EARLY FERMI GAMMA-RAY SPACE TELESCOPE OBSERVATIONS OF THE QUASAR 3C 454.3

The LAT (Atwood et al. 2009) simultaneously monitors ∼2.4 steradians on the sky, albeit with an effective area that varies significantly with the arrival direction of the incident photon with respect to the instrument pointing direction. In survey mode, the instrument z-axis is offset 35° North and South from the spacecraft zenith during alternate orbits in order to provide complete sky coverage every 3 hr. The LAT instrument was turned on 2008 June 25, and the data presented here were collected during the early check-out phase (which ended on August 11) until October 6. Most data were taken in survey mode, the source being observed at an average off-axis angle of 37°. The effective area decreases to about 80% and 70% of the on-axis value at 30° and 40°, respectively. About 7000 photons with E >100 MeV ascribed to 3C 454.3 were detected during the continuous survey mode extending from August 4 to October 6. The exposure at 1 GeV for this period was ≃6.5 × 10⁵ m² s.

For the period July 7–July 31, the instrument was alternately pointed at two sky regions centered on Vela and 3EG J1835+5918. In the later case, 3C 454.3 was observed with an average off-axis angle of 55°, i.e., close to the edge of the field of view, with a corresponding reduction of the effective area of about 50% with respect to the on-axis value.

The light curve obtained with the LAT is shown in Figure 1. We used the standard LAT analysis software, ScienceTools v9r7,⁵⁶ and performed a maximum likelihood fit of the model parameters. The source model included a point source for 3C 454.3, a component for the Galactic diffuse emission derived using the GALPROP code (Strong et al. 2004a, 2004b), and an isotropic component to represent, in combination, the extragalactic diffuse emission and the residual instrument background. The spectral shape of the isotropic component was derived from high Galactic latitude (|b|>60°) sky survey data accumulated over a similar time period as the data for 3C 454.3. The events were selected using the most stringent set of standard analysis cuts and correspond to the so-called diffuse class events, which comprise the highest quality photon events in terms of direction and energy reconstruction and background rejection. The events were extracted in the range 100 MeV–300 GeV and within a 15° acceptance cone centered on the location of 3C 454.3. This region is substantially larger than the 68% containment angle of the point-spread function (PSF) at the lowest energies (∼3°). This is necessary in order to constrain the diffuse emission components accurately. Since the likelihood calculation models the spatial distribution of events as a function of energy, this procedure naturally accounts for the narrower PSF at higher energies when fitting the point source spectral parameters.

Zoom In Zoom Out Reset image size

We note that 3C 454.3 was sufficiently bright that the background contribution within a few degrees of the source is only a small fraction of the source count rate and that the observed flux variations of the source do not correlate with the changing background level along the satellite orbit. Our current estimate of systematics is <20% on the flux.

Figure 1 shows that the source is consistently variable, with the rise and decay times of ∼3.5 days, corresponding to the doubling of the flux in 2 days. The spectral shape appears constant during this period within the systematic uncertainties. The flux F_E>100 MeV peaks at roughly 1.2 × 10⁻⁵ photons cm⁻² s⁻¹ on MJD 54657 (2008 July 10). Note that such an exceptional flux from any blazar was only observed twice during the EGRET era, for 3C 279 (Wehrle et al. 1998) and PKS 1622-297 (Mattox et al. 1997). In survey mode, the relative change of reconstructed flux due to uncorrected acceptance variations (calibration of the effective area as a function of the off-axis angle) has been determined to be about 5% for a daily light curve using the data from the (steady) bright pulsars. An additional 15% uncertainty has been estimated to affect the data collected in pointing mode presented here, because of the proximity of the source direction to the edge of the field of view.

We have also fitted the broadband spectrum of the combined data from the onset of the regular survey mode, August 3 through September 2. As with the daily light-curve analysis, we have used the standard "diffuse" class event selections and restricted the energy range to 200 MeV–300 GeV. Since 3C 454.3 has consistently been one of the brightest sources since the beginning of the mission, the data accumulated over the month of survey data are sufficient to allow us to consider models that are more complex than a simple power law in trying to fit the source spectrum.

Our basic model does use a simple power law for 3C 454.3, and that model yields a photon index of Γ = 2.40 ± 0.03_stat ± 0.09_sys. This is substantially softer than the index reported by AGILE, but the data were not contemporaneous. The difference could also be partially due to different bandpasses of the two instruments. The residuals to the power-law fit clearly indicate that the spectral model for the γ-ray emission from 3C 454.3 must be more complex; and in particular, the data show a steepening toward higher energies. A broken power-law model yields Γ_low = 2.27 ± 0.03_stat ± 0.09_sys, Γ_high = 3.5 ± 0.2_stat ± 0.15_sys, and a break energy of 2.4 ± 0.3_stat ± 0.3_sys GeV. The likelihood ratio test gives the probability for incorrectly rejecting the power-law model in favor of the broken power-law model as 5 × 10⁻¹². The break was consistently detected at similar energies when repeating the analysis for different one-month periods (August, September, and October). An instrumental effect can safely be ruled out, given the level of the systematic uncertainties. An unfolded νF_ν spectrum corresponding to this model is shown in Figure 2. We note that with the current data, we cannot distinguish between various models describing the steepening spectrum, e.g., broken power law versus exponentially cutoff power law, but such differentiation might be possible with more data. Nonetheless, this is the first indication of an observed break in the γ-ray spectrum of a high-luminosity blazar, calling for a complex spectral model beyond a simple power-law approximation. With this model, we determine the time-averaged flux for the 2008 August 3–September 4 epoch to be F_E>100 MeV = 3.0 ± 0.1 × 10⁻⁶ photons cm⁻² s⁻¹.

Zoom In Zoom Out Reset image size

Triggered by the detection of a high-flux level during the first intensive Fermi–LAT observations of 3C 454.3 in 2008 July/August, the broadband behavior of the source was monitored via an ad-hoc multiwavelength (MW) campaign covering the period 2008 July to 2009 January. These intensive observations involved ground-based monitoring at radio centimeter/millimeter/submillimeter wavelengths (e.g., Effelsberg 100 m, IRAM 30 m, SMA, and OVRO 40 m telescopes) and at IR/optical bands (e.g., Spitzer, Kanata/Hiroshima 1.5 m and REM telescopes, plus telescopes of the GASP-WEBT consortium). In addition, daily 2 ks observations were performed with the Swift satellite, providing X-ray data, as well as optical and UV coverage. Details of this MW campaign, including all collected data and the corresponding MW analysis of the source behavior between 2008 June and 2009 January at radio/IR/optical/UV/X-ray/γ-ray bands, will be presented in a subsequent paper (A. A. Abdo, et al. 2009, in preparation). For the purpose of the first LAT results presented here, we will use a subset of these MW data to construct a quasi-simultaneous spectral energy distribution (SED) of 3C 454.3 including radio, optical, UV, X-ray, and LAT γ-ray data. In Figure 3, we show the corresponding data collected during the period MJD 54685–54690 (2008 August 7–12).

Zoom In Zoom Out Reset image size

The radio centimeter/millimeter band data were obtained with the Effelsberg 100 m and IRAM 30 m telescopes in the framework of a Fermi-related monitoring program of potential γ-ray blazars (F-GAMMA project; Fuhrmann et al. 2007). At Effelsberg, observations were conducted quasi-simultaneously at eight frequencies between 2.6 and 42 GHz using cross-scans in azimuth and elevation direction. Consequently, pointing off-set correction, gain correction, atmospheric opacity correction, and sensitivity correction have been applied to the data (for details see Fuhrmann et al. 2008; Angelakis et al. 2008). The observations at the IRAM 30 m telescope were obtained with superconductor–insulator–superconductor (SIS) receivers (operating at 86, 142, and 228 GHz) and with calibrated cross-scans in azimuth and elevation direction. The receiver calibration was done using hot and cold loads (standard chopper wheel method) and the opacity corrected intensities were converted into the standard temperature scale. After Gauss fitting of the averaged subscans, each temperature was corrected for remaining pointing offsets and systematic gain-elevation effects. The conversion to the standard flux density scale was done using the instantaneous conversion factors derived from primary calibrator measurements (for details, see Ungerechts et al. 1998).

The optical, UV, and soft X-ray data were obtained from 2 ks pointings with the Swift satellite as a part of daily Target of Opportunity (TOO) observations during the campaign. For the screening, reduction and analysis of the data from Swift instruments we used standard procedures within the HEASoft version 6.5 software package with the calibration database updated as of 2008 June 26. The X-Ray Telescope (XRT) operated in photon counting mode, and the analysis was performed with the xrtpipeline task with default parameters and having selected photons with grades 0–12. We also checked for the presence of pileup. The Ultraviolet Optical Telescope (UVOT) data were integrated with the uvotimsum task and analyzed with the uvotsource task using source region radii of 5'' for the optical filters and 10'' for the UV filters, respectively. The background was extracted from a nearby source-free circular region with 50'' radius. The observed magnitudes were converted into flux densities according to standard procedures (Poole et al. 2008).

The amplitude of blazar variability can be as high as a factor of 100, so many important inferences with regard to the source structure can be derived most robustly from broadband spectra obtained simultaneously, with additional, crucial constraints from time-resolved broadband spectroscopy. To this end, we assembled the data as given above, contemporaneous with our 2008 August epoch. We plot a quasi-simultaneous SED using our MW data from August 7 to 12 in Figure 3. The overall SED of the source is broadly similar to that observed for other EGRET blazars, revealing two broad peaks: one in the millimeter-to-IR band and the other in the γ-ray band. Given the polarization measured in all accessible segments of the low-energy component, the most viable emission mechanism is the synchrotron process, produced by a distribution of relativistic electrons radiating in an ambient magnetic field.

The high-energy peak, in turn, could be due to Compton scattering of either the synchrotron photons internal to the jet (see, e.g., Ghisellini et al. 1985) or the radiation associated with the nucleus, such as the photons produced by the accretion disk, broad emission line region, or the thermal emission produced by the circumnuclear dust in the host galaxy (Dermer et al. 1992; Sikora et al. 1994). Radiation from ultrahigh energy protons or ions that interact with magnetic fields associated with the jet or its environment, including subsequent cascades, may be an alternative explanation for the high-energy component (e.g., Mücke et al. 2003; Atoyan & Dermer 2003; Böttcher et al. 2008).

Inferences regarding the source structure are strongly model dependent, even within leptonic emission models. A comparison of results presented for 3C 454.3 in Vercellone et al. (2009) with the analysis of the broadband temporal behavior of Sikora et al. (2008) clearly indicates that a wide range of parameters can describe the same data. Nonetheless, the different models do find roughly consistent values of the jet Doppler factor, δ ∼ 25, the magnetic field, B ∼ 1 Gauss, and the Lorentz factors of the electrons that radiate the bulk of the observed power, γ_max ∼ 10³. The largest disparity arises in the distance, r, of the dissipation region from the central black hole (i.e., the location of the "blazar emission zone"), and this points to differing origins of the seed photons that are Compton upscattered to the γ-ray range: UV disk photons reprocessed in the broad-line region (BLR) versus IR emission of the obscuring dust. In particular, Vercellone et al. claim r ∼ 3 × 10¹⁶ cm, while Sikora et al. argue for r ∼ 10¹⁹ cm. As we discuss below, our detection of the spectral break can put further constraints on the location of dissipation region, but ultimately, time-resolved broadband spectroscopy will be crucial.

The γ-ray flux from 3C 454.3 is clearly variable on timescales ranging from days to decades, with the average flux value measured during the LAT observations being somewhat higher than that found by AGILE and much higher than that measured by EGRET. Assuming a "concordance" cosmology (Ω_Λ = 0.73, Ω_matter = 0.27, and H_o = 71 km s⁻¹ Mpc⁻¹), we infer a luminosity distance of d_L = 5.5 Gpc and an apparent isotropic monochromatic luminosity of $L_{E_0} \approx 4 \pi d_L^2\, (\Gamma _{\rm low} - 1) \,E_0 \, F_{E> E_0} \approx 2 \times 10^{48}$ erg s⁻¹, where E₀ = 100 MeV and we have used the average 2008 August 3–September 4 flux of $F_{E_0} = 3.0 \times 10^{-6}$ ph cm⁻² s⁻¹. The bolometric luminosity is expected to be much higher than this since the high-energy spectral component peaks below 100 MeV. This finding, together with the observed rapid, large amplitude variability measured in the γ-ray band, puts strong constraints on the source parameters. If a stationary source of a given luminosity in γ-rays and X-rays were to be as compact as one would infer from the variability data, the source would be optically thick to e⁺/e⁻ pair production, assuming that the X-ray emission is cospatial with the γ-ray emission. This assumption is justified by the simultaneity of the flaring activity in both bands as observed during our MW campaign, and in particular, as seen from our full Swift XRT soft X-ray TOO observations (2008 July–September; A. A. Abdo, et al. 2009, in preparation). Similar simultaneous X-ray/γ-ray variability has been measured in the past for other sources, such as 3C 279 (Wehrle et al. 1998).

One obvious solution to the problem of excess pair-production opacity is to invoke relativistic motion and/or expansion of the source. Following the arguments given in Mattox et al. (1993, with corrections pointed out in Madejski et al. 1996), and adopting the doubling timescale of t_d ≈ 2 days and the observed X-ray flux of S_X ≈ 3 × 10⁻¹¹ erg cm⁻² s⁻¹ (as measured during our campaign) at the observed photon frequency ν_X ≈ 10¹⁸ Hz (corresponding to the photons that annihilate with the GeV γ-rays in the jet rest frame), we estimate the Doppler factor δ required for the photon–photon annihilation optical depth to be τ_γγ ⩽ 1. With the derived relation τ_γγ ∼ σ_T d²_L S_X/3 t_d c² E_X δ⁴, where we put the emission region linear size R = c t_d δ/(1 + z), the source-frame photon energy E'_X = (1 + z) hν_X/δ, and the intrinsic X-ray luminosity L'_X = 4πd²_L δ⁻⁴ S_X, one obtains δ ≳ 8 (note here that the estimated Doppler factor scales as δ ∝ t^−1/4_d). Of course, omitting the requirement of cospatiality of the X-ray and γ-ray emission regions relaxes this limit. Nonetheless, it is interesting to compare this constraint on the Doppler factor with the estimate obtained from the VLBI superluminal motion, δ = 24.6 ± 4.5 (Jorstad et al. 2005). As long as the velocity of the VLBI jet is the same as the velocity of the outflow within the blazar emission zone, this implies that the photon–photon annihilation effects involving the X-ray emission generated within the jet are negligible.

The LAT data imply a more complex spectrum than a simple power law, indicating significant softening of the photon spectral index by ΔΓ ∼ 1.2 toward higher energies. This is not consistent with a spectral change of ΔΓ = 0.5 that is expected from the typical "cooling break" associated with radiative losses. The observed softening may instead be due to an intrinsic decline or break in the particle distribution. It can also arise, at least partially, from "environmental" reasons, i.e., the underlying photon spectrum is modified by the photon–photon pair production. For 3C 454.3, this is unlikely to be due to the extragalactic background light (EBL) which would only affect the spectra from a source at z ⩽ 1 at energies ⩾40 GeV. Also, as argued above, the pair production effects internal to the jet are negligible, since the expected jet Doppler factor (δ ∼ 25) is much larger than that required for internal γ-ray transparency (δ ≳ 8). A photon field that is external to the jet but local to the blazar central engine would need to peak at ∼0.2 keV in order to account for the observed spectral softening at ∼2 GeV. This requirement would exclude photons produced in the BLR as the source of pair opacity since that emission peaks in the UV. However, the required X-ray photons could be produced in the innermost part of the putative accretion disk surrounding the central black hole and/or a hot corona above the disk.

We do not know the shape of the intrinsic, unabsorbed γ-ray spectrum, but the simplest assumption is that it is a straight power law, as is believed to be produced via first-order Fermi acceleration at shocks. Departures from such a spectral shape may involve both softening and hardening toward higher energies over the canonical power-law component, arising due to radiative particle energy losses and/or efficient stochastic acceleration acting near the shock front (see, e.g, Stawarz & Petrosian 2008). For a simple power law, we can estimate the opacity that the emerging γ-rays would encounter as a function of distance from the black hole, given the observed spectral break at 2 GeV (e.g., Reimer 2007). Assuming crudely that the break corresponds to the location where the pair-production opacity τ is of order unity, we can estimate the minimum distance at which the γ-rays are produced. We consider the soft X-ray target photons at ∼0.2 keV as being produced either in the inner parts of an accretion disk (assumed here to follow the behavior of a Shakura–Sunyaev disk spectrum) or in an X-ray-emitting hot corona above the disk. For a black hole mass estimate for 3C 454.3 of 4 × 10⁹M_☉ (Gu et al. 2001) and a mass accretion rate of ∼10% the Eddington rate, the bolometric (Shakura–Sunyaev) disk luminosity would be L_bol ∼ 2 × 10⁴⁶ erg s⁻¹. This is in good agreement with reports from Raiteri et al. (2007) of the quiet state UV-spectrum of 3C 454.3 that shows a rise in the SED, suggesting the onset of the "big blue bump" thermal accretion disk spectrum. It is also in good agreement with the luminosity estimate of the BLR from Pian et al. (2006) of order L_BLR ∼ 3 × 10⁴⁵ erg s⁻¹, implying a typical covering factor of ∼10%. The interaction of jet GeV photons with soft X-ray photons originating in the inner parts of an accretion disk implies photon–photon interactions at preferentially small angles. The angular dependence of the threshold target photon energy, E_thr = (2m_ec²)²/(2E_γ(1 − cos θ)), where θ is the interaction angle, requires correspondingly higher densities of energetic target photons than are expected to be emitted by a standard Shakura–Sunyaev disk around a 4 × 10⁹M_☉ black hole for which the thermal emission is weak at 0.2 keV and above.

On the other hand, quasars often show soft X-ray emission extending up to ∼100 keV that is believed to arise from Comptonization by a hot coronal plasma above the disk. This mechanism creates a characteristic flat continuum that can typically be modeled as an E⁻² photon spectrum with an exponential cutoff at ∼100 keV. Rapid variability of this flux indicates that it is emitted within the regions closest to the black hole. Adopting the typical values of the ratio L_X/L_bol of 1%–10% (Laor et al. 1997), we find that the condition corresponding to a unity optical depth for 2 GeV photons occurs at r ≲ 10⁻³–10⁻² pc, corresponding to ∼10–100 gravitational radii for M_BH = 4 × 10⁹M_☉. Thus, if the observed spectral break at 2 GeV were due to photon–photon pair production on the accretion disk corona radiation, the implied location of the γ-ray-emitting region should be extremely close to the black hole. Another scenario can be envisioned in which radiation from the accretion disk corona is subsequently Thomson-scattered toward the jet by the free electrons in the gas in the more distant BLR. However, this would require an unrealistically high product of covering fraction and Thomson optical depth that would not be consistent with the observed profiles of emission lines in AGN unless the electrons were located well within the BLR. Regardless, neither direct nor BLR-scattered accretion disk corona radiation are likely sources of the required pair opacity, because the reprocessed γ-radiation should then escape as soft and strong X-ray emission produced in a pair cascade developed in the intense photon field of the accretion disk/disk corona (Levinson & Blandford 1995; Ghisellini & Madau 1996), which is not consistent with the collected X-ray data in the context of the broadband spectrum (see, e.g., Figure 3).

We thus suggest that the observed 2 GeV break in the γ-ray spectrum of 3C 454.3 is due to an intrinsic break in the spectrum of radiating particles. In the context of models invoking Comptonization of external photons, such a break implies a characteristic Lorentz factor, γ_el, for the electrons radiating at the break. For inverse-Compton scattering of "seed" photons that have energy E_seed,rest in the rest frame of the blazar, the observed γ-ray photon energy is given by

$$\begin{equation} E_{\gamma } = E_{\rm seed, rest}\, \delta \, \Gamma _{\rm jet} \, (1+z)^{-1}\,\gamma _{\rm el}^2. \end{equation} \tag{ 1 }$$

Hence, the Lorentz factor of electrons radiating at E_γ = 2 GeV is

$$\begin{equation} \gamma _{\rm el} \approx 6 \times 10^3 \, (\delta / 10)^{-1/2} \, (\Gamma _{\rm jet}/10)^{-1/2} \, (E_{\rm seed, rest} / {\rm eV})^{-1/2}. \end{equation} \tag{ 2 }$$

The likely sources of seed photons in blazars are either UV photons from the accretion disk that are reprocessed in the BLR and which comprise primarily Lyα photons with E_seed,rest ∼ 10 eV, or IR emission from obscuring dust which have E_seed,rest ∼ 0.3 eV. These seed photon energies yield γ_el ∼ 10³ and ∼6 × 10³, for the cases of UV or IR seed photons, respectively, assuming the kinematic jet parameters δ = 25 and Γ_jet = 15 as inferred from VLBI observations. In principle, for those values of γ_el, the γ-ray radiation is close to the onset of the Klein–Nishina (KN) regime, but as has been argued by several authors (Zdziarski & Krolik 1993; Moderski et al. 2005), such a KN cutoff might be less severe than expected in the simplest scenarios. Regardless, we also comment here that such complex intrinsic γ-ray spectrum, if common, may present an obstacle toward the use of luminous blazars as "white light" sources to study the intergalactic diffuse UV/optical background radiation, first because it is not known whether the break can be reliably modeled, but also because the steepening spectrum reduces the photon statistics at the highest γ-ray energies, where the opacity of the diffuse background light is most pronounced. Finally, it is interesting to speculate about the issue of contribution of γ-ray—luminous blazars to the extragalactic γ-ray background. Specifically, if luminous blazars make a significant contribution to this background and if a spectral break at a few GeV is a common feature in such blazars, then evidence of such breaks should be present in the extragalactic diffuse emission. To settle this issue, sensitive, well-calibrated data extending to the highest energies accessible by Fermi are necessary, but detailed analysis of the diffuse extragalactic background in Fermi data is yet to be performed.