THE STRUCTURE AND EMISSION MODEL OF THE RELATIVISTIC JET IN THE QUASAR 3C 279 INFERRED FROM RADIO TO HIGH-ENERGY γ-RAY OBSERVATIONS IN 2008–2010

The data used here comprise two-year observations obtained between 2008 August 4 and 2010 August 6 (MJD 54682–55414). We used the standard LAT analysis software, ScienceTools v9r21. The events were selected using so-called diffuse class events. In addition, we excluded the events with zenith angles greater than 100° to avoid the contamination of the Earth-limb secondary γ radiation. The events were extracted in the range between 200 MeV and 300 GeV within a 15° acceptance cone of the region of interest (ROI) centered on the location of 3C 279 (R.A. = 195047, decl. = −5789, J2000). Below 200 MeV, the effective collection area of LAT for the diffuse class events drops very quickly and thus larger systematic errors are expected. The γ-ray flux and spectrum were calculated using the instrument response function (IRF) of "P6_V11_DIFFUSE" by an unbinned maximum likelihood fit of model parameters. We examined the significance of the γ-ray signal from the sources by means of the test statistic (TS) based on the likelihood ratio test.⁵⁴ The background models included a component for the Galactic diffuse emission along the plane of the Milky Way, which was modeled by the map cube file "gll_iem_v02_P6_V11_DIFFUSE.fits." An isotropic component (isotropic_iem_v02_P6_V11_DIFFUSE.txt) was also included to represent the extragalactic diffuse emission and residual instrumental background. Besides those components, the model in our analysis also included the emission from all nearby point sources inside the ROI from the first Fermi-LAT catalog (1FGL: Abdo et al. 2010c). The spectra of those sources were modeled by power-law functions except for a pulsar 1FGL J1231.1–1410 (=PSR J1231–1411), for which we included an additional exponential cutoff in its spectral modeling. During the spectral fitting, the normalization factors of the Galactic diffuse and isotropic components and the nearby sources were left as free parameters, and the photon indices of the nearby sources were fixed to the values from the 1FGL catalog except for 3C 273, whose photon index was allowed to vary freely. In the light curve analysis, we considered only two bright sources in the background model as nearby point sources, namely, 3C 273 and 1FGL J1231.1–1410, because other nearby sources had a negligible contribution to γ-ray signal, especially in such relatively short timescales (shorter than a week) for the light curves considered here. The fluxes used for the light curve were calculated by a simple power-law model fit using data in the given energy ranges.

The γ-ray light curve measured by Fermi-LAT can be seen in Figure 1. The figure shows the flux history above 200 MeV averaged over (a) one day intervals, (b) three day intervals, and (c) one week intervals. It also includes one week light curves of (d) the flux between 200 MeV and 1 GeV, (e) the flux above 1 GeV, and (f) the photon index in the range above 200 MeV.

Zoom In Zoom Out Reset image size

The γ-ray flux clearly shows variability. The source showed high-flux states between MJD 54700 and 54900, in which two prominent flares can be seen: one of the flares at ∼MJD 54800 and the other at ∼MJD 54880. During the second flare, a change in the optical polarization associated with a γ-ray flare was discovered (Paper I). We detected some flux variability between MJD 55000 and 55120, but after that, the source remained in a relatively low activity state until the end of the period considered in this paper. During this two-year period, the highest integral flux above 200 MeV occurred on MJD 54880 in the one day interval light curve with flux of F_{E > 200 MeV}  =  (11.8 ± 1.5) × 10⁻⁷ photons cm⁻² s⁻¹ and TS  =  306. By extrapolating the spectrum down to 100 MeV, an integral flux above 100 MeV on that day yields F_{E > 100 MeV}  =  (31.0 ± 6.0) × 10⁻⁷ photons cm⁻² s⁻¹, which is still a factor of 3–4 times lower than the flux of the brightest flare (∼1 × 10⁻⁵ photons cm⁻² s⁻¹) detected during the EGRET observations of the source (Wehrle et al. 1998; Hartman et al. 2001b).

We quantified the flux variability using one week interval data for energies above 200 MeV (full band), between 200 MeV and 1 GeV (soft band), and above 1 GeV (hard band). This is based on the "excess variance" method (Nandra et al. 1997; Edelson et al. 2002) after subtracting the contribution expected from measurement errors (σ_{err, i}). Using the mean square error 〈σ_{err, i}〉, the excess variance F_var can be described as (Vaughan et al. 2003)

$$\begin{equation} F_{\rm var} =\sqrt{\frac{S^2 - \langle \sigma _{{\rm err},i}\rangle ^2}{\langle F\rangle ^2}}, \end{equation} \tag{ 1 }$$

where S is the variance of the flux, and 〈F〉 is the mean value of the flux. The definition of associated error can be found in Vaughan et al. (2003). In the calculation, we excluded bins of 8, 82, 87, and 90 because the fit in the hard band failed due to poor statistics of the data samples. Resulting F_var values are 0.695 ± 0.015, 0.648 ± 0.017, and 0.839 ± 0.030 for the full, soft, and hard bands, respectively. The resulting values indicate that the flux of the hard band showed significantly stronger variability than that of the soft band. For comparison, F_var = 0.79 ± 0.02 for E > 300 MeV has been reported during the first 11 months of the Fermi scientific mission (Abdo et al. 2010f), when the source has clearly been more active.

A power density spectrum (PDS) for the three day binned light curve was calculated using a Fourier transform and is shown in Figure 2. The power density was normalized to fractional variance per frequency unit (rms² I⁻² day⁻¹) and the PDS points were averaged in logarithmic frequency bins. The white noise level was estimated from the rms of the flux errors and was subtracted from the PDS. A slope of 1.6 ± 0.2 was obtained from a linear fit to the binned PDS for frequencies up to 0.1 day⁻¹. The main uncertainty in the estimated PDS slope is due to the stochastic nature of the variability which leads to variations in the determined slope between different time-limited observations. An additional effect which can cause a systematic bias in the observed PDS slope is the red noise leakage (e.g., Chatterjee et al. 2008). In the present analysis this effect is not taken into account.

Zoom In Zoom Out Reset image size

Figure 3 shows plots of flux versus photon index (Γ) based on the weekly light curve results above 200 MeV (full band), between 200 MeV and 1 GeV (soft band), and above 1 GeV (hard band). The data that have TS > 10 were selected for the plots and are shown in gray points. An average photon index was calculated by fitting a constant value in each plot, corresponding to Γ_>200 MeV = 2.334 ± 0.015, Γ_{200 MeV − 1 GeV} = 2.20 ± 0.03, and Γ_>1 GeV = 2.48 ± 0.04 for the full, soft, and hard bands, respectively. The average photon index in the soft band shows a significantly harder spectrum than that in the hard band.

Zoom In Zoom Out Reset image size

We also derived photon indices resulting from an analysis where the data were sorted in five bins using week-long fluxes for each energy band, and plotted the results as red points. Those photon indices of each flux bin are also shown in the insets in Figure 3. For the full band, although the change of the photon index is rather small (ΔΓ ∼ 0.2) compared to the flux variation (spanning about an order of magnitude), a weak "harder when brighter" effect can be seen. Such an effect was also measured in other LAT blazars (Abdo et al. 2010e). The soft band also shows the weak "harder when brighter" effect with a slightly larger change of the photon index (ΔΓ ∼ 0.4). On the other hand, the photon index of the hard band changes only sightly (ΔΓ ∼ 0.1) and is statistically consistent with a constant value.

During the two year observations, the highest energy photon associated with 3C 279 was detected at MJD 54891.60745 with an estimated energy of 30.8 GeV. The event was converted in the front-thin layers (so-called front event) of the LAT detector and still remains even when we apply the cleanest event selection (so-called data clean event), which was developed for studying extragalactic γ-ray background (Abdo et al. 2010g). The reconstructed arrival direction of the event is 57 (= 0095) away from 3C 279, and is within the 68% containment radius of the LAT point-spread function (PSF; 0114 in the IRF of "P6_V11_DIFFUSE") for front events at 30.8 GeV. Based on our model fit of the epoch which contains that highest energy photon, we find the probability that the photon was associated with 3C 279 (as opposed to all other sources in the model including the diffuse emission and nearby point sources) is 88.6%.

In total, we found 10 events with estimated energies higher than 20 GeV within an 025 radius centered at 3C 279. All events lie within a 95% containment radius of the LAT PSF from 3C 279 and remain even after the "data clean selection" applied. The number of expected background events above 20 GeV within the 025 radius at the location of 3C 279 for the two year observations is only 0.16 events. The bottom panel of Figure 1 plots the arrival time distribution of those 10 events. All events except for two were detected between MJD 54780 and 54900 during the high-activity states. No photon above 20 GeV associated with 3C 279 has been detected after MJD 54914 during the two year observations.

We extracted the γ-ray spectra using data for the entire two-year period and following eight sub-periods (see also Table 1): (A) the initial quiescent state in the γ-ray band (MJD 54682–54728), (B) the first γ-ray flaring state (MJD 54789–54809), (C) an intermediate state (MJD 54827–54877), (D) the first five days of the second γ-ray flaring event (MJD 54880–54885), (E) the last three days of the second γ-ray flaring event (MJD 54897–54900), (F) during the isolated (first) X-ray flaring event (MJD 54950–54960; see Section 4), (G) during the second X-ray flaring event (MJD 55042–55045; see Section 4), and (H) a quiescent state (MJD 55240–55319). Those sub-periods were also selected taking into account observations in other energy bands. SEDs in the γ-ray band for each sub-period are presented in Figure 4. Each γ-ray spectrum was modeled using a simple power-law (PL; dN/dE∝E^−Γ), a broken power-law (BPL; $dN/dE \propto E^{-\Gamma _1}$ for E < E_brk and $dN/dE \propto E^{-\Gamma _2}$ otherwise), and a log parabola (LogP; $dN/dE \propto (E/E_0)^{-\alpha - \beta \log (E/E_0)}$) model. In the case of LogP model, the parameter β represents the curvature around the peak. We note that the choice of the reference energy E₀ in the LogP model does not affect the determination of the other two model parameters, and hence we fixed it at 300 MeV.

Zoom In Zoom Out Reset image size

The best-fit parameters calculated by the fitting procedure are summarized in Table 1. The integral fluxes above 100 MeV⁵⁵ derived using each spectral model are also included. The averaged γ-ray spectral shape for the two year observation significantly deviates from a single power law. A LogP model is favored to describe the γ-ray spectral shape over the simple PL model with the difference of the logarithm of the likelihood fits⁵⁶−2ΔL = 46.5 (corresponding to a significance level of ∼7σ),⁵⁷ and a BPL fit yields −2ΔL = 43.0. Even in some individual periods as defined above, the spectra deviate from a single power law: for example, the spectrum in the Period B yields −2ΔL = 13.6. This is consistent with our finding in Section 2.2 that the spectrum above 1 GeV is significantly softer than the spectrum below 1 GeV. We thus conclude that the γ-ray spectrum significantly deviates from a simple power law. The spectral break in 3C 279 is not as pronounced as that seen in the spectra, e.g., of 3C 454.3 (Ackermann et al. 2010). One the other hand, the BPL model returns break energies within a few GeV range regardless of the flux levels as observed in other bright FSRQs, such as 3C 454.3 and 4C+21.35 (Tanaka et al. 2011). Such a spectral feature could be due to γ–γ absorption to pair production by He ii Lyman recombination continuum UV photons from the emission line region (see, e.g., Poutanen & Stern 2010), or a break in the electron distribution (Abdo et al. 2009b). We consider the γ-ray emission region to be located significantly beyond the broad emission region (see the Discussion in Section 5), and this implies that the break in the electron energy distribution is the more likely explanation.

The Suzaku X-ray satellite (Mitsuda et al. 2007) observed 3C 279 as a part of multi-band studies of the object. The observations took place in two segments, with an interruption lasting roughly 1.5 days: (1) between 2009 January 19, 23:19:00 and 2009 January 22, 22:32:00 UTC (sequence number 703049010), and (2) between 2009 January 23, 20:45:00 and 2009 January 25, 03:00:00 UTC (sequence number 703049020). "Period C" (see Table 1) includes both Suzaku observations. The goals of the Suzaku observations were to monitor the soft-medium X-ray flux (0.3–12 keV) of the source with the X-ray Imaging Spectrometer (XIS; Koyama et al. 2007) and to take advantage of the data from the Hard X-ray Detector (HXD; Takahashi et al. 2007). The HXD consists of PIN silicon diodes for the lower energy band (10–70 keV) and GSO scintillators for the higher energy band (40–600 keV), to extend the spectral bandpass beyond the energies accessible with imaging instruments (>10 keV). The HXD nominal position was used for the observations to maximize its effective area. In the following analysis, the HXD/GSO data were not used because there was no significant detection of the source.

Although the observation conditions were nominal, the XIS1 data suffered from somewhat high and variable background, resulting in the total apparent counting rate ranging from one to three counts s⁻¹ in source-free regions for the entire chip. Still, the background-subtracted spectrum determined from the XIS1 data below 8 keV was entirely consistent with that from XIS0 and XIS3 and thus we included the background-subtracted XIS1 data in the spectral fitting. The total duration of good data accumulated by the XIS instruments was 191 ks. We used the standard ftools data reduction package, provided by the Suzaku Science Operations Center, with the calibration files included in CALDB ver. 4.3.1. For the analysis of spectra and light curves, we extracted the counts from a region corresponding to a circle with 260'' radius, centered on the X-ray centroid; we used a region of a comparable size from the same chip to extract the background counts. The net count rates were 0.47, 0.63, and 0.56 count s⁻¹ for XIS0, XIS1, and XIS3, respectively, with the typical count rate uncertainty in the entire observations of ∼3%. The data indicate no significant variability during the Suzaku observations.

The source was also detected in the HXD/PIN data, although the signal was relatively weak. We used the standard cleaned events, processed using the standard criteria applicable to the rev. 2.13 of the Suzaku HXD data processing software. This yielded 95.4 ks of good data, with a total count rate of 0.3 counts s⁻¹. For the background subtraction, we used the standard background files provided by the Suzaku team through HEASARC. We applied the standard tool hxdpinxbpi which accounts for the particle background as well as for the contribution of the cosmic X-ray background as appropriate for the effective area and the solid angle of the HXD. The net counting rate was 0.02 counts s⁻¹, with the formal statistical uncertainty of ∼10%. We note that this formal uncertainty is probably lower than the standard systematic error due to the background subtraction of 3% of the average background (corresponding to 0.01 counts s⁻¹). Nonetheless, even if the additional uncertainty is included, the source was still detected by the HXD/PIN.

For the spectral analysis, we used the XSPEC spectral analysis software. For the spectral fitting of the XIS data, we used the standard redistribution files and mirror effective areas generated with Suzaku-specific tools xisrmfgen and xissimarfgen. In the spectral fits, we used the counts corresponding to the energy range of 0.5–10.0 keV for XIS0 and XIS3, and 0.5–8.0 keV for the XIS1. We used all three XIS detectors simultaneously, but allowed for a small (a few %) variation of normalization. For the HXD/PIN data, we considered the data in the range of 20–50 keV and used the response file ae_hxd_pinhxnome5_20080716.rsp.

The source spectrum was modeled as an absorbed power law, with the cross-sections and elemental abundances as given in Morrison & McCammon (1983); other absorption models give similar results. The best-fit absorbing column was (3.1 ± 0.5) × 10²⁰ cm⁻², and the photon index was 1.76 ± 0.01. Inclusion of the HXD/PIN data in the fit did not change the fit parameters perceptibly. The χ² for the fit including the three XIS detectors and the HXD/PIN was acceptable, with 5061 for 5023 channels. The absorption inferred from the simple absorbed power-law model is marginally greater than the value inferred from the radio measurements of the column density of the material in the Galaxy of 2.0 × 10²⁰ cm⁻² (with an estimated error of ∼10%; Kalberla et al. 2005). We deem the difference not significant, since at such small column densities, it can be accounted for by even small systematic uncertainty in the knowledge of the effective area of the XIS instruments at the lowest end of the XIS bandpass. Furthermore, a modest additional column density is expected in the host galaxy of 3C 279. The observed model 2–10 keV flux is 8.0 × 10⁻¹² erg cm⁻² s⁻¹, with the statistical error of <2%, which is probably smaller than the systematic error resulting form the calibration uncertainty of the Suzaku instruments. We plot the Suzaku 3C 279 spectra in Figure 5.

Zoom In Zoom Out Reset image size

XMM-Newton observed 3C 279 once starting on 2009 January 21, 17:28 UT. The observation was largely devoid of flares (except for the period close to the end of the observation), and the total length of good data accumulated in the pointing was 16.8 ks. We used the standard Scientific Analysis System (SAS) data reduction package, provided by the XMM-Newton Science Operations Center. Since 3C 279 is a relatively bright source, we considered only the pn-CCD data. We note here that the spectra and light curves taken by MOS-CCDs were entirely consistent with the results inferred from the pn-CCD data.

For the analysis of spectra and light curves, we extracted the counts from within 40'' radius of the source; we used a region of the same size, from the same pn-CCD chip, to extract the background counts. The data indicate no significant variability during the XMM-Newton observation. The spectral analysis was performed using the XSPEC v.12 spectral analysis software with the standard redistribution files and mirror effective areas included in the SAS package. We used the counts corresponding to the energy range of 0.5–10.0 keV in our spectral fits.

The source spectrum was first modeled as an absorbed power law; the best-fit absorbing column was (2.2 ± 0.6) × 10²⁰ cm⁻², and the photon index was 1.77 ± 0.03, with χ² of 588 for 517 dof. The result is consistent with the spectral analysis results of the Suzaku observations as described in the previous section, which were performed during the same period as the XMM-Newton observation. We also considered a broken power-law model and found that the overall intrinsic source spectrum hardens with increasing energy. The absorbing column was (3.4 ± 0.7) × 10²⁰ cm⁻², and the low- and high-energy indices were, respectively, 1.83 ± 0.05 and 1.55 ± 0.2 with the break energy of 4.1 ± 0.8 keV. The resulting χ² was 563, for 515 dof. The broken power-law model is statistically only marginally superior to the simple power-law model, especially given that the absorption inferred form the simple power-law model is closer to the value inferred from Kalberla et al. (2005). For either model, the 2–10 keV flux is 7.7 × 10⁻¹¹ erg cm⁻² s⁻¹, with a statistical error of 5%, which is probably smaller than the systematic error resulting form the calibration uncertainty of the XMM-Newton pn-CCD.

RXTE carried out 321 observations between 2008 July 3 (MJD 54650) and 2010 August 12 (MJD 55420). Those include 52 observations based on the Cycle 12 Guest observer (GO) program and 269 observations based on the Core program in Cycles 12–14. The fluxes resulting from the Cycle 12 GO observations have been already reported in Paper I. Most of the individual observations have exposure times in a range from 1.0 to 2.5 ks. We analyzed the data from the Proportional Counter Array (PCA) following standard procedures using the rex script in HEASOFT v.6.9. Only signals from the top layer (X1L and X1R) of PCU2 were extracted for data analysis. The data were screened with the following data selection: source elevation above the horizon >10°, pointing offset smaller than 002, at least 30 minutes away from a South Atlantic Anomaly (SAA) passage and electron contamination smaller than 0.1. The background was estimated with standard procedures, and the detector response matrices were extracted with the RXTE tools (command PCARSO v.11.7.1). For the spectral analysis we re-binned the spectra into 11 channels. The spectra from the channels corresponding to nominal energies of 2.6–10.5 keV are adequately fitted by a single power-law model, absorbed by a fixed Galactic column density of 2.2 × 10²⁰ cm⁻² using the XSPEC v.12 software package. The value of the column density is based on our XMM-Newton results (in Section 3.2) and is also consistent with the value based on Kalberla et al. (2005).

In the HEASARC database,⁵⁸ there are 80 publicly available Swift X-Ray Telescope (XRT) observations between 2008 July 3 (MJD 54650) and 2010 August 12 (MJD 55420), which include 32 pointings based on an approved GI proposal in Cycle 4 (Proposal number: 5080069). The results of the flux history based on the data until 2009 May 31 have already reported in Paper I. Effective exposure times of these observations range between 1 and 3 ks, but some have longer exposure times, for example, 8.9 ks for ID:35019007 (MJD 54795), 22.5 ks for ID:35019009 (MJD 54797), 20.2 ks for ID:35019010 (MJD 54799), and 15.4 ks for ID:35019011 (MJD 54800). The XRT was used in the photon counting mode, and no evidence of pile-up was found. The XRT data were reduced with the standard software xrtpipeline v.0.12.6, applying the default filtering and screening criteria (HEADAS package, v.6.10). The source events were extracted from a circular region, 20 pixels in radius, centered on the source position. Exposure maps were used to account for PSF losses and the presence of dead pixels/columns. The background was determined using data extracted from a circular region, 40 pixels in radius, centered on (R.A., decl.: J2000) = (12^h56^m26^s, −05°49'30''), where no X-ray sources are found. Note that the background contamination is less than 1% of source flux even in the faint X-ray states of the source. The data were rebinned to have at least 25 counts per bin, and the spectral fitting was performed using the energy range between 0.3 keV and 10 keV using XSPEC v.12. The Galactic column density is fixed at 2.2 × 10²⁰ cm⁻² during the fittings as is the case in the RXTE data analysis.

Figure 6 shows a scatter plot between photon index and flux in the X-ray band as measured by Swift-XRT and RXTE-PCA. Generally, a "harder-when-brighter" trend can be seen. Only the highest flux point measured by Swift-XRT shows the photon index significantly harder (smaller) than 1.5.

Zoom In Zoom Out Reset image size

The Swift Ultra-Violet/Optical Telescope (UVOT; Roming et al. 2005) data used in this analysis included all of the observations performed during the time interval MJD 54650–55420. The UVOT telescope cycled through each of the six optical and ultraviolet filters (V, B, U, W1, M2, W2). The UVOT photometric system is described in Poole et al. (2008). Photometry was computed from a 5'' source region around 3C 279 using the publicly available UVOT FTOOLS data reduction suite. The background region was taken from an annulus with inner and outer radii of 275 and 35'', respectively. Galactic absorption in the direction of 3C 279 was adapted as given in Larionov et al. (2008), namely, A_V = 0.093, A_B = 0.123, A_U = 0.147, A_W1 = 0.195, A_M2 = 0.285, and A_W2 = 0.271. The measured magnitudes in each band during the two year observations are m_V = 15.6–18.7 (75 data points), m_B = 16.0–18.8 (80 data points), m_U = 15.1–18.0 (88 data points), m_W1 = 15.3–18.0 (84 data points), m_M2 = 15.4–18.2 (76 data points), and m_W2 = 15.5–18.0 (81 data points). All observed data points are shown in Figure 7.

Zoom In Zoom Out Reset image size

The GLAST-AGILE Support Program (GASP; Villata et al. 2008, 2009) is a project initially originating from the Whole Earth Blazar Telescope⁵⁹ (WEBT) in 2007. It is aimed to provide long-term monitoring in the optical (R band), near-IR, and mm–cm radio bands of 28 γ-ray-loud blazars during the lifetime of the AGILE and Fermi γ-ray satellites.

The observations of 3C 279 in the period considered in this paper were performed by the observatories listed in Table 2. The calibrated R-band magnitudes of the source were obtained through differential photometry with respect to the reference stars 1, 2, 3, and 5 by Raiteri et al. (1998). Near-IR data in the J, H, and K filters were acquired at Campo Imperatore and Roque de los Muchachos (Liverpool). When converting magnitudes into flux densities, optical, and near-IR data were corrected for Galactic reddening using A_B = 0.123 mag (Schlegel et al. 1998). We adapted the extinction laws by Cardelli et al. (1989), and the zero-mag fluxes by Bessell et al. (1998).

For the observations between 2008 August and 2010 August, the measured R-band magnitude ranged from 14.87 to 17.81 (673 data points). The R-band data have the best time coverage among the IR–optical–UV bands in our data thanks to the participation of a number of telescopes. The emission shows strong variability and the excess variance (F_var; Equation (1)) of the source R-band flux (i.e., in linear scale) is 0.853 ± 0.001. The near-IR magnitudes in the J, H, and K bands were measured in ranges of m_J = 14.91–15.59 (20 data points), m_H = 12.04–14.90 (68 data points), and m_K = 11.19–13.45 (20 data points). Those data points are shown in Figure 7. The radio flux densities were measured in ranges of F_5 GHz = 8.5–12.4 Jy (109 data points), F_8 GHz = 9.1–15.5 Jy (124 data points), F_14.5 GHz = 10.3–19.4 Jy (118 data points), F_22 GHz = 10–22 Jy (16 data points), F_37 GHz = 10–20 Jy (168 data points), F_43 GHz = 10–22 Jy (20 data points), F_230 GHz = 5.1–10.5 Jy (62 data points), and F_345 GHz = 6.0–6.8 Jy (7 data points). The light curves of the radio flux densities in those bands are plotted in Figure 8.

Zoom In Zoom Out Reset image size

We performed the V-, J-, and Ks-band photometry and polarimetry of 3C 279 using TRISPEC installed to the 1.5 m Kanata telescope located in the Higashi–Hiroshima Observatory.

TRISPEC has a CCD and two InSb arrays, enabling photopolarimetric observations in an optical and two near-IR bands simultaneously (Watanabe et al. 2005). We obtained 64, 42 and 17 photometric measurements in the V, J, and Ks bands, respectively. A unit of the polarimetric observing sequence consisted of successive exposures at four position angles of a half-wave plates: 0°, 45°, 225, 675. The data were reduced according to the standard procedures of CCD photometry. We measured the magnitudes of objects with the aperture photometry technique. We performed differential photometry with a comparison star taken in the same frame of 3C 279. Its position is R.A. = 12^h56^m1690, decl. = −05°50'430 (J2000) and its magnitudes are V = 13.660, J = 12.377, and Ks = 11.974 (Raiteri et al. 1998; Cutri et al. 2003). The photometric data have been corrected for the Galactic extinction with A_V = 0.093, A_J = 0.026, and A_Ks = 0.010. The measured optical and near-IR magnitudes by Kanata in the V, J, and Ks bands during the two year observations spanned m_V = 15.54–17.27 (56 data points), m_J = 13.00–14.58 (37 data points), and m_Ks = 11.21–11.47 (17 data points). Those data points are also shown in Figure 7.

We confirmed that the instrumental polarization was smaller than 0.1% in the V band using the observations of unpolarized standard stars. Hence, we did not apply any corrections for it. The zero point of the polarization angle is corrected as standard system (measured from north to east) by observing the polarized stars, HD19820 and HD25443 (Wolff et al. 1996). The polarization shows clear variability and the degree of polarization was measured in the range of 3%–36% during our two year observational campaign. As we reported in Paper I, we found a rotation of the polarization angle by 208° together with a sharp drop of the degree of polarization from ∼30% down to a few %. The event was coincident with a γ-ray flare (Periods D and E). In the second half of the two year observations, the source was generally in a quiet state in the optical band, and the degree of polarization was also relatively low.

The Gamma-Ray burst Optical/Near-infrared Detector (GROND; Greiner et al. 2008) mounted at the MPI/ESO 2.2 m telescope at LaSilla observatory in Chile observed the field of 3C 279 in two nights of 2008 July (2008 July 30 and 2008 July 31) and four nights in 2009 January (2009 January 19 to 2009 January 22). In each observation, a total of 4 images in each g'r'i'z' filter with integrations times of 35 s and 24 images of 10 s exposure in each JHK_s were obtained simultaneously.

GROND optical and near-IR data were reduced in standard manner using pyraf/IRAF (Tody 1993) similar to the procedure outlined in Krühler et al. (2008). The stacked images of each observation were flux calibrated against GROND observations of SDSS fields (Abazajian et al. 2009) taken immediately before or after the field of 3C 279 for the optical g'r'i'z', and magnitudes of 2MASS field stars (Skrutskie et al. 2006) for the JHK_s filters. All data were corrected for the expected Galactic foreground reddening of E_{(B − V)} = 0.029 according to Schlegel et al. (1998). Results of GROND observations are summarized in Table 4.

We observed 3C 279 with Spitzer Infrared Spectrograph (IRS), Multiband Imaging Photometer for Spitzer (MIPS), and Infrared Array Camera (IRAC) at several epochs in 2008 and 2009 under the Spitzer program PID50231 (PI: A. Wehrle; see Table 3). The observations were conducted once with each instrument in 2008 July and August and approximately daily during the instrument campaigns in the 2009 February–March visibility window.

For the IRS observations, high-accuracy blue peakup observations on a nearby star were used to center the spectrograph slit on the target. 3C 279 was observed with the low-resolution SL2, SL1, LL2, and LL1 modules, for three cycles of 14 s at each of two nod positions. Spitzer-IRS data reductions began with S18.7 Spitzer Science Center pipeline-processed, background-subtracted data. The background was removed by subtracting the alternate nod for each pointing. Additional processing steps were applied to clean bad data, remove fringes, and match and trim spectral orders. First, we cleaned bad, rogue pixels using the Spitzer Science Center procedure IRSCLEAN V2.0. One-dimensional spectra were then extracted using the standard point-source aperture and flux calibration in SPICE ver. 2.3. We used a custom spectral defringing tool to remove fringes introduced by the pointing-dependent instrumental flat field. This tool uses a predetermined flat-field fringing correction function, which is shifted to match and remove the observed fringes in the spectrum. Spectral orders were trimmed, and the SL2 and SL1 orders were scaled up by a factor of 1.06 to empirically correct for pointing-dependent point-source slit losses. Finally, the nod-spectra were averaged and combined into a single spectrum covering 5.2–35 μm rest wavelength. Figure 9 shows the reduced IRS spectra in the νF_ν representation.

Zoom In Zoom Out Reset image size

We used the pipeline MIPS images (ver. 18) for aperture photometry using 13'', 35'', and 50'' radius for 24, 70, and 160 μm bands, respectively, with aperture corrections from Tables 3.13, 3.14, and 3.16 of the MIPS Data Handbook (ver. 3.2). No 160 μm data were obtained in 2008 August because the array was not cooled during that campaign. 3C 279 has very low ecliptic latitude (02), hence the observed transients can be attributed to passing asteroids which appeared in various MIPS images. We used 6'' radius apertures for IRAC photometry on the pipeline data (ver. 18) with aperture corrections tabulated in Table 5.7 of the IRAC Data Handbook (ver. 3.0).

The Spitzer-MIPS photometric repeatability and absolute calibration uncertainties at 24 μm are, respectively, 0.4% and 4%; at 70 μm, 4.5% and 5%; and at 160 μm, 5% and 12% (Engelbracht et al. 2007; Gordon et al. 2007; Stansberry et al. 2007). We therefore adopt overall uncertainties of 10%, 10%, and 20% at 24, 70, and 160 μm, respectively. No color correction has been applied to the data because the slope and smoothness of the spectrum over the bandpasses are not known. Figure 10 describes the MIPS flux history during the six epochs from 2009 February 15 to 20 together with R-band flux for comparison. No significant flux variation is found in all MIPS bands during those epochs, which include period D.

Zoom In Zoom Out Reset image size

The Spitzer-IRAC calibration uncertainty is 3% overall and has photometric repeatability of 1.5% (Reach et al. 2005). We adopt the overall IRAC calibration uncertainty of 3%, but note the following characteristics of our images. In our IRAC frames, two standard comparison stars used in blazar monitoring were visible in the 3.6 μm images (Star 1 and Star 2).⁶⁰ One comparison star, Star 2, was visible in the 4.5, 5.8, and 8 μm images, located at the interstice of the chopping regions where the data are noisier than elsewhere. The spacecraft orientation, and hence the chopping orientation, was 180° different between 2008 July–August and 2009 March. The standard deviations in comparison Star 2's measurements in 2009 March at 3.6, 4.5, 5.8, and 8 μm are 0.68 mJy, 0.08 mJy, 0.08 mJy, and 0.09 mJy (5%, 1%, 2%, and 3%), respectively. The high 3.6 μm standard deviation was affected by a single high value on 2009 March 11, for which we found no obvious cause; excluding that value resulted in a standard deviation of 0.06 mJy (0.5%). In contrast, the flux of 3C 279 shows a steady decrease of 10%, 12%, 14%, and 13% at 3.6, 4.5, 5.8, and 8.0 μm, respectively, during the six epochs from 2009 March 10–16 as shown in Figure 11.

Zoom In Zoom Out Reset image size

Observations were obtained at mean frequencies of 92.5 and 227.5 GHz using the Combined Array for Research in Millimeter-wave Astronomy (CARMA; Bock et al. 2006). In all cases, the nominal signal to noise ratio exceeded 400 and calibration uncertainties dominated the errors. The source is bright enough to permit self-calibration on timescales of less than a minute and so atmospheric decorrelation was not expected to affect our results significantly even at the long baselines. However, observations in poor weather were not used due to the difficulty of reliably measuring pointing offsets in these conditions.

Data calibration and analysis was done with the MIRIAD software package (Sault et al. 1995). Flux densities were determined by first using phase self calibration with a short enough averaging interval to avoid any atmospheric phase de-correlation, then the flux density was determined from the vector average fringe amplitude at the position of 3C 279 over all baselines. For a strong point source such as 3C 279 this provides very robust and unbiased amplitude estimate independent of the weather or the interferometer baselines. We rely on regular system temperature measurements to provide flux calibration relative to the fixed system sensitivity. The absolute flux calibration of CARMA observations is usually quoted as 10%–15%. However, based on measurements made on the blazar 3C 454.3 we estimated the relative flux calibration at each frequency to be within 5% at 3 mm and 10% at 1 mm. The radio fluxes at 92.5 and 227.5 GHz measured by CARMA correspond to F_92.5 GHz = 11.7–19.7 Jy (14 data points) and F_227.5 GHz = 6.3–9.2 Jy (14 data points). Figure 8 includes the flux history of those radio data.

The Owens Valley Radio Observatory (OVRO) 40 m radio data were collected as part of an ongoing long-term, fast-cadence γ-ray blazar monitoring campaign, described in detail in Richards et al. (2011). Flux densities were measured in a 3 GHz bandwidth centered on 15.0 GHz using dual, off-axis 25 FWHM beams with 1295 separation. Dicke switching against a blank sky reference field to remove gain fluctuations and atmospheric and ground contamination were used. Flux densities from this program are found to have a minimum uncertainty of 4 mJy (mostly thermal) and a typical uncertainty of 3% for brighter sources. During the period included here, 3C 279 was observed as a pointing calibrator. The flux density scale was referred to the value for 3C 286 (3.44 Jy at 15 GHz; Baars et al. 1977) with a scale uncertainty of about 5%. The radio flux at 15 GHz measured by OVRO was ranging from 11.1 to 18.0 Jy among 124 data points during the two year observations. The light curve of the OVRO radio data is also plotted in Figure 8.

The multi-band light curves of 3C 279 are presented in Figure 12. They include (a) γ-ray flux above 200 MeV (Fermi-LAT), (b) 2–10 keV X-ray flux measured by Swift-XRT and RXTE-PCA, (c) optical-UV fluxes in R band (GASP), V band (Swift-UVOT and Kanata), and W2 band (Swift-UVOT), (d,e) degree and angle of optical polarization (Kanata and KVA), and (f) radio fluxes in the 230, 37, 15, and 5 GHz bands (GASP, CARMA, and OVRO). We note that the X-ray fluxes determined by Suzaku and XMM-Newton are entirely consistent with those plotted in Figure 12. The extensive data set obtained in many bands for 3C 279 allows us to make general statements regarding the relative flux variability in various spectral bands, and the relationship of the time series to each other. The first such feature of the multi-band light curves is a general—although not exact—trend where the IR through optical emission seems to be correlated with the γ-ray flux. We calculated the discrete correlation function (DCF; Edelson & Krolik 1988) to quantify the correlation of the flux variations between the γ-ray and other bands, and to determine whether we can measure any clear lag between the bands.

Zoom In Zoom Out Reset image size

For the DCF calculations, we use the γ-ray fluxes averaged over an interval of 1 day as shown in the top panel of Figure 1. The resulting DCF between γ-ray and optical R-band fluxes is shown in Figure 13. Positive values of "τ" correspond to flux variations in the γ-ray band lagging flux variations in the other bands. In the DCF between γ-ray and optical R-band fluxes, a peak can be seen close to zero lag. We fit the DCF data points in the range between −30 and 5 days using a Gaussian function of the form DCF(τ) = C_max × exp [(τ − τ₀)²/σ²], where C_max is the peak value of the DCF, τ₀ is the time at which the DCF peaks, and σ is the Gaussian width of the DCF. The fit yields a position of the peak at τ₀ = −10.7 ± 0.7 days, corresponding to a value of C_max = 1.07 ± 0.03 with a dispersion of σ = 19.4 ± 1.4 days. The result implies that the optical emission is possibly delayed with respect to the γ-ray emission by about 10 days.

Zoom In Zoom Out Reset image size

In the framework of the one-zone synchrotron + external-radiation Compton (ERC) models, the same electron population, of roughly the same energies, is responsible for the radiation in both the optical and γ-ray bands. There, the observed lag can result from different profiles of the decreasing magnetic and radiation energy densities along the jet: we show that idea quantitatively in the Appendix. As is shown there, a very steep drop of the external radiation energy density is required to explain the lag in a conical jet with magnetic field B'∝1/r, where r is the distance along the jet. This condition can be relaxed in the scenario involving the re-confinement of a jet (e.g., Daly & Marscher 1988; Komissarov & Falle 1997; Nalewajko & Sikora 2009). In such a case, the magnetic field intensity can drop more slowly than 1/r. If the lag of the optical emission is confirmed, the application of the results in the Appendix to the ∼10 day lag may imply the location of the active "blazar zone" at distances of a few pc in agreement with those postulated to explain the optical polarization swing (Periods D and E) in terms of a region containing an enhanced density of ultra-relativistic electrons propagating along a curved trajectory (Paper I). It is worth noting that similar γ-ray/optical lags have been reported during the outbursts of 3C 279 in early 1999 Hartman et al. (2001b), of PKS 1502+106 in 2008 (Abdo et al. 2010d), of PKS 1510−089 in early 2009 (Abdo et al. 2010a; D'Ammando et al. 2011), and of AO 0235+164 in late 2008 (Agudo et al. 2011b; Ackermann et al. 2012). On the other hand, no significant lags between γ-ray and optical signals have been detected in 3C 454.3 in late 2008 (Bonning et al. 2009; Jorstad et al. 2010), in 3C 66A in 2008 October (Abdo et al. 2011), and in OJ 287 in 2009 October (Agudo et al. 2011a). Based on investigations of long-term light curves of 3C 454.3 during 2008–2010, Raiteri et al. (2011) have shown that the optical and γ-ray flux variations are not always simultaneous and have proposed a geometrical scenario to explain the change in the γ/optical flux ratio during the outburst peaks in 3C 454.3. It is expected that the on-going multi-band monitoring of blazars will enable us to quantify such lags and find out how common they are.

Different behavior is apparent in the radio flux, where the energies of radio-emitting electrons are very different from the energies of the electrons involved in producing the observed optical and γ-ray emission. Variability appears to be much less rapid, and the excess variance (F_var; see definition in Equation (1)) in the radio regime is quite modest; for instance, 0.145 ± 0.004 at 37 GHz, 0.165 ± 0.001 at 15 GHz, and 0.104 ± 0.001 at 5  GHz. Those values are significantly less than ones in the γ-ray or optical bands. This suggests that the synchrotron emission from the γ-ray emitting region is self-absorbed at these wavelengths. The observed radiation is produced at much larger distances, where the light-travel effects smear out the sharp, rapid variability patterns observed in the optical and γ-ray bands.

Perhaps the most surprising behavior—and difficult to explain in the context of simple, one-component, single-zone models—is the relationship of the X-ray light curve to those in the IR–optical or γ-ray bands. In Paper I, we reported that the X-ray time series exhibits a relatively rapid, symmetrical flare at ∼ MJD 54950 (Period F) with a duration of ∼20 days, which is not accompanied with any prominent IR/optical or γ-ray flares. As we argued in Paper I, the hard (rising in νF_ν representation) X-ray spectrum is unlikely to be the "tail" of the synchrotron emission, but instead, it is more likely to be produced by the low-energy end of the electron distribution radiating via inverse-Compton process.

The continuing monitoring of the object in the X-ray band revealed another X-ray flare at ∼MJD 55040 (Period G), ∼90 days after the first X-ray flare. The separation of the two X-ray flares is remarkably close to the temporal separation of the two γ-ray flares, with the two pairs delayed with respect to each other by ∼155 days. Figure 14 presents the calculated DCF between γ-ray and X-ray fluxes, which shows a modest peak at ∼ − 155 days with a correlation coefficient of 0.6–0.7 and indicates no correlation between the γ-ray and the X-ray bands with zero lag. While confirming the physical connection of the two pairs would be very important, we cannot currently envision any situation where the two would be causally connected; the 155 day lag would imply the distance of the X-ray flare production ∼155Γ²_j lt-day ∼ 50(Γ_j/20)² pc and at such a distance should be accompanied by radio flares, which are not seen in our data. In such a scenario, the X-ray flares should be significantly broadened compared to the γ-ray flares; however, we observe a similar temporal structure in both bands. Furthermore, we note that there are some optical and γ-ray peaks that might well be associated with the second X-ray flare. Hence, it is possible that the two prominent γ-ray/optical flares (Periods B and D), together with the subsequent two X-ray flares (Periods F and G), form a sequence of four events separated by a similar time intervals. Those intervals, in turn, can be possibly determined by instabilities in the jet launching region. Here, the different broadband spectra during these events may result from small changes of parameters, such as the jet direction, Lorentz factor, and/or location and geometry of the dissipation event.

Zoom In Zoom Out Reset image size

A weak (and sporadically almost absent) correlation between X-rays and other spectral bands can also result from such processes that preferably contribute to radiation in the X-ray band. They can be related to the following three mechanisms/scenarios.

• 1.  
  Bulk-Compton process. This involves Compton-scattering of ambient optical/UV light by the cold (non-relativistic) electrons in the jet. This mechanism is most efficient close to the accreting black hole where the processes responsible for the variability of X-rays may operate independently of those at larger distances and producing there variable non-thermal radiation (Begelman & Sikora 1987). A drawback of this scenario can be that the bulk-Compton spectrum is predicted to have a similar shape as the spectrum of the external radiation field (Ackermann et al. 2012), which significantly differs from what we observe in the X-ray band.
• 2.  
  Inefficient electron acceleration. Acceleration of the relativistic electrons at proton-mediated shocks is likely to proceed in two steps: in the first one low-energy electrons may be pre-accelerated via, for example, some collective processes involving protons; in the second step, they may participate in the first-order Fermi acceleration process. If under some conditions the electron–proton coupling is inefficient, the fraction of electrons reaching the Fermi phase of acceleration will be small. In this case the X-rays, originating from lower energy electrons, are produced efficiently, while the γ-rays and optical radiation that involve more relativistic electrons are not.
• 3.  
  The X-rays can be also contributed by hadronic processes, specifically by the pair cascades powered by protons losing their energy in the photo-mesonic process (Mannheim & Biermann 1992). For this process to be efficient, it requires extreme conditions (Sikora et al. 2009; Sikora 2011); however, operating in the very compact central region, at distances less than few hundred gravitational radii, it may occasionally dominate in the X-ray band.

Figures 15 and 16 show broadband SEDs of 3C 279 in all periods as defined in Table 1. In addition, we also extracted an SED using data taken on 2008 July 31 (MJD 54678), which has a good energy coverage of the synchrotron emission component including Spitzer and GROND data, although the γ-ray data by Fermi-LAT are not available at that time because this was before the beginning of normal, all-sky science observations with Fermi-LAT. Both SEDs for Period D (2009 February; corresponding to the brightest γ-ray flare coincident with the optical polarization swing) and Period F (2009 April; corresponding to the first isolated X-ray flare) have already been partially reported in Figure 2 of Paper I. New Spitzer-MIPS data points are included in the SED for Period D in this paper. In Period C, there are observations by MAGIC, which provide upper limits above 100 GeV (Aleksić et al. 2011). For comparison, we also include VHE γ-ray fluxes detected with the MAGIC telescope in 2006 February as gray points (Albert et al. 2008) in Figure 15.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

This is the richest set of time-resolved spectra ever collected for this source. The spectral coverage of the synchrotron bump is unprecedented, allowing us not only to constrain the parameters of the emission models, but also to study their time evolution. As we discussed in Section 2.4, the shape of the γ-ray spectrum deviates from a simple power law, in similarity to other FSRQ blazars. Strong variability, over one order of magnitude, is evident in near-IR/optical/UV and γ-ray bands. This contrasts with the moderate variability in the radio/mm and X-ray bands.

Particularly interesting is the behavior of this source in the mid-IR band, around ∼10¹³ Hz, where significant spectral variability is observed. In the low state in Period A, the mid-IR spectrum is relatively soft and can be extended with a power-law shape to the optical/UV band. In this case, the synchrotron component peaks in the mm/sub-mm band (∼10¹¹–10¹² Hz). However, in the high state in Periods D and E, the mid-IR spectrum is much harder and shows a significant curvature. In Period D, there is a clear spectral break at ∼3 × 10¹² Hz (∼100 μm). The spectral index between the 70 μm (∼4.3 × 10¹² Hz) and 160 μm (∼1.9 × 10¹² Hz) points is α_{70 − 160} = 0.35 ± 0.23, taking into account systematic errors described in Section 3.9. The synchrotron peak is located in the mid-IR band, at a frequency one order of magnitude higher than in the low state. This indicates that there are two independent synchrotron bumps, possibly produced at different locations. The mid-IR-peaking component, seen only in the IR/optical/UV flaring state, is characterized by a strong and rapid variability. The mm/sub-mm peaking component is more persistent and dominates when the source is in the low state. The complex shape of the SED between the mm band and the 70 μm point in Period D requires a coexistence of these two components. A similar scenario of multiple synchrotron components was investigated in the case of 3C 454.3 by Ogle et al. (2011).

In the X-ray band, despite the smaller variability amplitude, we observe some spectral changes. In particular, in Periods F and G, which represent the two isolated X-ray flares, the spectrum is very similar and harder than on average. Figure 15 shows that these flares are not energetically important. If we extrapolate the X-ray spectra with power laws to the γ-ray band, we underpredict the observed γ-ray flux in Periods B, C, D, and H. Periods B, C, and D coincide with the high-activity γ-ray state. This indicates that the X-ray flux cannot originate from the same emission component as the γ-ray flux, at least in the flaring state. Because the γ-rays are correlated with the optical flux but not with the X-ray flux, the γ-rays can be related to the mid-IR-peaking synchrotron bump while the X-rays may correspond to the mm/sub-mm peaking synchrotron bump. We explore this possibility when modeling the SEDs at Periods A and D in Section 5.2.

An intrinsically spherically symmetric emitting region is expected to produce the observed electric polarization vector aligned with the projected velocity of the emitting region. Nalewajko (2010) presented a simple model of its trajectory to explain the event of simultaneous smooth variations of the polarization degree and angle during the polarization swing which has been reported in Paper I. This model adopts a constant jet Lorentz factor Γ_j = 15 and can be used to predict the viewing angles for a given observation time. For Period D we estimate θ_{obs, D} ∼ 15, while for Period E: θ_{obs, E} ∼ 24. Between Periods D and E (Δt ∼ 15 days), the emitting region propagates over a distance Δr ∼ Γ²_jc(Δt) ∼ 2.8 pc.

In Figure 17 we show Model D1 fitted to the SED in Period D at r = r_IR and Model E1 fitted to the SED in Period E at r = r_IR + Δr. Model parameters are listed in Table 5. Both models use the magnetic field scaled to the same value at the distance of 1 pc. In order to explain the difference in the luminosity ratio of the ERC component and the synchrotron component, which decreased by factor ∼4 between Model D1 and Model E1, we assume a distribution of the comoving IR radiation energy density dropping steeply with distance, adopting β_IR ∼ 4. This corresponds to a strongly stratified torus structure, with a significant concentration of hot dust very close to the sublimation radius (see, e.g., Mor & Netzer 2012). We should note that, although the relatively soft γ-ray spectrum was observed at Period E (Γ = 2.64 ± 0.32, see Table 1), the peak of the ERC-IR component in the Model E1 falls at ∼800 MeV, in the Fermi-LAT band.

Zoom In Zoom Out Reset image size

The far-IR spectral break in Period D requires a sharp break in the electron distribution function at γ_br1 = 800. As the cooling break is expected at γ_c ∼ 3m_ec²/(2σ_TRu'_ext) ∼ 660 (where R is a radius of source emission region and u'_ext is the energy density of the external radiation in the jet comoving frame), γ_br1 is located just within the fast-cooling regime. The electron distribution in the fast-cooling regime cannot be harder than p = 2, hence the resulting synchrotron spectral index α = (p − 1)/2 should be larger than 0.5. In fact, the mean value of the observed spectral index between 70 μm and 160 μm is smaller than 0.5 (α_{70 − 160} = 0.35 ± 0.23), which cannot be explained if the electron cooling is efficient. However, the uncertainty of the measurement does not allow us to reject this scenario.

Alternatively, if the jet precession can cause the observed γ-ray flare event with the polarization swing, the γ-ray/optical emission can be generated much closer to the central black hole, even within the broad-line region (see also in Paper I). Therefore, we also attempted to model Period D placing the emitting region at r_BEL. For Γ_j = 15, with model parameters fitted using the synchrotron and ERC components, the X-ray flux is overproduced by the SSC process. To alleviate this problem, we increased the jet Lorentz factor to Γ_j = 20. In Figure 18, we show Model D2 with parameters listed in Table 5. The magnetic field strength scaled to the distance of 1 pc is almost the same as the value in Model D1. Because of a smaller size of emission region and higher energy density of the locally produced synchrotron radiation, the synchrotron self-absorption is able to produce a spectral cutoff at a higher frequency of ∼3 × 10¹² Hz (∼100 μm), consistent with the far-IR break. This interpretation has an advantage that it also could explain the observed hard spectral index between 70 μm and 160 μm, even smaller than 0.5, independently of details of the electron energy distribution.

Zoom In Zoom Out Reset image size

The low-energy synchrotron component, dominating the mm/sub-mm band, must be produced in a much larger region, placing it far outside the broad-line region. In Figure 18, we present Model A2, fitted to the SED at Period A. We kept the Lorentz factor and the magnetic field consistent with Model D2, but we set the source at the distance ∼4 pc, the same as in Model E1. This low-state model of Period A can reproduce both observed X-ray and γ-ray spectra by a single broken power-law electron distribution. The γ-ray spectral index is consistent with the IR/optical/UV spectral index. The synchrotron self-absorption is effective at ∼10¹¹ Hz and the spectral peak is located in the mm/sub-mm band.

Those results suggest the existence of two synchrotron components: one peaking in the mm/sub-mm band and the other peaking in the mid-IR band. The component with the peak in the mid-IR band is more variable, and can be produced at shorter distances, within the broad-line region, where the far-IR break can be explained by the synchrotron self-absorption.