SUZAKU OBSERVATION OF THE BRIGHTEST BROAD-LINE RADIO GALAXY 4C 50.55 (IGR J21247+5058)

Radio galaxies are a class of active galactic nuclei (AGNs) spouting powerful radio jets whose axis is not aligned along the line of sight (LOS). While the formation mechanism of relativistic jets is not fully understood, it must be closely related to mass accretion flow onto the central supermassive black hole. In fact, studies of Galactic black holes have revealed that the relative power of the jets to accretion critically depends on the state of the accretion disk, which is predominantly determined by the mass accretion rate (e.g., see Fender et al. 2004). For AGNs, however, detailed studies of the accretion disk state in relation to the jet formation are still limited. Broadband observation of nearby, bright radio galaxies hence give us ideal opportunities to unveil this problem, because, unlike "blazars," the innermost disk can be well observable in the X-ray band with much smaller contribution from the jet emission.

4C 50.55 (IGR J21247+5058) is a Fanaroff–Riley type II broad-line radio galaxy (BLRG), discovered by International Gamma-Ray Astrophysics Laboratory (INTEGRAL) in its first survey catalog (Bird et al. 2004). The source is also detected in the first nine months data of the Swift Burst Alert Telescope (BAT) survey of AGNs (Tueller et al. 2008). Though being the brightest BLRG in the hard X-ray sky above 10 keV, 4C 50.55, located at (l, b) = (93.32, 0.3937), had been unrecognized as a bright X-ray source until the INTEGRAL and Swift era, due to the obscuration by the Galactic plane. This source is also of great interest for understanding the accretion flow onto supermassive black holes at high fractions of Eddington luminosity, which is estimated to be L_bol/L_Edd ∼ 0.4 (see Section 3.5). Massetti et al. (2004) determined the redshift to be z = 0.020 ± 0.001 based on the detection of a broad Hα emission line in the optical spectrum. From the observed flux densities at 1.4 GHz with the Very Large Array (VLA), Molina et al. (2007) estimate the inclination angle to be θ ∼ 35°, assuming a Lorentz factor of γ = 5 and a typical power ratio between the core and total known for radio galaxies (Giovannini et al. 1988). The inferred viewing geometry of the nucleus is very consistent with the classification of 4C 50.55 as a BLRG.

The initial results on the X-ray spectra are reported by Molina et al. (2007), who used the XMM-Newton and Swift/XRT data combined with a time-averaged INTEGRAL spectrum. They find that the spectra below 10 keV are apparently hard, and complex, multiple layers of absorber are required. Iron-K features are not significantly detected and the strength of reflection components, R ≡ Ω/2π, where Ω is the solid angle of the reflector, is not tightly constrained (R ≲ 0.9). In this paper, we present the first simultaneous broadband X-ray data of 4C 50.55 observed with Suzaku (Mitsuda et al. 2007), obtained in a different epoch of the XMM-Newton observation. Suzaku carries three CCD cameras (then available) called the X-ray Imaging Spectrometers, XIS-0, XIS-3 (Front-side Illuminated XIS; XIS-FI), and XIS-1 (Back-side Illuminated XIS; XIS-BI), and a collimated-type instrument called the hard X-ray detector (HXD), which consists of Si PIN photodiodes and gadolinium silicate (GSO) scintillation counters. The XISs and HXD-PIN covers the 0.2–12 keV and 10–70 keV, respectively. Since AGNs are time variable, the simultaneous coverage is critical for studying the continuum shape over the broad band, in particular to accurately constrain the Compton reflection component. Moreover, the Suzaku exposure is quite deep (a net exposure of ∼100 ks), and thus provides us with the best quality data set so far obtained from this source.

The organization of our paper is as follows. First, we describe our Suzaku observation and data reduction in Section 2. Next, we present the analysis and results in Section 3. The results of the Swift/BAT data are also presented, whose spectrum averaged over 22 months is utilized to constrain the highest energy band up to 200 keV. We plot the spectral energy distribution (SED) of 4C 50.55 from the radio to gamma-ray bands to discuss the possible contribution of the jet component. Finally, we discuss our results in comparison with the previous studies of this target and other radio galaxies in Section 4. In all spectral analysis, we apply the Galactic absorption fixed at N^gal_H=1.0 × 10²² cm⁻² (Kalberla et al. 2005). The cosmological parameters (H₀, Ω_m, Ω_λ) = (71 km s⁻¹ Mpc⁻¹, 0.27, 0.73; Komatsu et al. 2009) are adopted in calculating the luminosities. The errors attached to spectral parameters correspond to those at 90% confidence limits.

4C 50.55 was observed with Suzaku from 2007 April 16 14:05 (UT) to April 18 11:04 (UT) (observation ID 702027010), focused on the nominal center position of the HXD. The net exposures after data screening, described below, are 85.0 ks (XIS-0, 1), 80.0 ks (XIS-3), and 54.7 ks (PIN). For the HXD, we use only data collected with the PIN diodes, because we find that the signal from the source in the GSO data is not significant over the systematic error of ≈2% in the non X-ray background (NXB) (Fukazawa et al. 2008). We use FTOOLS (heasoft version 6.6.2) to extract data, and XSPEC version 11.3.2ag for spectral fitting.

2.1. XIS Data Reduction

2.2. HXD-PIN Data Reduction

3.1. Swift/BAT Data

Figure 1 shows the long-term light curve of 4C 50.55 in the 14–195 keV band obtained with Swift/BAT from 2004 December 15 to 2009 October 10. Each data point corresponds to the averaged flux for 16 days. Time variability is clearly noticed. The epoch of our Suzaku observation is indicated by the dashed line in the figure. The time-averaged BAT spectrum over the first 22 months covering the 14–195 keV band is plotted in Figure 2. We find that it can be fit with a single power law of Γ = 1.68 ± 0.25. The time-averaged 14–195 keV flux is 1.7 × 10⁻¹⁰ erg cm^-2 s⁻¹ (based on the best-fit power-law model), which corresponds to a luminosity of 1.5 × 10⁴⁴ erg s^-1 in the rest-frame 14.3–199 keV band. In the subsequent subsections, we will apply more realistic spectral models to this spectrum.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

3.2. Suzaku Light Curve

We make the Suzaku light curves of 4C 50.55 with a 5760 s bin, the orbital period of the satellite, to remove any possible modulation related to the orbital condition. Figure 3 shows the X-ray light curves obtained with the XIS-FIs (2–10 keV, upper) and with the HXD/PIN (15–40 keV, middle), and their hardness ratio (PIN/XIS, lower). The zero point of time corresponds to the start time of the Suzaku observation. The XIS light curve in the 2–10 keV band indicates a rapid flux increase by ∼20% in the last 20 ks exposure. The timescale of this variability is ∼10⁴ s, indicating that the emission region is within ∼30r_g ($r_{\rm g} \equiv \frac{GM}{c^2}$ is the gravitational radius) for a black hole mass of M = 10^7.8M_☉ (see Section 3.5). By contrast, the light curve in the 15–40 keV band does not show evidence for significant variability above the statistical errors in the same epoch. Thus, there is a hint of a decrease in the hardness ratio between the PIN and XIS, although its significance is marginal. For the following spectral analysis, we separate the observation period into two epochs, epoch 1 (0–1.4 ×10⁵ s) and epoch 2 (1.4 ×10⁵–1.7 ×10⁵ s).

Zoom In Zoom Out Reset image size

3.3. Spectral Analysis with Phenomenological Models

3.3.1. Individual Fit to the Suzaku and Swift Data

We analyze the Suzaku spectra of epoch 1 and epoch 2 separately, by performing simultaneous fits to those of the XIS-FIs in the 1–9 keV band, the XIS-BI in the 1–8 keV band, and the HXD/PIN in the 12–60 keV band. Given the fact that the spectrum of 4C 50.55 is subject to a heavy Galactic absorption, we do not utilize the XIS-BI data below 1 keV to avoid any possible uncertainties in the response. We find that the effects of pileup are significant above 9.0 keV for the XIS data. Further, we discard the XIS data of the 1.7–1.9 keV band because of calibration uncertainties associated with the instrumental Si-K edge. In the simultaneous fit, the flux normalizations for the XIS-FIs and XIS-BI are set free, while that of the PIN is fixed at 1.18 relative to that of the XIS-FIs (Maeda et al. 2008).

We first apply a single power-law model (modified by the Galactic absorption fixed at N^gal_H=1.0 × 10²² cm^-2) to the XIS and PIN data in epoch 1, covering the 1–60 keV band. The fit is found to be far from acceptable (χ²/dof = 13035/810) and results in a very flat slope, Γ ≈ 1.2. This is because the spectral shape below 10 keV is much harder than that above 10 keV. Next we apply a cutoff power-law model, in the form of E^−Γ × exp(−E/E_cut), over which a narrow Gaussian is added to represent an iron-K emission line at 6.4 keV. We obtain Γ ≈ 0.83 and E_cut ≈ 14 keV with χ²/dof = 4616/808. There remain strong absorption features around 1 keV suggesting intrinsic absorption, however. When we add another absorption at the source redshift (z = 0.02), the fit becomes much better, yielding Γ = 1.39 ± 0.02, E_cut = 45 ± 2 keV, and N_H = (0.67 ± 0.01) × 10²² cm^-2 with χ²/dof = 997/807. Finally, similar to the analysis done by Molina et al. (2007), we consider a double absorber model with two different column densities, N¹_H and N²_H, whose covering fraction is f and (1 − f), respectively. The fit is significantly improved (χ²/dof = 940/805) with Γ = 1.61^+0.03_−0/05, E_cut = 140⁺⁴⁶₋₂₀ keV, N¹_H≈8.3 × 10²² cm^-2, N²_H ≈0.73 × 10²² cm^-2, and f = 0.19. The photon index becomes a reasonable value for AGNs. Note that our "double absorber" model has three free parameters for the absorber, while the "double partial covering" model ("pcf*pcf*" in the XSPEC terminology) adopted by Molina et al. (2007) has four. Since the latter model does not give a significant improvement for our data (Δχ² ≈ 1), we adopt the former as a base continuum model for the following analysis.

The high quality Suzaku spectra are quite useful to constrain the reflection component, which is indicated by the presence of the iron-K emission line. Thus, we include it by utilizing our modified version of the "pexriv" reflection code (Magdziarz & Zdziarski 1995) that assumes a cutoff power-law continuum and contains a self-consistent fluorescence iron-K line calculated according to the same algorithm as described in Zycki et al. (1999). The additional free parameter is the relative reflection strength, R(≡Ω/2π), while we fix the ionization parameter, temperature, and inclination angle at 0, 10⁵ K, and 35° (Molina et al. 2007), respectively. The solar Fe abundance by Anders & Grevesse (1989) (Fe/H = 4.68 × 10⁻⁵) is assumed. Considering that the reflection most likely occurs in the accretion disk, we smear both the reflected continuum and iron-K emission line by the "diskline" kernel (Fabian et al. 1989). The innermost radius r_in is set as a free parameter by assuming an emissivity law of r⁻³ for a fixed outer radius r_out = 10⁵r_g. When constraining r_in, we utilize only the XIS spectra around the iron-K band (3–9 keV for the XIS-FIs and 3–8 keV for the XIS-BI) and fix all the other parameters except for the normalization. Thus, the spectral fit is performed by iteration; after r_in is determined from the XIS-only fit, it is then fixed when finally determining the continuum parameters in the XIS+PIN fit.

The results of the spectral fit to the individual Suzaku spectra in epochs 1 and 2 are summarized in Table 1. Figure 4 shows the XIS+PIN spectra folded by the energy responses, over which the best-fit models are plotted, with residuals in the lower panel. The expanded figure of the XIS spectra between 3 and 9 keV in epoch 1 is plotted in Figure 5. To emphasize the iron-K line feature, the residuals when the line is excluded from the model are shown in the lower panel of this figure. From epoch 1, we obtain Γ = 1.61 ±   0.05, E_cut = 80⁺³⁶₋₁₉ keV, and R = 0.18 ±   0.04. The innermost radius is constrained to be r_in = 720 r_g(>340 r_g) from the XIS data. For the analysis of the epoch 2 spectra, we assume the same parameters of the reflection component (including its absolute flux) as those found from the epoch 1 data, since it is very unlikely that it varied on such a short timescale of <10⁴ s. The parameters of the absorption are fixed to the epoch 1 values as well. Thus, only free parameters are Γ, E_cut, and the normalization. Finally, we also perform a spectral fit to the time-averaged Swift/BAT spectrum by adopting the same model. The reflection strength is fixed at R = 0.18 referring to the epoch 1 result. The best-fit model is overplotted in Figure 2, whose parameters are summarized in Table 1.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Table 1. Best-fit Parameters of Cutoff Power-law Model to the Individual Suzaku and Swift Data

Parameters	Epoch 1	Epoch 2	Swift/BAT
af2–10 keV	7.4 × 10−11	8.5 × 10−11	⋅⋅⋅
af10–60 keV	1.6 × 10−10	1.6 × 10−10	⋅⋅⋅
af14–195 keV	⋅⋅⋅	⋅⋅⋅	1.7 × 10−10
bNgalH	1.0e	1.0e	⋅⋅⋅
Γ	1.61 ± 0.05	1.64 ± 0.03	1.68+0.23−0.25
Ecut [keV]	80+36−19	68+29−16	127+340−57
R(=Ω/2π)	0.18 ± 0.04	f	0.18e
cf	0.19+0.03−0.04	0.19e	⋅⋅⋅
dN1H	7.6+1.5−1.4	7.6e	⋅⋅⋅
dN2H	0.73 ± 0.03	0.73e	⋅⋅⋅
EW [eV]	22	19	⋅⋅⋅
rin[rg]	720e (>340)g	720e	720e
χ2/dof	916.0/805	190.8/175	8.594/5

Notes. Errors are 90% confidence level for a single parameter. ^aObserved fluxes in the 2–10 keV, 10–60 keV, and 14–195 keV bands, in units of erg cm^-2 s⁻¹. ^bGalactic absorption column density in units of 10²² cm^-2. ^cCovering fraction. ^dIntrinsic absorption column density at the source redshift in units of 10²² cm^-2. ^eParameters fixed at these values. ^fWe assume the same reflection component as that determined in epoch 1. ^gWe constrain the inner radius only from the 3–9 keV XIS spectra in epoch 1, which is then fixed at the best fit when determining the continuum parameters.

Download table as:  ASCIITypeset image

3.3.2. Simultaneous Fit to the Suzaku and Swift Data

From the above analysis, we find no significant differences in the spectral parameters (except for the normalization) within the statistical errors between the Suzaku epoch 1, epoch 2, and Swift/BAT data, although there is a hint that the spectrum became slightly softer in epoch 2. Thus, to best constrain the continuum parameters, in particular the cutoff energy, we study the Suzaku spectra (either of the two epochs) in the 1–60 keV band and Swift spectrum in the 14–195 keV band simultaneously, in all following analyses. The flux normalization between the Suzaku (XIS-FIs) and BAT spectra are set free to take into account the time variability. In analyzing the spectra of epoch 2, we always fix the reflection component to that determined from the epoch 1 data.

Table 2 summarizes the results using the same phenomenological model (cutoff power-law) as adopted in Section 3.3. For epoch 2, we consider two extreme cases as the cause of the spectral variability from epoch 1 that (1) only the continuum changed without change of the absorber and that (2) only the absorber changed with the same continuum except for its normalization. We obtain similarly good fits for the two cases, and thus both possibilities are plausible from the spectral analysis. In reality, however, it may be difficult to explain such a short time (<10⁴ s) variability by the absorber alone. If the absorber makes Kepler motion at ∼1000 r_g, a typical location of the broad-line region in AGNs, it moves only ∼r_g in 10⁴ s, by assuming the black hole mass of 10^7.8 M_☉. Thus, unless the emitting region is extremely small (like < several r_g), it is unlikely that crossing blobs in the LOS can cause the large variability as observed.

Table 2. Best-fit Parameters of Cutoff Power-law Model to the Combined Suzaku and Swift Data

Parameters	Epoch 1	Epoch 2 (1)	Epoch 2 (2)
aL2–10 keV	8.1	9.3	9.0
bNgalH	1.0e	1.0e	1.0e
Γ	1.65 ± 0.04	1.69 ± 0.02	1.65e
Ecut [keV]	105+28−19	127+25−23	105e
R(=Ω/2π)	0.17 ± 0.04	f	f
EW [eV] (Fe Kα line)	21	18	18
cf	0.21 ± 0.03	0.21e	0.40+0.26−0.15
dN1H	7.9+1.4−1.3	7.9e	2.9+1.6−1.0
dN2H	0.75 ± 0.03	0.75e	0.49+0.17−0.33
χ2/dof	926.9/812	201.1/182	192.7/181

Notes. Errors are 90% confidence level for a single parameter. ^aIntrinsic luminosity in the 2–10 keV band corrected for both Galactic and intrinsic absorptions in units of 10⁴³ erg s^-1. ^bGalactic absorption column density in units of 10²² cm^-2. ^cCovering fraction. ^dIntrinsic absorption column density in units of 10²² cm^-2. ^eParameters fixed at these values. ^fWe assume the same reflection component as that determined in epoch 1.

Download table as:  ASCIITypeset image

3.4. Spectral Analysis with Comptonization Model

In this subsection, we analyze the spectra of 4C 50.55 with a physically motivated model instead of the phenomenological "cutoff power-law" model, which is a mathematical approximation of the X-ray spectra of AGNs. Such analysis of AGN spectra has been very limited so far, since it requires high quality broadband data. As we will discuss in Section 3.5, the contribution from the jet components is very small in the X-ray band. Hence, we consider that the origin of the continuum emission is predominantly thermal Comptonization of soft (ultra-violet) photons off hot electrons in the corona located above the accretion disk. Accordingly, we adopt a thermal Comptonization model, thComp (Zycki et al. 1999), for the primary continuum. It has two free parameters, the slope Γ and electron temperature kT_e. The electron scattering optical depth τ_e is related to T_e and Γ by the following formula (Sunyaev & Titarchuk 1980):

$$\begin{eqnarray} &&\tau _{\rm e} = \sqrt{2.25 + \frac{3}{(T_{\rm e}/511 \ {\rm keV})[(\Gamma + 0.5)^2 -2.25]}}-1.5. \end{eqnarray} \tag{ 1 }$$

For seed photons, we assume a multicolored disk component with the innermost temperature of 0.01 keV. This choice is not important to constrain the Comptonization parameters. As described in the previous subsection, the reflection component and fluorescence iron-K line are self-consistently included, which are blurred by the "diskline" profile with the same parameters as obtained above (r_in = 720 r_g).

We perform a simultaneous fit to the Suzaku (epoch 1 or 2) and Swift/BAT spectra using this model. The best-fit parameters are summarized in Table 3. We obtain the electron temperature of ≈30 keV with an optical depth of ≈3. Again, the reflection component in epoch 2 is fixed at the same one determined from epoch 1. We find that the continuum parameters are consistent with each other between epochs 1 and 2 within the statistical errors at the 90% confidence level. Figure 6 shows the results in the two epochs, in the form of unfolded spectra (i.e., those corrected for the effective area of the instruments) in units of EI(E), where E is photon energy and I(E) the energy flux.

Zoom In Zoom Out Reset image size

Table 3. Best-fit Parameters of Comptonization Model to the Combined Suzaku and Swift Data

Parameters	Epoch 1 and BAT	Epoch 2 and BAT
aL2–10 keV	8.4	9.6
bNgalH	1.0f	1.0f
Γ	1.78 ± 0.02	1.81 ± 0.01
kTe [keV]	31+10−6	46+37−13
c(τe	2.8+0.4−0.5	2.1+0.5−0.8)
R(=Ω/2π)	0.16 ± 0.04	g
EW [eV] (Fe Kα line)	19	17
df	0.27 ± 0.02	0.27f
eN1H	8.6+1.1−1.0	8.6f
eN2H	0.80 ± 0.03	0.80f
χ2/dof	942.5/812	200.8/182

Notes. Errors are 90% confidence level for a single parameter. ^aIntrinsic luminosity in the 2–10 keV band corrected for both Galactic and intrinsic absorptions in units of 10⁴³ erg s^-1. ^bGalactic absorption column density in units of 10²² cm^-2. ^cElectron-scattering optical depth calculated from Equation (1). ^dCovering fraction. ^eIntrinsic absorption column density in units of 10²² cm^-2. ^fParameters fixed at these values. ^gWe assume the same reflection component as that determined in epoch 1.

Download table as:  ASCIITypeset image

3.5. Multi-wavelength Spectral Energy Distribution

In Figure 7, we summarize the available multi-wavelength fluxes of 4C 50.55 to discuss its SED from the radio to γ-ray bands. In the radio band, we utilize those obtained by the VLA and the Giant Metrewave Radio Telescope (GMRT) compiled by Molina et al. (2007). In the GeV γ-ray band, the upper limit derived from the one-year Fermi data (Abdo et al. 2009) is indicated by the arrow. We plot the results of the cutoff power-law fit to the Suzaku spectra in epochs 1 (black) and 2 (red) as well as the time-averaged Swift/BAT data, correcting for both Galactic and intrinsic absorptions. To examine the origin of the time variability, we analyze the difference spectrum of XIS-FIs between epochs 1 and 2. Fitting with a single power law with the same double absorption as given in Table 1, we find Γ ≈ 1.8. This result is also plotted in the figure (green curve).

Zoom In Zoom Out Reset image size

Assuming that the global intrinsic SED of the jet emission in 4C 50.55 is similar to that of jet-dominated sources (Kubo et al. 1998) with correction for an estimated beaming factor (δ ∼ 1 for 4C 50.55 and δ ∼ 10 for blazars), the predicted X-ray emission is a factor of ∼10^1–4 less than the observed X-ray flux. We thus conclude that X-ray emission due to a jet via synchrotron emission or inverse Compton should be very small in the total X-ray emission of this source. Based on the Suzaku results, the ratio of the core luminosity at 5 GHz and that in the 2–10 keV band is estimated to be log R_X = −3.6. Thus, while 4C 50.55 should be classified as a radio-loud object (Terashima & Wilson 2003), its radio to X-ray power ratio is much lower compared with typical blazars and more powerful radio galaxies, like 3C 120 (log R_X = −2.1; Kataoka et al. 2007). This confirms the conclusion by Molina et al. (2007), who did not find any evidence for an additional power-law component representing the jet emission in the XMM-Newton spectrum.

We note that the optical fluxes of 4C 50.55 are highly uncertain at present due to the extinction (both Galactic and intrinsic ones), which could greatly affect the empirical estimate of the black hole mass. In fact, the extinction-corrected AGN continuum flux at 5100 Å adopted by Winter et al. (2010) falls far below an extrapolation from the X-ray spectra toward lower energies. As a simpler approach, assuming typical SEDs of AGNs (Elvis et al. 1994; Grupe et al. 2010), we roughly estimate the 5100 Å flux to be (0.7–8) × 10⁻¹⁰ erg cm⁻² s⁻¹ from the BAT flux averaged for 22 months. Then, from the luminosity at 5100 Å, (0.6–7) × 10⁴⁴ erg s⁻¹, and the width of the H_β line, 2320 km s⁻¹ FWHM (Winter et al. 2010), we derive the black hole mass of 4C 50.55 to be ∼10^7.8±0.3 M_☉, using the formula given by Vestergaard & Peterson (2006). This is about 10 times higher than the estimate (10^6.58±0.07 M_☉) by Winter et al. (2010). Nevertheless, the suggestion that the black hole in 4C 50.55 accretes with a high Eddington ratio still holds. With a typical bolometric correction factor (L_bol/L_2–10 keV= 30), we estimate the fraction of Eddington luminosity of 4C 50.55 to be L_bol/L_Edd ∼ 0.4.

We have obtained the first simultaneous, broadband X-ray spectra of 4C 50.55 over the 1–60 keV band with Suzaku. These provide one of the best quality high energy data so far obtained from BLRGs. From the SED, we conclude that there is little contribution from jets in the X-ray emission, which is thus dominated by Comptonization by hot corona. We find that the overall continuum is represented by a cutoff power law, or a thermal Comptonization model, with complex intrinsic absorptions. We show that at least two absorbers with different covering fractions and column densities are necessary to model the spectrum. The photon index and cutoff energy of 4C 50.55 are within the distribution of previous observations of BLRGs (Grandi et al. 2006), although both are somewhat smaller (Γ ≈ 1.6 and E_cut ≈ 100 keV) compared with typical values found from Seyfert galaxies (Γ≈1.9 and E_cut∼200 keV; Dadina 2008). These findings on 4C 50.55 are fully consistent with the results by Molina et al. (2007) from the XMM-Newton and INTEGRAL data observed on earlier epochs. We estimate that the corona responsible for Comptonization is optically thick for scattering τ_e ≈ 3 and is relatively cool T_e ≈ 30 keV. The optical depth is larger and the temperature is lower than those obtained from Seyfert galaxies with thermal Comptonization models (e.g., Lubinski et al. 2010), as expected from the comparison in Γ and E_cut.

From our observations, time variability of the X-ray flux on both long (≳10⁶ s) and short (∼10⁴ s) timescales is detected. The averaged 2–10 keV flux was the highest in our Suzaku observation among those in the previous observations reported in Molina et al. (2007) by a factor of 1.3–2.5. The hard X-ray flux in the 17–100 keV band (1.8 × 10⁻¹⁰ erg cm⁻² s⁻¹) was also higher by 1.4 than the averaged flux over 22 months obtained with Swift/BAT (1.3 × 10⁻¹⁰ erg cm⁻² s⁻¹). The large long-term variability in the hard X-ray above 10 keV suggests that it is mainly produced by the intrinsic emission, not purely by the change of the absorber as discussed by Molina et al. (2007) in the framework of a "patch torus" model (Elitzur & Shlosman 2006). While there remains a possibility that Compton thick absorber (N_H>10²⁴ cm⁻²) may completely cover a part of the emission region to cause the flux variability, it would produce signals of heavy obscuration such as a deep iron-K edge feature in the X-ray spectra, which are not seen in the data. Thus, at least variability of the continuum flux is required to explain these results, while that of the absorber can also contribute to the variability, in particular below 10 keV (see, e.g., Risaliti et al. 2005 for NGC 1365).

We significantly detect an iron-K emission line and obtain a tight constraint on the reflection component, even though it is quite weak as reported by Molina et al. (2007). The reflection strength, R ≃ 0.2, is much smaller compared with Seyfert 1 galaxies, which typically have R ∼ 1 (Dadina 2008). Even correcting for a variability effect that the direct continuum flux was higher in our observations by a factor of ≈1.4 than the 22 month average, the reflection is still small, R ≃ 0.3. The weak reflection is also consistent with the narrow iron-K emission line, which indicates that the reflection is mainly produced by relatively outer parts of the disk (hence with a small solid angle), unlike the results from typical Seyfert 1 galaxies (Dadina 2008). Our 4C 50.55 result confirms the trend reported by Sambruna et al. (2002) for radio-loud AGNs.

The analysis of the iron-K line profile yields an (apparently) large innermost radius, r_in ∼ 700 r_g, by assuming an emissivity law of r⁻³. Our diskline result suggests it unlikely that a "standard disk" extends down to close to the innermost stable circular orbit (ISCO) around the black hole, <6 r_g. Instead, it is possible that the disk is there but its inner part is covered by an optically thick corona, as estimated by our Comptonization model fit, which would smear out relativistic iron-K broad lines. Note that the obtained r_in value does not directly mean that the disk is truncated at that radius, because (1) the estimated r_in critically depends on the emissivity profile and (2) there may be another line component from distant parts, such as the torus. If the scale height of the X-ray irradiating corona is sufficiently small, then one would expect a flatter slope for the emissivity law, even close to r⁻². In this extreme case, we were not able to obtain good constraints on r_in from our data. To examine the second possibility, we apply a two-component line model to the XIS 3–9 keV spectra, consisting of a narrow Gaussian at 6.4 keV and a broad disk line, which represents that from the torus and disk, respectively. We obtain a worse fit than the single diskline fit by Δχ² = 5 even with a larger degrees of freedom, suggesting that the two-component model is not a good description of the data. Nevertheless, when r_in of the diskline component is fixed at 10 r_g as found from 3C 120 (Kataoka et al. 2007), both lines are found to be significant with equivalent widths of 19 ± 6 eV and 24 ± 15 eV, respectively. Thus, we do not completely exclude the possibility for the presence of a moderately broadened iron-K line in the observed spectra of 4C 50.55.

The inferred geometry of the accretion disk in 4C 50.55 (i.e., truncated and/or inner parts covered by corona) may be common features of AGNs with powerful jets. Recent Suzaku studies indicate that radio galaxies also have relatively narrow iron-K emission lines, e.g., r_in> 20 r_g for 3C 390.3 (Sambruna et al. 2009) and r_in> 44 r_g for 4C +74.26 (Larsson et al. 2008) from the single diskline fit, and r_in = (9 ± 1) r_g for 3C 120 from the multiple-component fit (Kataoka et al. 2007). This result is in accordance with an expectation from theories that jets are more easily produced by radiatively inefficient accretion flow than by a standard disk.

Another key parameter to understand the accretion flow is the Eddington ratio, which is estimated to be L_bol/L_Edd ∼ 0.4 for 4C 50.55 (Section 3.5). Similarly, we also estimate that of 3C 120 to be L_bol/L_Edd ∼ 0.5, using the 2–10 keV flux (Kataoka et al. 2007) and the black hole mass of 10^7.7 M_☉ (Peterson et al. 2004). Thus, these two sources may belong to a very similar class of AGNs, except for the radio loudness to the X-ray flux (log R_X = −2.1 for 3C 120 and log R_X = −3.6 for 4C 50.55), which could be partially explained by the small inclination angle of 3C 120 (i < 14°; see Kataoka et al. 2007) compared with 4C 50.55 (i ∼ 35°). The physical reason for the difference in their X-ray spectra that the reflection component and iron-K lines are stronger in 3C 120 (R ∼ 0.7) is not clear at present.

4C 50.55 and 3C 120 are rare objects having distinctively high fractions of Eddington luminosity compared with other typical BLRGs, for instance, L_bol/L_Edd = 0.01–0.07 for 3C 390.3 (Sambruna et al. 2009; Lewis & Eracleous 2006), ∼0.04 for 4C +74.26 (Larsson et al. 2008), and 0.001–0.002 for Arp 102B (Lewis & Eracleous 2006). By analogy to the Galactic black holes, these low Eddington ratio sources likely correspond to the low/hard state, where the accretion disk is accompanied by steady jets, while normal Seyfert galaxies may correspond to the high/soft state, where the disk extends close to the ISCO with quenched jet activity. The accretion flows in 4C 50.55 and 3C 120 could be explained as a high luminosity end of the low/hard state. Alternatively, they may be another state achieved with even higher mass accretion rates than in the high/soft state, where the disk structure is also similar to that found in the low/hard state (i.e., truncated disk). For this possibility, it is interesting to note the similarity to the high Eddington ratio Galactic black hole GRS 1915+105, which exhibits a similarly narrow iron-K emission line over a Comptonization-dominated continuum, implying that the inner disk is fully covered by a corona (Ueda et al. 2010); in GRS 1915+105, a compact jet is also detected in a steady state with a hard spectrum, so-called in Class χ (see e.g., Fender & Belloni 2004). In summary, the unified picture of accretion flows over a wide range of black hole mass is far from established. Further systematic studies of the accretion disk structure of radio-loud AGNs at various accretion rates based on detailed X-ray spectroscopy and multi-wavelength data are very important to reveal these fundamental problems.

We thank Gerry Skinner for providing the Swift/BAT light curve of 4C 50.55, and the Suzaku team for the calibration of the instruments. Part of this work was financially supported by the Grant-in-Aid for Scientific Research 20540230 (YU) and 20740109 (YT), and by the Grant-in-Aid for the Global COE Program "The Next Generation of Physics, Spun from Universality and Emergence" from the Ministry of Education, Culture, Sports, Science and Technology (MEXT) of Japan.