 1998), showing only small line flux variations (≈25%) with factors of 50 in optical continuum variability.

Historically, the UV thermal bump from disk emission has been observed to be ∼7 × 10⁴⁶ erg s⁻¹ in the faint state of 3C 454.3 following the 2005–2006 outburst (Villata et al. 2007; Raiteri et al. 2007b). Comparing our estimate of the total broad-line luminosity, ≈1 × 10⁴⁵ erg s⁻¹, we infer that either the covering factor is ≈1/70 or, in order for the covering factor to be closer to the value of 0.1 observed in other quasars (Baldwin & Netzer 1978; Smith et al. 1988), the disk luminosity must have declined by a factor of seven. The lack of variability seen in the UV Big Blue Bump during both active and quiescent jet states (Bonnoli et al. 2011; Bonning et al. 2012), however, favors the former explanation.

However, the very highest γ-ray jet flaring periods are associated with significant increases in the BLR emission. The 2009 December flare phase (MJD 55166–55173; F₁₀₀ = 22 × 10⁻⁶ photons cm⁻² s⁻¹; Ackermann et al. 2010), corresponded to significant increases in the Mg ii and Hγ line fluxes during the plateau phase of that flare period. The 40% increase in Mg ii line flux seen by León-Tavares et al. (2013) in 2010 November occurs during the plateau phase (MJD 55503–55515), whereas we saw a larger increase. The 2σ deviation in Mg ii and the 4σ increase in Hγ during the flaring phase (MJD 55516–55522; F₁₀₀ = 66 × 10⁻⁶ ph cm⁻² s⁻¹) are much larger than the increase seen by León-Tavares et al. (2013); these results are consistent because they do not sample the same time periods and, in particular, they do not sample the γ-ray flaring period (as defined by the Ackermann–Abdo convention¹¹). In our case, because we normalize using highly accurate photometry, we avoid systematic errors that come from the difficulty of absolutely calibrating the spectra and maintaining calibration accuracy over long time intervals.

The 2010 April flare (MJD 55280–55300; F₁₀₀ = 16 × 10⁻⁶ photons cm⁻² s⁻¹) showed no significant line variability in any of the lines analyzed in this work. While the Fermi γ-ray fluxes for the 2009 December and 2010 April flaring phases are similar to the 2010 November plateau phase intensity, the 2010 April flare has slightly lower peak intensity and is the only flare that did not show a detectable 7 mm core ejection event (Jorstad et al. 2012). Furthermore, in contrast with the significant increases in polarization during the 2009 and 2010 flares, the 2010 April flare also lacks a significant increase in the polarized (non-thermal) flux. This flare may therefore be produced in a different manner than the other two flares discussed here. As such, the difference in line flux response may be due to a lack of strong γ-emitting emission originating within the BLR. The low polarization seen near MJD 55700, which also corresponds to low Fermi γ-ray flux, also suggests that no γ-emitting plasma is present. This is in contrast with the high polarization that is seen during high γ-ray states and when emission line variability is observed.

The Hγ line shows stronger variations than Mg ii and Hβ in both the 2009 December and 2010 November flaring periods. This may be because the Mg ii and Hβ lines are more challenging to measure due to systematic effects. Hβ suffers from a falloff in detector sensitivity at the red end of the spectrograph; Mg ii is diluted by the underlying Fe ii emission known to be present at the same wavelength interval (this systematic effect has not been accounted for in the quoted uncertainties of the Mg ii line strength). The Fe ii contribution is expected to be strongest in the wings of the Mg ii line and at a local minimum around the center of the Mg ii line (Wills et al. 1985; Goad et al. 1999) and thus is not likely to contribute significantly to our line flux measurements. In addition, León-Tavares et al. (2013) decomposed the Fe ii emission but it showed the same variation as the Mg ii line (see their Figure 2), suggesting that its contribution does not obscure the emission line variation or else that it is not possible to disentangle it from Mg ii. The smaller variability of Mg ii compared with Hγ is likely due to the stratification of the BLR with respect to ionization parameter and gas density (Clavel et al. 1991; Goad et al. 1999). It should also be noted that Hγ falls within the highly variable atmospheric water vapor absorption feature from ∼8000–8400 Å, which could in principle affect the line flux we measure. However, any atmospheric absorption would attenuate the measured emission line flux, such that the true line flux is greater than that reported.

As the broad emission lines vary by less than a factor of two over most of the observation and the sampling is not dense enough, it is not possible to do reverberation mapping. However, constraints can be placed on the variability timescale during the 2010 November flare period given the coincidence of the γ-ray flare period with the deviation from the mean line fluxes. The peak γ-ray flux occurs within 2 days of the peak Hγ line flux (MJD 55520 and 55518, respectively), so the light travel time from the broad-line clouds to the jet acceleration region is ⩽2δ days. This work shows that the Mg ii emission coincides with the strongest flaring period in 2010 November, as does the Hγ emission and both lines increase near the peak of the 2009 December γ-ray flaring period. This suggests that during the strongest flaring periods the jet is bright enough to significantly increase the photoionizing flux in the BLR, as also suggested by León-Tavares et al. (2013). If the disk was responsible for the changes we see, we would expect each flare to be associated with changes in the BLR luminosity with a marked delay, which is not observed.

Previous studies have shown correlations among millimeter, infrared, optical, UV, and γ-ray emission (Bonning et al. 2009; Jorstad et al. 2012; Wehrle et al. 2012; Abdo et al. 2011; Vercellone et al. 2011) during the 2010 November flaring period. These correlations argue for the cospatiality of the synchrotron radiation with the lower energy photons on which inverse Compton scattering takes place. That the variability of the BLR luminosity coincides with the strongest γ-ray flaring periods suggests that the BLR must lie further from the central source than the γ-emitting region, at least during the 2009 December and 2010 November flaring events. The fact that we see an increase in broad-line strength at the same time as the Doppler-boosted jet emission implies that broad-line clouds along the line of sight are affected by the jet flare.

The location of the γ-emitting region is actively debated. Many studies suggest that it lies within the BLR (Böttcher 2007; Kataoka et al. 2008; Tavecchio et al. 2010; Ghisellini et al. 2010; Poutanen & Stern 2010; Stern & Poutanen 2011; Abdo et al. 2011), which is at sub-pc scales (0.03–1 pc) (e.g., Peterson 1993, 2006), while others conclude it is well outside the BLR, tens of pc from the black hole (Marscher et al. 2008; Jorstad et al. 2010, 2012; Agudo et al. 2012). We know from very long baseline interferometry imaging that at least some jet flares are located at large distances from the central source (∼10 pc) (e.g., TEMZ model; Marscher et al. 2011; Agudo et al. 2011). 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 (r_em ∼ 0.14 pc), using γγ-opacity arguments.

An independent argument can be used to determine whether the γ-emitting region could exist within the BLR. The correlation of the millimeter and γ-ray light curves imposes an upper limit on the synchrotron self-absorption frequency to be ∼7 mm. The spectral energy distribution for 3C 454.3 presented in Bonning et al. (2012) was fit with a one-zone synchrotron and inverse Compton model (Coppi 1992; Ghisellini et al. 2007) with the following parameters: size: R_γ ∼ 10¹⁷ cm, Doppler factor: δ ∼ 20, and magnetic field strength: B ∼ 1 G, similar to the fits of Bonnoli et al. (2011). For this model, the self-absorption frequency, ν_ssa ⩽ 3 mm, is consistent with a γ-emitting region located within the canonical BLR.

The beaming of the jet emission adds photoionizing flux to a small fraction of the gas in the BLR, similar to that described in the mirror model (Ghisellini & Madau 1996). A systematic study of other blazars would confirm whether increased broad-line fluxes are generally associated with the strongest jet flares. Then, by comparing the line variability characteristics of other (misaligned) quasars and radio galaxies, one would have an independent probe of the parent population of blazars.

The optical–infrared continuum and optical emission line flux variability of the blazar 3C 454.3 was measured throughout 3.3 yr of continuous Fermi observations. Five dramatic γ-ray flares coincided with large increases in the millimeter, optical–infrared, and UV bands but the broad emission lines varied much less, being consistent for the most part with a constant line luminosity. Two notable exceptions were seen, however, during the two strongest γ-ray flares. The coincidence of the increases in the broad emission lines at the peak of the two highest γ-ray flaring states (and core ejections) suggests that the γ-emitting region, at least in those flares, must be interior to the BLR gas.

To strengthen these results, much finer cadence optical spectroscopy is needed during the highest γ-ray flaring states of 3C 454.3, meaning at least daily monitoring during and in the weeks immediately following a large γ-ray flare. Other bright blazars should be monitored spectroscopically to see whether they show similar evidence of photoionization by a compact γ-ray source. We have undertaken such a study, with results to be reported in a future paper.

We thank the anonymous referee for valuable improvements to this manuscript. We also thank Frederick Walter, queue scheduler for the SMARTS 1.5 m + RCSpectrograph, for providing the order blocking data and conversations regarding the performance of the spectrograph. SMARTS observations of LAT-monitored blazars are supported by Yale University and Fermi GI grant NNX 12AP15G. J.C.I. has received support from NASA-Harriet Jenkins Pre-doctoral Fellowship Program, NSF Graduate Research Fellowship Program (DGE-0644492), and the National Research Council's Ford Foundation Dissertation Fellowship. C.D.B., M.M.B, and the SMARTS 1.3 m and 1.5 m observing queue also receive support from NSF grant AST-0707627. We are grateful for photometry and polarimetry from Paul Smith's monitoring program at the Steward Observatory, which is supported by Fermi Guest Investigator grants NNX08AW56G, NNX09AU10G, and NNX12AO93G.