Exploring the Connection between Parsec-scale Jet Activity and Broadband Outbursts in 3C 279

The 7 mm VLBA observations of the source were obtained in the course of a program¹⁴ of monthly monitoring of bright γ-ray blazars at 43 GHz. The details of the observations can be found in Jorstad et al. (2017). The data reduction is performed using the Astronomical Image Processing System (AIPS) and Difmap (Shepherd 1994). Details of the data reduction are described in Jorstad et al. (2005, 2017). We model-fit the brightness distribution of the source using circular Gaussian components using Difmap, and the model-fit parameters (peak flux density, distance from (0, 0) coordinates, position angle, and size) are listed in Table 2 in the appendices. The polarization parameters (polarized flux density and electric vector position angle (EVPA)) are determined by model-fitting the stokes Q and U in (u, v) data. The polarized flux density is given by $\sqrt{{Q}_{\mathrm{flux}}^{2}+{U}_{\mathrm{flux}}^{2}}$ and the EVPA by 0.5 × arctan(U_flux/Q_flux).

The 230 and 345 GHz data are provided by the Submillimeter Array (SMA) Observer Center¹⁵ database (Gurwell et al. 2007). The 230 GHz total flux density data are complemented by the data from the Polarimetric Monitoring of Active Galactic Nuclei at Millimetre Wavelengths (POLAMI) program¹⁶ (Agudo et al. 2018a, 2018b; Thum et al. 2018). POLAMI is a long-term program to monitor the polarimetric properties (Stokes I, Q, U, and V) of a sample of around 40 bright AGNs at 3.5 and 1.3 mm wavelengths with the IRAM 30 m telescope near Granada, Spain. The program has been running since 2006 October, and it currently has a time sampling of ∼2 weeks. The XPOL polarimetric observing setup has been routinely used as described in Thum et al. (2008) since the start of the program. The reduction and calibration of the POLAMI data presented here are described in detail in Agudo et al. (2010, 2014, 2018a). We also make use of the 15 GHz Owens Valley Radio Observatory (OVRO) data publicly available at the OVRO website.¹⁷ Details of the observations and data reduction are described in Richards et al. (2011).

During our campaign period, the source was simultaneously observed in six optical/UV filters (U, B, V, W1, W2, and M2) by Swift/UVOT (Roming et al. 2005). We used the uvotsource tool from HEASoft v6.16 for the data analysis (Poole et al. 2008). A step-by-step data reduction procedure is described in Rani et al. (2017), to which we refer for details. We have also included optical/IR data at the B, V, R, J, and K passbands from the Small and Moderate Aperture Research Telescope System (SMARTS5) monitoring program. The reduction and analysis of this data are described in Bonning et al. (2012).

We also make use of the optical R-passband flux and polarization observations provided by the Steward Observatory blazar monitoring program¹⁸ ; the details of the observations and data reduction are described in Smith (2016). Additional optical R-band polarimetric observations were obtained at the 1.83 m Perkins telescope of the Lowell Observatory (Flagstaff, AZ) using the PRISM camera.¹⁹ The details of the observations and data reduction are given in Jorstad & Marscher (2016). We also used polarimetric and photometric data from the AZT-8 (Crimean Astrophysical Observatory) and LX-200 (St. Petersburg, Russia) telescopes. The details of the observations and data reduction are given in Larionov et al. (2008).

Finally, the data were corrected for the Galactic extinction using the coefficients of Schlegel et al. (1998) for different filters.²⁰ We converted the magnitudes to fluxes using the central wavelengths for each filter as calibrated by Poole et al. (2008) for the space-based observations. For the ground-based observations, the zero-point fluxes²¹ corresponding to the observed wavelengths were used for magnitude-to-flux conversion.

X-ray observations (0.3–10 keV) of the source were obtained using the Swift/XRT telescope (Burrows et al. 2005) in both the photon-counting and windowed-timing mode. The data are analyzed using the standard xrtpipeline v0.13.2. The spectra in individual time bins were modeled with XSPEC v12.9.0 to extract the photon counts and indices. We used the absorbed power law with the total H i column density fixed to the Galactic value (N_{H i} = 0.212 × 10²¹; Kalberla et al. 2005) to model the 0.3–10 keV spectrum, which provided an acceptable fit at all epochs. The details of the observations and data reduction are discussed in Rani et al. (2017).

Here we employed the 100 MeV–300 GeV data of the source observed in survey mode by the Fermi/LAT (Atwood et al. 2009). The LAT data were analyzed using the standard ScienceTools (software version v10.r0.p5) and instrument response function P8R2_SOURCE_V6. Photons in the Source event class were selected for the analysis. We analyzed a region of interest (ROI) of 20° centered at the position of 3C 279 using a maximum-likelihood algorithm (Mattox et al. 1996). For the unbinned likelihood analysis,²² all 3FGL (Acero et al. 2015) sources within the ROI were included. We also included the Galactic diffuse background (gll_iem_v06.fits) and isotropic background (iso_P8R2_SOURCE_V6_v06.txt) emission components.²³ Model parameters for sources within 5° of the center of the ROI were kept free. Model parameters for variable sources (variability index ≥72.44 in the 3FGL catalog) were also set free, while the model parameters of nonvariable sources beyond 5° were fixed to their catalog values.

To characterize the variability properties of the source, we computed the photon-flux light curve of the source at E above E₀, where E₀ is the decorrelation energy (Lott et al. 2012), which minimizes the correlations between integrated photon flux and photon index. During our campaign, we obtained E₀ = 219 MeV. We generated the constant uncertainty (15%) light curve above E₀ through the adaptive-binning method following Lott et al. (2012). The bin size ranges between 0.07 and 31.06 days with an average of 5.43 days. The adaptive binned light curve is produced by modeling the spectra by a simple power law, N(E) = N₀ E^−Γ, N₀ : prefactor, and Γ : power-law index.

The observed broadband flux density variations in 3C 279 during a major γ-ray outburst are plotted in Figure 1. The source became active at GeV energies in early 2013 November, and the flaring activity continued until 2014 August. Hereafter, we will use JD' (defined as JD—2,455,000) for time. In Figure 1, the cyan dashed lines labeled "A" and "D" mark the beginning and end of the outburst, respectively. The source displayed multiple flares across the whole electromagnetic spectrum over this period, which we visually inspected and labeled as follows. The blue dotted-dashed lines (1–6) are marked close to the peak of the rapid γ-ray flares labeling the broadband flares "G" for γ-rays, "X" for X-rays, "O" for optical, and "R" for radio, followed by the number. For instance, the γ-ray flares should be read as G1–G6.

Zoom In Zoom Out Reset image size

Flare 2 is observed close in time at γ-ray and X-ray energies. The first two flares are accompanied by an increase in the source brightness at radio frequencies. At optical frequencies (panel (c)), the source is still in the quiescent phase. The peak of flare G3 coincides with a minimum in the X-ray light curve but has optical and radio counterparts and a delayed response in X-rays. In comparison to X1 and X2, X3 has a rather low magnitude. Because of the gap in observations at X-ray energies, it is hard to determine the absence/presence of an X-ray counterpart for flare G4, but the flare has a near-simultaneous optical and radio counterpart. A prominent optical flare (labeled "3.5") is seen between flares 3 and 4. The flare O3.5 is accompanied by similar activity at the radio (230 GHz) band; however, the observed variations are quite mild at X-ray and γ-ray energies. Flare 5 is the only one that is observed quasi-simultaneously at all energies (radio to γ-rays). Flare 5 has a double feature in γ-rays and X-rays (see Figures 2 and 4).

Zoom In Zoom Out Reset image size

The other two interesting states in the broadband activity are marked "B" and "C." After the first five γ-ray flares, a local minimum (B) was observed simultaneously at all frequencies. The local minimum was followed by a concurrent flare (flare 6) at radio, optical, and γ-ray energies; no significant activity was observed in the X-ray regime during this period. Period C marks the end of flare 6, which happened to be almost simultaneous for all light curves. A continuous decay in the source brightness was observed between periods C and D, which we refer to as the post-outburst plateau.

3.1.1. Gamma-Ray Variability

Figure 2 displays details of the observed γ-ray flaring activity of 3C 279. The rapid flares (labeled 1–6) were superimposed on top of a broad outburst. The source started to brighten from a flux level F (E > 219 MeV) ∼ 3.3 × 10⁻⁷ photons cm⁻² s⁻¹ on JD' ∼ 1610 and returned to the same brightness level on JD' ∼ 1850. The green lines in Figure 2 mark the observed brightness level (∼3.3 × 10⁻⁷ photons cm⁻² s⁻¹) of the source before and after the outburst period. Significant variations in γ-ray photon index were observed during this period. The estimated Γ value is 2.7 at the beginning and end of the outburst. A significant spectral variation is observed for the individual flares (1–6). For each flare, the spectrum gets harder at the peak, and later it softens again. The fastest observed γ-ray flare during this activity period had a flux-doubling timescale of ∼2 hr and a very hard γ-ray spectrum with Γ = 1.7 (Hayashida et al. 2015).

In Figure 3, we plot the γ-ray photon flux values as a function of photon index. A clear spectral hardening with an increase in source brightness can be seen here. Since we investigate the correlation above the decorrelation energy (E₀), the estimated correlation is not sensitive to the LAT instrumental bias. A Pearson correlation analysis reveals a significant correlation between the two. Formally, we obtained r_P²⁴  = −0.33 and a p-value = 0.0019, which suggests a significant anticorrelation at a confidence level of >99%. It is therefore evident that not only the major outburst but also the individual γ-ray flares are characterized by significant spectral hardening, although with substantial scatter.

Zoom In Zoom Out Reset image size

3.1.2. X-Ray Variability

The X-ray energy flux and photon index light curves are shown in Figure 4. Compared to the γ-ray light curve, the X-ray observations were rather sparse. We could identify two prominent flares in the X-ray light curve. Apparently, the first X-ray flare coincides with flare G2 in γ-rays, and the second X-ray flare coincides with flare G5. While the X-ray energy flux is varying by a factor of more than four, the photon index remains consistent with the mean value of Γ_X = 1.47 ± 0.01.

Zoom In Zoom Out Reset image size

3.1.3. Variability at Optical Frequencies

The observed near-IR, optical, and UV light curves of the source are shown in Figure 5. Similar to γ-rays, there are two modes of flaring activity at optical frequencies. Fast flares (timescale ∼20–30 days) are superimposed on the broad outburst. No optical observations are available before JD' = 1635 (2013 December). The flaring behavior is very similar for the near-IR, optical, and UV light curves.

Zoom In Zoom Out Reset image size

The average IR–optical–UV (hereafter IOU) spectrum over the entire period of our observations is shown in Figure 6 (top). The observed IOU spectrum is steep until an upturn at UV frequencies. The excess emission could be due to thermal dominance either from the accretion disk or a broad-line region. The observed IOU spectrum can be well described by a power law, F(ν) ∝ ν^−α, where α is the spectral index. For the average spectrum, we obtained α = 1.91 ± 0.08.

Zoom In Zoom Out Reset image size

To investigate the spectral variations during the broad outburst, we constructed the IOU spectrum every 20 days. The spectral index is estimated via fitting the spectrum by a power law as described above. Prominent variations in the IOU spectral indices can be seen in Figure 6 (bottom). The IOU spectrum is steeper at the beginning of the outburst (α_IOU ∼ 2, Flare 2), and the spectrum gets harder as the source brightness increases. The estimated α_IOU is ∼1.6 (Flare 6) at the peak of the outburst (JD' ∼ 1720). However, the spectral variations are quite marginal during the decay phase of the outburst, with α_IOU remaining nearly consistent with a value of ∼1.6.

Figure 7 presents the optical polarization variations observed in the source. The fractional polarization drops from ∼25% to ∼5% within the first 50 days of our observations; this period coincides with the beginning of the γ-ray outburst. The sharp drop in percentage polarization (PP) was followed by a slow rise (superimposed with rapid variations) until 2014 August. At the end of the outburst in 2014 August, the PP returned to its original value of ∼25%. The optical polarization angle (EVPA, χ) variations, on the other hand, are quite modest (Δχ ∼ 60°). Throughout the campaign period, the EVPA remained consistent with an average value of ∼50°, which is along the parsec-scale jet direction of ∼−135° (Jorstad et al. 2017).

Zoom In Zoom Out Reset image size

3.1.4. Variability at Radio Frequencies

In Figure 8, we plot the radio flux density light curves sampled at the 15, 86, 230, and 350 GHz bands. The observed variations appear quite similar at 15 and 230 GHz bands, the most densely sampled frequencies. Faster variations (for instance, the flare between 3 and 4) superimposed on the broad outburst can be seen here. A prominent flare is also apparent in the 86 GHz light curve; however, the data sampling is quite sparse close to the peak of the outburst. The data sampling is even worse at 350 GHz, and as a result, the flaring activity is less evident.

Zoom In Zoom Out Reset image size

The PP (top) and EVPA (bottom) curves at the 86 and 230 GHz radio bands are shown in Figure 9. There are significant variations in the 230 GHz band PP curve. For instance, the PP drops from 15% to 5% between JD' = 1600 and 1700, and it increases again to 15% at JD' ∼ 1720. The observed polarization was quite high (∼15%) for the rest of the period. On the other hand, the PP variations at the 86 GHz band and the EVPA variations at both the 86 and the 230 GHz bands were quite modest (Δχ ≤ 25°).

Zoom In Zoom Out Reset image size

3.2.1. Fractional Variability

To quantify how the amplitude of flux variations changes as a function of frequency, we compared the fractional variability amplitude (F_var) of the observed light curves. For a given light curve with an average flux $\overline{x}$ and sample variance S², F_var is defined as (Vaughan et al. 2003)

$$\begin{eqnarray}&&{F}_{\mathrm{var}}=\sqrt{\displaystyle \frac{{S}^{2}-\overline{{\sigma }_{\mathrm{err}}^{2}}}{{\bar{x}}^{2}}},\end{eqnarray} \tag{ 1 }$$

where $\overline{{\sigma }_{\mathrm{err}}^{2}}$ is the mean of the squared measurement uncertainties. The uncertainty on F_var due to the measurement-error fluctuations was derived through Monte Carlo simulations by Vaughan et al. (2003, Equation (B2)).

Figure 10 shows the calculated F_var at different frequencies. A gradual increase in F_var with frequency is visible in the radio regime, and this trend continues until the IOU regime. The first peak in the F_var-versus-frequency curve is observed at the optical U passband, which is followed by a dip in the X-ray regime. Compared to the lower frequencies, the F_var is extremely large at γ-rays; F_var > 1 implies more than 100% variations.

Zoom In Zoom Out Reset image size

In Figure 11, we display an example image of the 3C 279 jet. The top panel shows the circular Gaussian components (yellow circled crosses) superimposed on the total intensity contours. The bottom panel shows the composite superresolved image convolved with a beam size of 0.1 mas (0.63 pc). In the composite image, the contours represent the total intensity, and the color scale represents the polarized intensity. The yellow line segments mark the EVPA direction.

Zoom In Zoom Out Reset image size

The total intensity of the jet is described using circular Gaussian components, and we list the model fit parameters in Table 2 in the appendices. To study the evolution of flux density and the polarization of components, as well as the component motion and its evolution as a function of time, we cross-identify the features using the following assumption. We assume that the changes in the flux density, distance from the VLBI core, position angle, and size should be small for the time period between adjacent epochs to prevent a potentially large systematic error arising from the incorrect cross-identification.

3.3.1. Component Motion

The VLBI core centered at R.A. 12 56 11.167, decl. −05 47 21.53 is chosen as a reference point and is fixed to coordinates (0, 0) to study the kinematics of individual components. The absolute position of the core could change because of changes in opacity, pressure, or instabilities in the jet (see Hodgson et al. 2017 and Lisakov et al. 2017 for details). However, whether we fix the core position to (0, 0) or leave it free does not affect the relative separation of the components with regard to the core. During the course of our campaign period, we identified a total of six components—C1, C2, C3, NC1, NC2, and NC3—in addition to the core. Figure 12 shows the evolution of the radial distance of the components from the core (a) and their trajectories in the X–Y plane²⁵ projected on the sky (b). The flux density variations in the individual knots are shown in Figure 12(c).

Zoom In Zoom Out Reset image size

In 2013 August, the jet's brightness could be well described using five components (C1–C3 and two unidentified components) in addition to the core (see Table 2 in the Appendices). A new component (NC1) appeared in 2013 November, followed by the ejection of two more components in 2014 May (NC2) and December (NC3). The core brightened before the ejection of a new component, and later the component appeared in the jet. All of these components exhibit significant motion, especially in the radial direction. As seen in Figure 12(b), the components tend to follow the same linear trajectories in the X–Y plane up to a distance of ∼0.5 mas. This is in contrast to what we usually observe for blazars. Wiggly/curved trajectories of components are often seen for blazars (see Chatterjee et al. 2008; Lister et al. 2013, Rani et al. 2014; 2015; and references therein).

To study the component motion in the radial direction, we fitted the radial separation of the components relative to the core using r(t) = μ × (t − T₀), where μ is the angular speed in mas yr⁻¹ and T₀ corresponds to the ejection time of the components from the core. The solid lines in Figure 12(a) represent the best-fit functions. The fit provides the angular speed (μ), which we used to calculate the apparent velocity following ${\beta }_{\mathrm{app}}=\tfrac{\mu .{D}_{L}}{c.(1+z)}$, where D_L is the luminosity distance and z is the redshift of the source. In Table 1, we list the angular and apparent speeds of the components. Back-extrapolation of the components' motion is used to estimate the ejection time of the new components (NC1–NC3). The calculated ejection times of the new components are listed in Table 1. The highest detected apparent speed (β_app) can be used to get an estimate of the lowest possible Lorentz factor, ${\rm{\Gamma }}\geqslant {(1+{\beta }_{\mathrm{app}}^{2})}^{0.5}$, and maximum viewing angle, θ_view ≤ sin⁻¹(1/β_app). Using β_app = 22.4, we get Γ ≥ 22.4 and θ_view ≤ 26.

Table 1.  Kinematical Parameters of the Jet Emission

Component	μ (mas yr−1)	βapp (c)	T0 (days)
C1	0.49 ± 0.04	15.8 ± 1.2	⋯
C2	0.58 ± 0.03	18.4 ± 0.8	⋯
C3	0.69 ± 0.02	21.8 ± 0.7	⋯
NC1	0.71 ± 0.02	22.4 ± 0.7	${1558}_{+15}^{-28}$
NC2	0.55 ± 0.03	17.5 ± 0.8	${1638}_{+40}^{-27}$
NC3	0.61 ± 0.04	19.2 ± 1.2	${1810}_{+26}^{-19}$

Note. μ: angular speed. β_app: apparent velocity. T₀: component's core separation time.

Download table as:  ASCIITypeset image

Figure 13 presents a comparison of the variations observed in PP at optical and radio frequencies (the left y-scale is for the optical data, while the right is for the radio data). Although it is difficult to make a one-to-one comparison between the PP variations at the two frequencies because of sparse sampling of the radio data, we do see a similarity. For instance, in the beginning, the PP is high at both the optical and radio bands. A drop in PP (between JD' 1600 and 1680) is observed almost simultaneously at the two frequencies. An increase in PP appears after the dip observed in the optical and 230 GHz data but not in the 43 GHz VLBI data. The PP in the core region remains rather constant after the dip. On the other hand, the PPs at the optical and 230 GHz radio frequencies continue rising until they regain their initial value at the end of the observations.

Zoom In Zoom Out Reset image size

Figure 13 (bottom) shows a comparison of the EVPA variations at the optical and radio frequencies. In contrast to PP, EVPA variations are rather modest. A peak-to-peak (minimum-to-maximum) change of ∼60° is observed in the optical and 43 GHz VLBI data, and in the 230 GHz data, it is ≤25°. The average EVPA values at the three frequencies are also quite similar (∼40°–60°). This indicates that the optical and radio polarization angles (EVPAs) are roughly parallel to the jet direction²⁶ (∼−135°). We noticed a change in the VLBI core EVPA during the ejection of the new components (NC2 and NC3), which could be because of a different EVPA orientation of the new component(s).

The PP (top) and EVPA (bottom) variations observed in the new ejected components are shown in Figure 14. The PPs of the components gradually increase after their separation from the core and stay as high as ≥20%. This could result from either a transition from the optically thick to thin regime at 43 GHz or a progressive ordering of the magnetic field with distance down the jet. The EVPAs of the new components (especially NC2) show some modest variations but still remain aligned along the jet axis.

Zoom In Zoom Out Reset image size

In Figure 15, we show a comparison of optical flux density and polarization variations. We used an inverse y-scale for the optical PP data. During the first three γ-ray flares (1–3), the optical flux is rising, and a decay is seen in the PP. The peaks of flares 4 and 5 coincide with the dips in the PP curve, indicating a clear anticorrelation between the two. The anticorrelation is even more evident in the PP–versus–optical flux plot on the right. A formal linear Pearson correlation test suggests a significant anticorrelation between the optical flux and polarization data; we obtained r_P = −0.76 at the >99.99999% confidence level.

Zoom In Zoom Out Reset image size

A comparison of the optical PP and γ-ray photon flux variations is shown in Figure 16 (left; again, an inverse scale is used for the optical PP data). There are hints of an apparent anticorrelation between the two. For instance, at the onset of the γ-ray outburst (JD' ∼ 1610), the optical PP is quite high (∼25%). Also, the peaks of the γ-ray flares (4 and 5) roughly coincide with the dips in the PP curve. Finally, the optical PP regained its original value (∼25%) at the end of the outburst (JD' ∼ 1820). The anticorrelation is even more evident in the optical PP–versus–γ-ray photon flux plot shown in the right panel of Figure 16. The statistical significance is tested via the linear Pearson correlation analysis, which suggests a significant (r_P = −0.27 at a confidence level = 98.5%) anticorrelation between the two.

Zoom In Zoom Out Reset image size

Figure 17 shows the PP–versus–spectral index plot. We again used the linear Pearson correlation analysis to test the significance of the correlation between PP and optical/γ-ray spectral indices. Formally, we obtained r_P = 0.63 at the 99.9998% confidence level for the PP–versus–γ-ray photon index data (see top panel), which suggests a significant correlation between the two. The PP-versus-optical spectral index plot suggests a correlation (r_P = 0.57) at the 93.7% confidence level. A positive correlation between PP and optical/γ-ray spectral indices indicates that higher PP is observed for a steeper spectrum. A significant hardening of the optical/γ-ray spectrum is observed during bright flares (see Sections 3.1.1 and 3.1.3 for details).

Zoom In Zoom Out Reset image size

A change in PP could be simply explained by a change in either the B-field order (i.e., an increase in PP indicates a more ordered magnetic field) or the spectral index of the emitting particles. We could expect to see some increase in PP due to steepening of the electron spectrum or synchrotron spectrum (degree of maximum polarization = (α + 1)/(α + 5/3), where α is the spectral index). For instance, steepening of a spectrum by one magnitude (Δα = 1) produces a ≤5% increment in PP magnitude. Although the PP significantly correlates with α, our observations suggest variations of 20% (ΔPP = 20). This, in turn, suggests that the PP variations cannot be solely explained via steepening of the spectrum, as a change in the spectral index has only a very slight effect on the PP.

Since the mean polarization is the maximum divided by the square root of the number of cells²⁷ involved in the emission, the observed anticorrelation between γ-ray flux and optical PP could be because of the presence of more turbulent cells during the outburst than at other times. The jet plasma is highly dynamic; plasma instabilities (e.g., the current-driven kink instability and/or magnetohydrodynamical instabilities; Nalewajko & Begelman 2012; Nishikawa et al. 2016) could occur and vary with time, which basically drives the turbulence. In short, instabilities drive turbulence, which can influence the number of cells in a given emission region. This, in turn, implies that the turbulence could enhance the efficiency of particle acceleration, which is why there is an outburst and also why the spectrum is flatter during the outburst and steeper before and after the outburst when the PP is higher. An observational test of the turbulence cell model (TEMZ; Marscher 2014) is to compare the standard deviation of random PP fluctuations (compared to the systematic variations). In case of the presence of only turbulent component(s), $\sigma (\mathrm{PP})=0.5\langle \mathrm{PP}\rangle$. We used an equal time bin when the PP is not changing systematically to determine σ(PP). For 3C 279, the calculated σ(PP) = (0.08–0.25) $\langle \mathrm{PP}\rangle$, which suggests the presence of a polarized steady component as well or the presence of multiple emission components with different polarizations.

To further test the hypothesis of the presence of a steady polarized component in addition to the multiple turbulent components, we investigate the variations in Stokes Q and U parameters. Photometric and polarimetric measurements obtained on the same day were used to calculate the fractional Stokes parameters, Q_frac = PP cos(2 × EVPA) and U_frac =PP sin(2 × EVPA). Figure 18 (left) shows the variations in optical R-band Q_frac and U_frac. A clear difference in the Stokes Q_frac and U_frac variations can be noticed here. The observed dependence between the Stokes parameters is given in the right panel of Figure 18 (red: optical R band; magenta: 43 GHz radio core; blue: 230 GHz; green: 86 GHz). The Q_frac, U_frac centroid clearly deviates from (0, 0). We noticed a similar shift at the 86 and 230 GHz radio bands and in the 43 GHz radio core. The observed shift of the Q, U centroid with respect to (0, 0) can be explained by the presence of a steady component with a polarization direction aligned along the jet axis (Blinov & Hagen-Thorn 2009). Moreover, a similarity of EVPA values at optical and radio frequencies with the VLBI core EVPAs supports the alignment of optical/radio polarization direction along the jet axis (Section 3.4). As suggested by Blinov & Hagen-Thorn (2009), if the steady polarized component has a higher degree of polarization, one can expect to see an anticorrelation between the flux and degree of polarization. The observed PP at the beginning/end of the outburst (see Figures 15 and 16) is ∼25%, which means that the steady polarized component is higher than 25%. A steady polarized component having its direction parallel to the jet axis could be the toroidal magnetic field component of the jet. Alternatively, the presence of multiple emission components with different polarizations could be plausible as well.

Zoom In Zoom Out Reset image size

Given the superposition of multiple modes of flaring activity, it is quite challenging to understand the broadband flaring behavior of the source. The major outburst (A–D) is accompanied by faster flares across the whole electromagnetic spectrum (see Figure 1). Interestingly, a new component, "NC2," is ejected from the core during the period of the first three (1–3) γ-ray flares. The next three flares (4–6) were followed by a continuous decay in the source brightness at the γ-ray, optical, and radio frequencies between periods C and D. Later, the ejection of a new component "NC3" from the core suggests a delay between the high-energy activity and the component's core-separation timing.

Simulations (as described in the appendices) were used to test whether or not the connection between component ejection and γ-ray flaring activity is just by chance. First, we simulated a total of 1000 light curves. Next, we checked how often we see a γ-ray flare in the simulated data during the ejection of NC2 (JD' = 1611–1678) and NC3 (JD' = 1790–1836). Given the superposition of multiple modes of flaring activity, defining a flare could be tricky. We choose a simplistic approach here. As the typical duration of the rapid γ-ray flares (1–6) is 10–30 days, we define a continuous rise in the γ-ray flux above 6.6 × 10⁻⁷ photons cm⁻² s⁻¹ (two times above the quiescent level) for ≥20 days as a flare and tested how often this condition is met in our simulated light curves. During the NC2 ejection period, we found a flare in 28 out of 1000 cases; in short, the ejection of NC2 is connected to the γ-ray flaring activity with a 97.2% confidence level. Similarly, we found that the γ-ray flares are related to NC3 with a 95.8% confidence level. We also tested the significance of connection between γ-ray flares observed within 50 days before and after the ejection of the components. Our calculations suggest that the chance coincidence of NC2 being associated with a flare(s) 50 days before or after its ejection is below 90%. In the case of NC3, there are fair chances (94.2%) that its ejection could be related to a flare(s) observed 50 days prior to the ejection, while the confidence level of its ejection being connected to a flare observed with 50 days after its ejection is below 90%. Moreover, we tested the chance coincidence for the different flare durations. For a short flare duration of 10 days, the probability of observing a flare by chance during the ejection of NC2 (and NC3) is less than 10%; i.e., the confidence level of NC2 (and NC3) being associated with a γ-ray flare is ≃90%. In the case of a long-duration flare (30 days), the confidence level is >96%.

During 2013–2014, we observed two modes of flaring activity (faster flares sitting on top of the long-term outburst) in 3C 279 at radio, optical and γ-ray energies. Although X-ray counterparts were observed for two flares, the long-term outburst seems to be missing. The broadband spectral energy distribution modeling during this period (reported in Hayashida et al. 2015) suggests that single-zone emission models that matched the optical and γ-ray spectra failed to describe the X-ray emission, which suggests the presence of multiple emission regions.

Unlike the last three flares (4–6), the first three γ-ray flares (1–3) are missing their optical/radio counterparts, except for flare 3, which is observed at optical bands. Moreover, the IOU spectrum is relatively softer (α_IOU ∼ 1.8–2.0) during the first episode of flares. Later, the IOU spectrum gets harder (α_IOU ∼1.6). A new component, NC2, appeared in the jet during flares 1–3. Ejection of NC2 during the first three flares indicates that these high-energy flares have to be produced either at the VLBI core or very close to the core region; otherwise, we would have observed a delay between the high-energy flares and the component ejection, which we did not. A delay is observed between the last three flares (4–6) and the ejection of NC3. This indicates that the high-energy region is located further upstream of the VLBI core (closer to the central black hole). Moreover, the PP–versus–γ-ray/optical flux and spectral index correlation (discussed in Section 3.5) suggests the presence of multiple turbulent cells in a given emission region and that the number of these cells is variable with time.

We interpret the observed broadband flaring activity in the following manner (see Figure 19). A disturbance at the base of the jet could cause a shock wave and propagate down the jet and could be related to the onset of the major outburst (Figure 19(a)). Later, the moving shock reaches the 43 GHz core, producing the first three flares (1–3). Given the different B-field properties of cells, the particle acceleration efficiency could be different for different cells. Since what we observe at a given frequency is their collective emission, concurrent broadband flaring activity may or may not be observed. The moving shock disturbs the jet outflow and becomes knot NC2 as it separates from the core (Figure 19(b)). In the second episode of flaring activity (4–5), the emission region seems to be located further upstream of the core (closer to the central black hole; see Figure 19(c)). Later, the disturbance/shock wave passes the core region, resulting in the ejection of NC3 (Figure 19(d)). Our analysis therefore supports two possible energy dissipation sites in 3C 279. The location of the dissipation zone depends on the parameters of a disturbance, especially the Lorentz factor (Γ); for disturbances with lower Γ, the dissipation zone is expected to be closer to the black hole than for those with larger Γ (Katarzyński & Ghisellini 2007). Although the values of β_app of NC2 and NC3 are similar within the uncertainties, it is possible that Γ_NC2 > Γ_NC3 that would result in the dissipation zone during flares 1–3 being farther from the black hole and closer to the VLBI core. Both observations (Jorstad et al. 2005; Rani et al. 2015; Hodgson et al. 2017) and simulations (Gómez et al. 1997; Fromm et al. 2012) support the presence of standing shocks in parsec-scale AGN jets. In fact, at short mm-radio bands, the VLBI core could be a standing shock as well (Marscher 2008; Hodgson et al. 2017). Passage of a moving shock through standing shocks is quite efficient in accelerating particles to high energies, which could explain why some sites are more favored than others for energy dissipation (Gómez et al. 1997; Meyer et al. 2015).

Zoom In Zoom Out Reset image size