X- AND γ-RAY PULSATIONS OF THE NEARBY RADIO-FAINT PSR J1741−2054

The launch of the Fermi γ-ray Space Telescope offered the first opportunity to study a sizeable population of γ-ray pulsars. The Fermi Large Area Telescope (LAT; Atwood et al. 2009) has discovered pulsed γ-ray signals from more than 150 objects (Abdo et al. 2013), revolutionizing our view of them and giving birth to new high-energy pulsar sub-families, such as millisecond (see, e.g., Abdo et al. 2009b; Ransom et al. 2011; Keith et al. 2011; Espinoza et al. 2013) and radio-quiet γ-ray pulsars (see, e.g., Abdo et al. 2009a; Saz Parkinson et al. 2010; Pletsch et al. 2013), as numerous as the classic family of young, radio-loud pulsars (Caraveo 2013). The wealth of detections confirms the importance of the γ-ray channel in the overall energy budget of rotation-powered pulsars and paves the way to understanding the three-dimensional structure and electrodynamics of neutron star magnetospheres. Indeed, radio and γ-ray light curves contain a great deal of useful information about pulsar emission processes (see, e.g., Watters & Romani 2011; Pierbattista et al. 2012, 2014), confirming that models with emission originating at high altitudes in the magnetosphere (e.g., outer and slot-gap; Cheng et al. 1986; Harding & Muslimov 2004) are favored over models with near-surface emission (e.g., polar cap; Harding 2013).

Fitting γ-ray and radio light curves simultaneously is a promising way to constrain pulsar geometric parameters (e.g., Pierbattista et al. 2014). Using the information in the (magnetospheric) non-thermal pulsar X-ray light curves could further improve the approach, adding another piece to the pulsar emission puzzle. This approach could localize the emitting region(s) responsible for the non-thermal pulsed X-ray emission with respect to the high altitude gamma-ray emitting region.

Few X-ray light curves have been exploited for modeling magnetospheric emission, compared to γ-ray profiles. This is largely due to the lack of high-quality X-ray light curves, primarily due to the occasional and non-targeted observational efforts. At this time, ∼60 out of 77 young pulsars in the second Fermi LAT pulsar catalog (2PC; Abdo et al. 2013) have been detected in X-rays (Becker 2009; Marelli et al. 2011; Marelli 2012; Abdo et al. 2013), but X-ray pulsations have been detected from fewer than half of them. Only nine Fermi pulsars have both the high X-ray fluxes and the long dedicated X-ray observations needed to disentangle the thermal and the non-thermal pulsations. Only five of these (Crab, Vela, Geminga, PSR J0659+1414, and PSR J1057−5226) have been characterized by a multi-bin phase-resolved X-ray spectral analysis (De Luca et al. 2005; Manzali et al. 2007; Weisskopf et al. 2011). Of these, only the Crab (and possibly Geminga, Jackson & Halpern 2005) yielded a non-thermal light curve with a photon index varying with phase, a behavior that is still puzzling (Harding et al. 2008; Tang et al. 2008; Hirotani 2008).

With the notable exception of the Crab among young pulsars, the multiwavelength behavior of isolated neutron stars is complex, with radio, optical, X-ray, and γ-ray light curves usually unaligned, pointing to different emitting regions in the pulsar magnetosphere. The rich phenomenology, in particular including the X-ray information, has not yet been fully exploited for modeling the radiation processes of pulsars, leaving a number of questions unsolved.

Here we report the results of a deep XMM-Newton observation intended for phase-resolved X-ray spectral analysis of the Fermi pulsar J1741−2054 (hereafter J1741). The middle-aged J1741 (τ_c = 390 kyr) was discovered in a blind pulsation search of a Fermi LAT point source (Abdo et al. 2009a). For a moment of inertia I = 10⁴⁵ g cm², its period P = 413 ms and period derivative $\dot{P}$ = 1.7 × 10⁻¹⁴ s s⁻¹ give a spin-down energy loss rate $\dot{E}$ = 9× 10³³ erg s⁻¹, clearly on the low side of the γ-ray pulsar distribution (see Figure 1 in Abdo et al. 2013). The pulsar was then detected in archival Parkes radio observations, showing a remarkably low dispersion measure, DM = 4.7 pc cm⁻³ (Camilo et al. 2009). The Galactic electron density model of Cordes & Lazio (2002) yields a distance of ∼0.38 kpc, making J1741 one of the closest γ-ray pulsars known. At this distance, the low observed 1400 MHz radio flux density, S_1.4 ∼ 160 μJy, makes it the faintest radio pulsar known.

At the position obtained from LAT timing analysis (Ray et al. 2011), Abdo et al. (2009a), Romani et al. (2010), and Marelli et al. (2011) found the X-ray counterpart using both Swift and Chandra data. The Chandra observation also revealed diffuse, faint X-ray emission due to a pulsar wind nebula (PWN) trail extending some 2' at position angle P.A. = 45° ± 5° (north through east). This extended structure was also associated with a 20'' long Hα bow shock. Accurate bow shock modeling by Romani et al. (2010) suggests a pulsar velocity v_p ∼ 150 km s⁻¹ directed 15° ± 10° out of the plane of the sky.

Our deep XMM-Newton observation of J1741 started on 2013 February 28 at 19:50:39 UT and lasted 70.9 ks. The PN (Struder et al. 2001) of the European Photon Imaging Camera (EPIC) was operating in Small Window mode (time resolution of 5.6 ms over a 4' × 4' field of view, FOV), while the Metal Oxide Semi-conductor (MOS) detectors (Turner et al. 2001) were set in Full Frame mode (2.6 s time resolution and a 15' radius FOV). The thin optical filters were used for the PN and MOS cameras. For our analysis, we used the XMM-Newton Science Analysis Software (SAS) v13.0. To fully characterize both the pulsar and the nebula, we also used the available Chandra Advanced CCD Imaging Spectrometer (ACIS; Garmire et al. 2003) observation of the field, performed on 2010 May 21 and lasting 48.8 ks (these data were included in Romani et al. 2010). To better characterize the pulsar, we also took advantage of the ∼300 ks of data collected as part of the Cycle 14 Chandra Visionary Project "A Legacy Study of the Relativistic Shocks of PWNe" by R. Romani. We retrieved "level 1" data from the Chandra Science Archive and used the Chandra Interactive Analysis of Observation (CIAO) software v.4.3. For all the data sets, we followed the standard data processing approach, using the same procedures reported in Marelli et al. (2013).

The Fermi LAT data set we used to extend the γ-ray ephemeris of J1741 spans five years, from 2008 August 4 to 2013 August 4. P7REP Source class events were selected with reconstructed energies from 0.1 to 100 GeV from an area within 20° of the source position. We excluded γ-rays collected when the LAT was not in nominal science operations mode, when the spacecraft rocking angle exceeded 52°, when the Sun was within 5° of the pulsar position, and those with measured zenith angles >100°, to reduce contamination by residual γ-rays from the bright limb of the Earth. We performed a binned maximum likelihood analysis, as reported in Abdo et al. (2013). We used the Fermi Science tools v09r32p04, instrument response functions P7REP_SOURCE_V15, and the Galactic and isotropic models obtained by the LAT collaboration from the analysis of four years of data. The analysis tools, instrument response functions, and diffuse emission models are available from the Fermi Science Support Center, http://fermi.gsfc.nasa.gov/ssc. The source models were taken from the two-year catalog (Nolan et al. 2012) and 2PC. Post-fit spatial residuals did not reveal the need for any additional sources, beyond those in the two-year catalog, in our model of the region. The pulsar γ-ray spectrum is consistent with a power law with an exponential cutoff with a photon index Γ = 1.10 ± 0.10 and cutoff energy E_c = 0.92 ± 0.06 GeV. These results are in full agreement with those in 2PC.

The Fermi Science Tool gtsrcprob combines the spectral results with the LAT's energy-dependent point-spread function (PSF) to assign each event its probability of coming from the pulsar (Kerr 2011). We used only barycentered events with a probability higher than 0.01 for the following timing analysis. The rotation ephemeris used in 2PC spans only 3 yr: we extended it, using a weighted Markov Chain Monte Carlo algorithm (MCMC; see, e.g., Wang et al. 2013). Adding six months of data in each iteration, we re-evaluated the timing solution using the weighted H test (see, e.g., De Jager & Busching 2010). The highest H value of 4782.59 (20 harmonics) for five years of data was found for F₀ = 2.41720384698 Hz, F₁ = −9.93133e-14, F₂ = −5.924e-25, and F₃ = 5.75e-32, with epoch zero at MJD 55631.0002.¹⁰

With this ephemeris, we assigned a rotational phase to each γ-ray event and filled a 100 bin >0.1 GeV phase histogram with bin uncertainties using the photon weights, following Abdo et al. (2013; see Section 6).

We performed a γ-ray phase-resolved spectral analysis on the five-year data set. For 20 phase bins, we re-ran the binned spectral analysis, leaving only one or more of the pulsar parameters (normalization, photon index, and energy cutoff) free to vary. The Test Statistic value (see, e.g., Mattox at al. 1996) does not vary significantly by freeing one or two of the parameters so that the addition of the additional degrees of freedom are not justified. Moreover, with both the photon index and the cutoff energy free, their fit with a constant is acceptable (null hypothesis probability nhp = 6 ×  10⁻⁴ and 6 ×  10⁻³ for the photon index and cutoff, respectively). Thus, we found no indication of spectral variation as a function of the pulsar phase.

Following the method reported in Marelli et al. (2013), we detected and selected the active galactic nucleus (AGN) like serendipitous sources in the XMM-Newton and Chandra FOVs (see Figure 1). The spectra of these sources were fitted together, linking their hydrogen column densities N_H to assess the average Galactic absorption column. The resulting value, N$_H^{\rm gal}$ = (4.54 ± 0.36) × 10²¹ cm⁻² (90% confidence error), is slightly greater than the value of 3×10²¹ cm⁻² obtained from the 21 cm H i sky survey of Kalberla et al. (2005). This points to the absence of thick molecular clouds in the XMM-Newton FOV. Since N$_H^{\rm gal}$ obtained from X-ray absorption probes all types of nuclei along the line of sight, whereas the latter value samples only atomic hydrogen, this observed value of N_H is to be expected (Dickey & Lockman 1990).

Zoom In Zoom Out Reset image size

Apart from the pulsar itself, the main feature apparent in the PN FOV is the PWN emission. To find the best extraction regions for the pulsar and the nebula, as well as to better evaluate the PWN shape, we created exposure-corrected radial (around the pulsar) and linear (parallel and orthogonal to the main axis of the nebula) brightness profiles in the 0.3–10 keV energy range. The radial profile is consistent (nhp = 0.14) with the nominal instrument PSF¹¹ only up to 40'' from the pulsar due to the relative faintness of the nebula. The linear profile, with 1' width centered on the pulsar and orthogonal to the main axis of the nebula, is also consistent (nhp = 0.009) with the PSF, proving the absence of a detectable nebular component in the orthogonal region. The presence of the nebula in the profile parallel to the main axis is apparent, extending up to ∼2' from the pulsar and with flux decreasing with distance (apparent in Figure 1, see also Romani et al. 2010). From the deep Chandra observations, the unusual shattered shape of the tail-like nebula is apparent, appearing to be divided into three blobs.

Using a 25'' radius extraction circle around the pulsar, we obtained 9794, 2983, and 3073 counts in the 0.3–10 keV energy range in the PN and the two MOS detectors, respectively, with a background contribution of less than 4%. We also used 19,380 pulsar counts from the Chandra observations. For the spectral analysis, we followed Marelli et al. (2013). To better constrain the column density, we fitted the nebular spectra together with the pulsar spectrum. We obtained 1508, 557, and 533 counts in the 0.3–10 keV energy range in the PN and the two MOS detectors, respectively, with a background contribution of 40% and 25%. We obtained 5633 nebular counts, with a 30% background contribution from the Chandra observations.

One-component models are not statistically acceptable for the pulsar spectrum ($\chi ^2_{r}$ = 2.2, nhp ∼0 for an absorbed power law and $\chi ^2_{r}$ = 11.6, nhp = 0 for an absorbed blackbody, where the χ² has been reduced for 781 degrees of freedom, or "dof"). The pulsar spectrum is consistent ($\chi ^2_{r}$ = 1.18, nhp = 2 × 10⁻³, 779 dof) with a combination of a power-law component with a photon index of Γ = 2.68 ± 0.04 (90% confidence errors) and a blackbody with a temperature of T = (7.07 ± 0.19) × 10⁵ K, and an emitting radius R = (5.39$_{-0.71}^{+0.81})\times d_{380}$ km (where d₃₈₀ is the distance of the pulsar in units of 380 pc, derived from the dispersion measure of the pulsar using the NE2001 electron model of the Galaxy; Camilo et al. 2009), absorbed by a column density of N_H = (1.21 ± 0.01) × 10²¹ cm⁻², about one-fourth of the Galactic value. The PWN spectrum is consistent with an absorbed power law with Γ = 1.74±0.07. The unabsorbed fluxes of the non-thermal and thermal components of the pulsar spectrum are 5.47 ± 0.13 × 10⁻¹³ and 7.63 ± 0.19 × 10⁻¹³ erg cm⁻² s⁻¹, respectively. The unabsorbed nebular flux is 1.40 ± 0.09 × 10⁻¹³ erg cm⁻² s⁻¹. We note that a three-component model is not statistically compelling (for a power law plus double blackbody, we obtain $\chi ^2_{r}$ = 1.18, nhp = 3 × 10⁻³, 777 dof). We also note that a composite non-thermal plus a magnetized neutron star atmosphere model (nsa in XSpec; assuming a neutron star with a radius of 12 km, mass of 1.4 M_☉, and a surface magnetic field of 10¹³ G) gives a very poor fit. We obtain a $\chi ^2_{r}$ = 1.22, nhp = 7 × 10⁻⁵, 779 dof, with a lower emitting temperature of (3.02 ± 0.12) × 10⁵ K.

To study the possible spectral variation of the nebula with angular distance from the pulsar, we divided the nebula into three different regions on the basis of their angular separations from the pulsar. We chose to equally divide the main axis of the ellipsoidal nebular region in order to consider the three blobs. The different PSFs, as well as different exposure maps, of the two X-ray instruments have been taken into account. All the parameters of the spectra from the corresponding regions in Chandra and XMM-Newton observations were linked, also freezing the absorption column to the previously fitted value. We see marginal evidence for a spectral variation (in fact, linking the photon index for the three spectra, we obtain $\chi ^2_{r}$ = 1.45, nhp = 6.4× 10⁻⁵, 184 dof), with Γ_near = 1.72 ± 0.09 (90% confidence errors), Γ_med = 1.90 m ± 0.10, and Γ_far = 1.69 ± 0.09.

For the timing analysis, a PATTERN selection was performed as by Marelli et al. (2013) and the X-ray photon arrival times were barycentered to the Chandra source position (R.A., decl. (J2000) 17:41:57.28, −20:54:11.8) from Romani et al. (2010), which is consistent with the γ-ray timing position. We then phase-folded the X-ray photons using our Fermi LAT timing solution, contemporaneous with our XMM-Newton data set. We extracted 11,507 PN events in the 0.15–10 keV energy range from the 25'' circle centered on the pulsar. We repeated the exercise for photons in different energy ranges.

Pulsations with >20σ significance appear in the 0.15–10 keV energy range (H-test = 585; tail probability = 0, and $\chi ^2_{r}$ = 29, 19 dof, according to a χ² test on the folded curve testing a constant model). The profile has a single peak, lagging the radio peak by 0.6 in phase. Light curves for different energy ranges are shown in Figure 2, phase-aligned with the γ-ray and radio profiles. Using the phase-averaged composite spectrum, we evaluated the background contribution, the pulsar blackbody, and the power-law components in different energy ranges, as in Caraveo et al. (2010). We obtained a background-subtracted pulsed fraction of 36.1% ± 1.5% in the 0.3–10 keV energy range. Dividing the energy range into low energy (0.3–0.7 keV, where the blackbody accounts for ∼60% of the counts) and high-energy bands (0.7–10 keV, where the power law accounts for ∼90% of the counts) we obtained pulsed fractions of 35.8% ± 1.5% and 36.4% ± 1.5%, respectively. We know the pulsed fraction in each band and the percentage of blackbody, power-law, and background contributions so that we can get the net pulsed fraction of the two spectral components. The pulsar's power law is pulsed at ∼38% and the blackbody component at ∼36%. Although at lower energies, the curve is quasi-sinusoidal (a fit with a sinusoid gives $\chi ^2_{r}$ = 1.31, 17 dof, nhp = 0.17) due to the blackbody contribution, the profile at higher energies is not compatible with a sinusoid ($\chi ^2_{r}$ = 2.70, 17 dof, nhp = 10⁻⁴). The correlation of the power-law photon index with its normalization (see, e.g., Figure 4) somewhat biases the normalization profile in Figure 3, while the counts profiles in Figure 2 are better suited for comparing their shapes. The peak positions at low (0.3–0.7 keV) and high (0.7–10 keV) X-ray energies are separated by less than 0.1 in phase, but they are offset by about a half-rotation from the main γ-ray pulse and the radio pulse (following Abdo et al. (2013), the radio pulse leads the γ-ray pulse by 0.074 ± 0.006 in phase).

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

To search for possible variation of the X-ray spectral parameters with rotational phase, we first fitted the on- and off-pulse spectra. We define the on-pulse interval to be between phase 0.3 and 0.8 (values obtained from a fit with a double-step model) and the remaining phase bins as off-pulse. Fitting only the power-law counts (E > 0.7 keV), we fixed the column density and the photon index to the phase-averaged values, obtaining an acceptable nhp = 0.047 ($\chi ^2_{r}$ = 1.21, 143 dof). Freeing both the normalization and photon indices we obtain an improved fit (nhp = 0.22, $\chi ^2_{r}$ = 1.09, 141 dof). An f test (Bevington 1969) shows that the probability for a chance improvement is 2.6$\times 10\smash{^{-4}}$, pointing to a significant effect when both the power-law normalization and photon index are free to vary. Next, we fixed the power-law values obtained in the previous fit and evaluated the blackbody component in the 0.3–0.7 keV energy band. Thawing only the blackbody normalization, we obtain an nhp = 0.74 ($\chi ^2_{r}$ = 0.93, 128 dof). A variation of the blackbody temperature with pulsar phase is not statistically compelling; however, the high pulsed fraction at low energies implies significant pulsation of the blackbody normalization.

The blackbody normalization and power-law normalization and photon index vary with pulsar phase. We divided our data set into 10 phase bins, each with width 0.1 and containing ∼1000 counts, and fitted the obtained spectra fixing the column density and linking blackbody temperatures. We obtained the curves reported in Figure 3, with the photon index gradually increasing with phase and the normalizations following the light curve, with no statistically significant differences between blackbody and power-law peaks. The variation in photon index is also shown by the confidence contours for the 10 phase bins, plotted in Figure 4 for the plane of the index and the power law normalization. The variation with blackbody normalization is similar.

By analyzing the new XMM-Newton observation of PSR J1741−2054, we fully characterized the high-energy emission of this nearby middle-age radio-faint pulsar.

Its nebular emission is typical of pulsar wind nebulae, both for its non-thermal spectrum and for the flux decrease with distance along the tail emanating from the pulsar. A hint of nebular spectral variation with distance from the pulsar is detected. The shattered shape of the nebula is peculiar and a deeper analysis of the new Chandra data is needed to better understand this unusual behavior. These observations could also provide a measurement of the pulsar proper motion. Such an analysis is beyond the scope of this work.

Modeling the X-ray spectrum of the pulsar requires a composite model, summing thermal and non-thermal components. Both components are ∼35%–40% pulsed, with single-peaked light curves and maxima phase-aligned to within 0.1 in phase. While the thermal light curve is compatible with a sinusoid, the non-thermal profile has a sharper peak. The best-fitting thermal spectrum yields a temperature (∼7 × 10⁵ K), compatible with the theoretical expectations for the cooling of a 390 kyr old pulsar (Pons et al. 2009). We note that the best-fit temperature of the nsa model (∼3 × 10⁵ K) is below the theoretical expectations, further disfavoring such a model. A pulsed component from thermal cooling has already been noted for several pulsars (Caraveo et al. 2004; De Luca et al. 2005; Manzali et al. 2007), and that of PSR J1057−5226 has a similar pulsed component. Such pulsations can be ascribed to a dependence of the observed emitting area on the line of sight. Halpern & Ruderman (1993) describe a magnetospheric "blanket" caused by cyclotron resonance scattering off the plasma in the magnetosphere that could screen the thermally emitting surface during specific phase intervals, depending on the magnetic field configuration and viewing geometry. Anisotropic heat transfer from the pulsar interior can also explain flux variation across the neutron star surface (Greenstein & Hartke 1983). If the thermal component is due to the cooling of the entire neutron star surface (12 km radius) from the thermal normalization in the best-fit spectrum (which depends only on the pulsar distance and the emitting radius), we can derive the pulsar distance to be ∼850 pc, with 3σ limits of ∼650 and ∼1100 pc (e.g., Halpern et al. 2007). On the other hand, thermal emission from polar caps heated due to downstreaming of e^± (Harding & Muslimov 2002), as seen in the case of PSR J0007+7303 in the CTA 1 supernova remnant, is expected to be generated from much smaller regions (<100 m, based on a simple "centered" dipole magnetic field geometry, De Luca et al. 2005; Marelli et al. 2013), which would imply an unrealistic pulsar distance smaller than 10 pc. Distances lower than a few hundred parsecs are greatly disfavored due to the non-negligible value of the column density (about one fourth of the total Galactic column density in this direction). We note that a distance of 850 pc would result in a 110% γ-ray efficiency, defined as $L_\gamma /\dot{E}$, with L_γ the luminosity above 100 MeV. The distance range cited above implies an efficiency range of 60%–180%. Similarly, the X-ray efficiency would be ∼1.2% (0.7%–2.1%). Such an unrealistic γ-ray efficiency could be explained by a beaming factor f_Ω less than 1 (as defined in 2PC), or by a moment of inertia larger than 10⁴⁵ g cm².

A 10 bin phase-resolved X-ray spectral analysis reveals variations in the X-ray photon index in addition to the phase-varying normalizations of the two spectral components, with a softer spectrum during the on-pulse phases (Figure 3). It is difficult to compare the behavior of J1741 with that of other Fermi pulsars, since the Crab is the only one for which such variation was detected, and the Crab's non-thermal spectrum becomes softer in the primary-pulse maximum and harder during the bridge between the two maxima (Weisskopf et al. 2011). Although some models have been developed to explain the Crab's optical-to-γ-ray behavior (see, e.g., Harding et al. 2008; Tang et al. 2008; Hirotani 2008), the physics behind the pulsar's X-ray photon index variations are still unclear.

Moreover, unlike the Crab, the γ-ray, X-ray, and radio peaks of PSR J1741−2054 are not aligned, pointing to a clear difference in the geometry and/or altitude above the neutron star surface of the different emitting regions. Indeed, the γ-ray and X-ray peaks are phase-offset by roughly a half rotation, as is also seen in PSRs J0007+7303, J1057−5226, and J0659+1414 (De Luca et al. 2005; Caraveo et al. 2010). Although these differences are expected from different models for the radio and γ-ray bands (Abdo et al. 2013), no model is able to account for the offset between γ-ray and X-ray peaks. The alignment between the thermal and non-thermal X-rays (seen also in other pulsars, e.g., J0659+1414 and J1057−5226), as well as the phase lag with the γ-ray emission coming from the outer magnetosphere, can suggest that the non-thermal emission is generated in a region near the pulsar poles (e.g., in a polar cap emission model). Also, the low X-ray luminosity of radio-quiet pulsars in the X-ray band (Marelli et al. 2011) suggests that the radio and X-ray emission regions may be in close proximity. The origin of the phase lag between the radio and X-ray light curves is unclear. A comprehensive study of pulsar high-energy light curves and phase-resolved spectra will be crucial to understanding the X-ray emission mechanisms and geometry.

This work was supported by the ASI-INAF contract I/037/12/0, art.22 L.240/2010 for the project "Calibrazione ed Analisi del satallite NuSTAR." C.S. was supported in part by NASA ADAP grant NNX13AE64G. G.R.S. is supported by an NSERC discovery grant.

The Fermi LAT Collaboration acknowledges generous ongoing support from a number of agencies and institutes that have supported both the development and the operation of the LAT as well as scientific data analysis. These include the National Aeronautics and Space Administration and the Department of Energy in the United States; the Commissariat à l'Energie Atomique and the Centre National de la Recherche Scientifique/Institut National de Physique Nucléaire et de Physique des Particules in France; the Agenzia Spaziale Italiana and the Istituto Nazionale di Fisica Nucleare in Italy; the Ministry of Education, Culture, Sports, Science and Technology (MEXT), High Energy Accelerator Research Organization (KEK), and Japan Aerospace Exploration Agency (JAXA) in Japan; and the K. A. Wallenberg Foundation, the Swedish Research Council, and the Swedish National Space Board in Sweden. Additional support for science analysis during the operations phase is gratefully acknowledged from the Istituto Nazionale di Astrofisica in Italy and the Centre National d'Études Spatiales in France.

Special thanks to David A. Smith and Tyrel Johnson for reviewing the paper.

Facilities: CXO (ACIS) - Chandra X-ray Observatory satellite, XMM (EPIC) - Newton X-Ray Multimirror Mission satellite, Fermi (LAT) - Fermi Gamma-Ray Space Telescope (formerly GLAST)