SIMULTANEOUS OBSERVATIONS OF PKS 2155−304 WITH HESS, FERMI, RXTE, AND ATOM: SPECTRAL ENERGY DISTRIBUTIONS AND VARIABILITY IN A LOW STATE

The underlying particle distributions of blazars are usually studied by matching broadband observations with predictions from radiative models. Since these sources are highly variable, simultaneous observations are essential. The most energetic BL Lac spectra extend up to TeV energies, and positive detections have usually indicated flaring states. However, with their improved sensitivity, the new generation of Atmospheric Cherenkov Telescopes (ACTs), which has more than quadrupled⁷⁹ the number of known extragalactic very high energy (VHE) sources, finds a few of these sources in marginally variable states with consistent detections after short exposures. One of these objects, the blazar PKS 2155−304 at z = 0.116, is an ideal target for such studies. Crucial information is expected from the Fermi Gamma-ray Space Telescope, since its improved sensitivity over EGRET would constrain dramatically the existing models that predict a wide variety of fluxes in the 100 MeV–10 TeV energy range. Since the HESS experiment detects this source in a low state within ∼1 hr, significant daily detections were guaranteed and the source was targeted for an 11-day multiwavelength campaign.

The HESS observations of PKS 2155−304 took place during MJD 54701−54715, for a total of 42.2 hr. After applying the standard HESS data-quality selection criteria, an exposure of 32.9 hr live time remains (MJD 54704−54715), at a mean zenith angle of 183. The data set has been calibrated using the standard HESS calibration method (Aharonian et al. 2004). The analysis tools and the event-selection criteria used for the VHE analysis are presented in F. Aharonian et al. (2009, in preparation). The events have been selected using "loose cuts," preferred for their lower energy threshold of 200 GeV and higher γ-ray acceptance. A 02 radius circular region centered on PKS 2155−304 was defined to collect the on-source events. The background was estimated using the "Reflected Region" method (Aharonian et al. 2006b). Those observations yield an excess of 8800 events, a signal with a significance of 55.7σ calculated following Li & Ma (1983). Using standard cuts an excess of 3612 events with a significance of 68.7σ is found. An independent analysis and calibration (Benbow 2005) yields similar results.

The data from the Large Area Telescope (LAT; Atwood et al. 2008) have been analyzed by using ScienceTools version 9.7, which will be publicly available from the HEASARC in the future. Events having the highest probability of being photons (class 3, called "diffuse") and coming from zenith angles <105° (to avoid Earth's albedo) were selected. The diffuse emission along the plane of the Milky Way, mainly due to cosmic-ray interactions with the Galactic interstellar matter, has been modeled using the 54_59Xvarh7S model prepared with the GALPROP code (Strong et al. 2004a, 2004b) which has been refined with Fermi–LAT data taken during the first three months of operation. The extragalactic diffuse emission and the residual instrumental background have been modeled as an isotropic power-law component and included in the fit. Photons were extracted from a region with 10° radius centered on the coordinates of PKS 2155−304 and analyzed with an unbinned maximum likelihood technique (Cash 1979; Mattox et al. 1996) using the Likelihood analysis software provided by the LAT team. Because of calibration uncertainties at low energies, data in the 0.2–300 GeV energy band were selected.

A total of 75 ks of exposure was taken with RXTE, spread over 10 days coinciding with the HESS observations, and a 6.4 ks exposure with Swift was made toward the end of the campaign. The data taken with the Proportional Counter Array (PCA; Jahoda et al. 1996) and the X-ray Telescope (XRT; Burrows et al. 2005) instruments were analyzed using the HEASOFT 6.5.1 package using the Guest Observer Facility recommended criteria. The XRT data were extracted from a 56'' slice, both for the source and the background. Since the rate was less than 10 Hz, no pile-up is expected in the Windowed Timing (WT) mode.

During the multiwavelength campaign, a total of 106 observations were taken with the 0.8 m ATOM optical telescope (Hauser et al. 2004) located on the HESS site. Integration times between 60 s and 200 s in the Bessel BVR filter bands were used. Photometric accuracy is typically between 0.01 and 0.02 mag for BVR.

2.1. Spectral Analyses

2.2. Light Curves

The light curves from HESS, Fermi, RXTE, and ATOM are shown in Figure 1, where the HESS runs (∼28 minutes) were combined to derive nightly flux values. The average integrated flux above 200 GeV (5.56 ± 0.13_stat ± 1.11_sys) × 10⁻¹¹ ph cm⁻² s⁻¹, corresponds to ∼20% F_{Crab>200 GeV}, or ∼50% higher than the quiescent state of 2003 (Aharonian et al. 2006a) and 70 times lower than its peak flaring flux (Aharonian et al. 2007). The positive excess variance σ²_XS, indicating variability, allows a fractional rms of F_var,VHE = 23% ± 3% (see Vaughan et al. 2003 for definitions of σ²_XS and F_var) to be derived, which is three times less than the high state flaring variability reported by Aharonian et al. (2007). A spectrum was obtained for each night when possible, otherwise two or three nights were combined. No indication of spectral variability was found during those observations, with a limit on the nightly index variations of ΔΓ < 0.2.

Zoom In Zoom Out Reset image size

The Fermi light curve shows the photon fluxes for the high energy (HE) range, 0.2–300 GeV, and the photon spectral indices for each interval. Each bin is the result of a power-law fit, using the background values found on the overall time-averaged fit, and centered on the HESS observations. The light curve fit to a constant has a χ² probability of p(χ²) = 0.95, clearly consistent with a constant flux. The normalized excess variance of −0.16 ± 0.09 sets a 90% confidence level limit of F_var,HE ⩽ 20% on the fractional variance (Feldman & Cousins 1998).

The X-ray light curve, derived from spectral fits of the nightly RXTE (and Swift) data sets, shows flux doubling episodes on timescales of days, similar to the optical and VHE measurements. The lowest fluxes of ∼3–6 × 10⁻¹¹ erg cm⁻² s⁻¹ are at the same level as those seen in the low state (Aharonian et al. 2005b) but with larger fluctuations, F_var,X = 35% ± 0.05%. The time history of the fitted spectral indices in Figure 1 show clearly that the X-ray spectrum hardens significantly, ΔΓ_x ≈ 0.5, as the 2–10 keV flux increases.

The ATOM fluxes are ∼5 times higher than the low state found in Aharonian et al. (2005b), but the V-band magnitudes reported here are in the range 12.7–13 which is well on the lower side of the measurements of PKS 2155−304 reported by Foschini et al. (2008) when the source was quoted to be in a low state with V-band magnitudes in the range 12–12.7. The host galaxy flux is estimated to be ≈10⁻¹¹ erg cm^-2 s^-1 (Kotilainen et al. 1998), hence most of the optical flux can be attributed to the central AGN. The average fractional rms over all bands is F_var,opt ∼ 8% ± 0.5%. The B − R light curve is compatible with a constant, p(χ²) = 0.66, indicating little or no optical spectral variability.

The two-component broadband spectra of high energy-peaked BL Lac objects (HBLs) are typically modeled with synchrotron self-Compton (SSC) scenarios (e.g., Band & Grindlay 1985). Despite the simplicity of these models, they have been successful in reproducing many blazar spectral energy distributions (SEDs) and make definite predictions for the flux and spectral variability that should be seen in the two components. In particular, for typical parameters, the electrons responsible for the X-ray emission also produce the VHE emission; and if the underlying particle distributions were to vary, the resulting flux and spectral changes in the VHE band should be related to variations in the X-rays. In fact, for the 2006 July flare, a nonlinear relationship was seen between the X-ray and VHE bands, though the observed variability patterns do not quite fit the simple SSC model in detail (Costamante 2008).

In Figure 2, we overlay a model SSC calculation that roughly fits the time-averaged SED. The electron distribution model parameters, a three-component power law with indices p₀ = 1.3, p₁ = 3.2, p₂ = 4.3 ($dn/d\gamma \propto \gamma ^{-p_i}$), minimal and maximal Lorentz factors γ_min = 1 and γ_max = 10^6.5, break electron Lorentz factors γ₁ = 1.4 × 10⁴, γ₂ = 2.3 × 10⁵, and total electron number N_tot = 6.8 × 10⁵¹, have been set to reproduce the shape of the lower energy component of the SED. The overall SED is then adjusted with the remaining parameters: radius of the emitting region in the comoving frame, R = 1.5 × 10¹⁷ cm; bulk Doppler factor, δ = 32; magnetic field, B = 0.018 G. Even though we regard this fit as a "straw-man" model, it is perhaps reassuring that the joint Fermi–HESS time-averaged spectra can be reasonably well described as SSC emission. Katarzyński et al. (2008) found similar values for R, B, and δ in their SSC description of a steady large jet component in the SED of PKS 2155−304.

Zoom In Zoom Out Reset image size

Some features of this model calculation are particularly noteworthy. The electrons that produce the synchrotron X-ray emission have Lorentz factors >γ₂. When the power-law component for those electrons is omitted from the calculation, the dot-dashed curve in Figure 2 results. For this particular set of parameters, the electrons that produce the X-rays have higher energies than the electrons that produce the VHE emission. Furthermore, the lack of a significant impact on the shape of the SSC component when those electrons are removed indicates that Klein–Nishina effects suppress any significant contribution by those electrons to the emission at ∼TeV energies.

These features of this calculation allow that there need not be a correlation between the X-ray and VHE fluxes; and in fact, this is what is observed. In contrast with the 2006 July flare, we do not find any evidence of flux correlation between the X-ray and HESS bands with a Pearson's r of 0.12 ± 0.1 between these bands. Furthermore, the 2–10 keV X-ray spectra show spectral variability consistent with an underlying electron distribution for which the cooling timescales are of order the flux variability timescales, i.e., the spectra are softer when the flux is lower, with changes in photon index of ΔΓ_x ≈ 0.5 (Figure 1); whereas, the VHE emission shows no evidence for significant spectral variability despite flux variations of a factor of 2. Since radiative cooling timescales vary inversely with electron energy, this supports the conclusion that the electrons responsible for the synchrotron emission in the X-ray band have higher energies than the electrons that produce the inverse-Compton emission in the VHE range, assuming they are part of the same overall nonthermal distribution.

Even though this all fits in with our straw-man SED calculation, the variability patterns in the optical, X-ray, HE, and VHE bands suggest a much more complex situation. In the absence of spectral variability, the mechanisms that would produce the observed flux variability in the VHE band are rather constrained. Increases in flux could be driven by injection of particles with a constant spectral shape, and decreases in flux could be caused by particle escape from the emitting region or by expansion (adiabatic) losses, assuming those latter two processes can operate independent of particle energy. However, since the electrons that produce the VHE emission must be in the weak radiative cooling regime, a more natural mechanism for the flux variability would be that changes in the seed photon density are driving the variability. Comparing the daily flux values in the optical and the VHE bands, we find indications of fairly strong correlations that suggest that the optical emission provides the target photons for the IC emission. In the B, V, and R bands, the correlations with the HESS fluxes have Pearson's r values in the range 0.77–0.86 with uncertainties ⩽0.09. This correlated behavior is readily apparent in the light curves shown in Figure 1, and these results provide the first quantitative evidence of correlated variability between the optical and VHE bands on these timescales for an HBL.⁸⁰ Confirmation of this behavior, not only from this source but also from other VHE emitting blazars in a low state, would provide important constraints on emission models for these objects.

In the context of a single-zone SSC model, we would expect that any flux variability in the optical bands should also appear as variability in the Fermi–LAT energy range. To illustrate this, we plot, as the dashed curve in Figure 2, the SED that results if we omit contributions from electrons with energies >γ₁. For the original model parameters, the electrons that produce the optical-soft X-ray emission also produce the bulk of the IC component, including the HE and VHE emission. Since we do not find any indication of a correlation between the optical and HE fluxes, this suggests that the optical emission may arise from a separate population of electrons than those responsible for the HE and VHE emission. If so, then these electrons probably also occupy a distinct physical region with different physical parameters (magnetic field, size scale, bulk Lorentz factor). Multizone SSC models of this kind have already been proposed to account for the "orphan" γ-ray flare in 1ES 1959+650 during 2002 May (Krawczynski et al. 2004).

Although the 0.2–300 GeV photon fluxes measured by Fermi are consistent with being constant, we find more significant variations of the photon spectral index in the daily analyses (p(χ²) = 0.19). The fitted values range from fairly soft, Γ = 2.7 ± 0.7, to extremely hard, Γ = 1.1 ± 0.4. These values, along with the constant, intrinsic VHE index of Γ_VHE ≈ 2.5 derived from the HESS data, imply spectral breaks between the HE and VHE bands of ΔΓ as large as 1.4. Very sharp spectral breaks (ΔΓ ≳ 1) would require rather narrow electron distributions and would therefore pose difficulties in fitting a broad lower energy component in the context of a single-zone model. Interestingly, we find a significant anticorrelation between the nightly X-ray fluxes and the Fermi–LAT spectral indices of r_XΓ = −0.80 ± 0.15. A fit to a linear model is preferred over a constant at the 2.6σ level, with a slope of −0.14 ± 0.05. If the electrons that produce the X-rays are at higher energies than those that produce the TeV emission, the cause for such a correlation would be difficult to understand. An important caveat in considering these results is that the Fermi coverage for PKS 2155−304 was relatively uniform over each 24 hr period, whereas the optical, X-ray, and VHE observations were restricted to 4–6 hr intervals each night. Hence, the Fermi observations are not strictly simultaneous with the other measurements, so it is possible that some of the observed HE spectral variability occurred outside of the nightly observing windows.

As the first multiwavelength campaign of an HBL that includes Fermi and an ACT instrument, these observations have yielded results that strongly challenge the standard models for these sources. Having caught PKS 2155−304 in a low state, we see that its spectral and variability properties are significantly different than its flaring, high state behavior. The variability patterns, in particular, defy easy explanation by the usual SSC models and should provide valuable constraints for models that attempt to describe the emission mechanisms in blazar jets.

The full HESS and Fermi–LAT acknowledgements can be found on Web sites http://www.mpi-hd.mpg.de/hfm/HESS/acknowledgements and http://www-glast.stanford.edu/acknowledgements.