ABSTRACT
We present an analysis of the gamma-ray data obtained with the Large Area Telescope (LAT) on board the Fermi Gamma-ray Space Telescope in the direction of SNR W49B (G43.3−0.2). A bright unresolved gamma-ray source detected at a significance of 38σ is found to coincide with SNR W49B. The energy spectrum in the 0.2–200 GeV range gradually steepens toward high energies. The luminosity is estimated to be 1.5 × 1036 (D/8 kpc)2 erg s−1 in this energy range. There is no indication that the gamma-ray emission comes from a pulsar. Assuming that the supernova remnant (SNR) shell is the site of gamma-ray production, the observed spectrum can be explained either by the decay of neutral π mesons produced through the proton–proton collisions or by electron bremsstrahlung. The calculated energy density of relativistic particles responsible for the LAT flux is estimated to be remarkably large, Ue,p>104 eV cm−3, for either gamma-ray production mechanism.
Export citation and abstract BibTeX RIS
1. INTRODUCTION
Galactic cosmic rays are widely believed to be accelerated in supernova remnants (SNRs) through the diffusive shock acceleration process (e.g., Blandford & Eichler 1987). Several SNRs have recently been detected with the Large Area Telescope (LAT) on board the Fermi Gamma-ray Space Telescope; specifically SNRs W51C, Cassiopeia A, W44, and IC443 (Abdo et al. 2009b, 2010a, 2010c, 2010d). Except for Cas A, the LAT-detected SNRs are known to be interacting with molecular clouds. The GeV emission from such SNRs is expected to be dominated by the hadronic gamma rays due to the decay of π0 mesons, since the ambient dense molecular cloud would enhance the proton–proton collisions (Aharonian et al. 1994). The observed gamma-ray sources associated with cloud-interacting SNRs are all seen to be spatially extended in the LAT data. Based on the extension and its comparison with radio data, it is concluded that the gamma-ray emission comes from SNRs not from pulsars/pulsar wind nebulae (PWNe). The LAT spectra of these SNRs steepen above a few GeV. Although electron bremsstrahlung cannot be ruled out, π0-decay emission is the most plausible explanation for the observed LAT data (Abdo et al. 2009b, 2010c). The breaks in the observed spectra may be accounted for by an energy-dependent escape of accelerated protons at SNRs (Aharonian & Atoyan 1996).
The gamma-ray measurements in the TeV range provide direct support for the acceleration of particles up to ∼100 TeV in SNR shells (Aharonian et al. 2007). The TeV gamma rays in SNR RX J1713.7−3946, one of the most prominent examples of TeV-emitting SNRs, can be ascribed to the decay of π0 mesons produced in pp collisions (e.g., Berezhko & Völk 2008) if the average magnetic field strength is larger than ≃15 μG (Uchiyama et al. 2007). However, the emission mechanism remains unsettled largely because of poorly constrained physical conditions in the gamma-ray-emitting zone in SNR RX J1713.7−3946. Other examples are valuable for discriminating the origins of the gamma-ray emission.
SNR W49B (G43.3−0.2) has a bright radio shell and centrally peaked thermal X-ray emission. The interaction between W49B and molecular clouds was evidenced by observations of mid-infrared lines from shocked molecular hydrogen (Reach et al. 2006). H i Zeeman observations also suggest the interaction (Brogan & Troland 2001). Near-infrared [Fe ii] emission exhibits filamentary structures, tracing radiative shocks (Keohane et al. 2007). The age of W49B is estimated to be in the range of ∼1000–4000 yr (Pye et al. 1984; Hwang et al. 2000), and the distance is estimated to be 8–11 kpc (Radhakrishnan et al. 1972; Moffett & Reynolds 1994; Brogan & Troland 2001). The radio continuum map shows a shell structure with a diameter of ∼4' (∼10 pc at 8 kpc). The radio flux density is 38 Jy at 1 GHz. The radio emission is linearly polarized and the spectral index is α = 0.48 (Green 1988) in the frequency range 0.3–30 GHz, indicating a synchrotron origin. No optical emission is detected from the source due to the severe extinction through the Galactic plane. Although the ATNF pulsar database (Manchester et al. 2005)53 lists seven pulsars with the spin-down luminosity >1 × 1034 erg s−1 within 1
0 of the SNR position, no pulsar candidate has been reported within 04. Prior to our LAT observations, gamma-ray emission had not been detected in the GeV or TeV bands.
Here we report the LAT observations in the direction of SNR W49B. A GeV gamma-ray source spatially coincident with W49B is designated as 0FGL J1911.0+0905 in the initial source list published by the Fermi-LAT collaboration, which includes the 205 most significant sources based on the observation in the first three months (Abdo et al. 2009a). It is also designated as 1FGL J1910.9+0906c in the year-1 catalog (1FGL catalog; Abdo et al. 2010f). In this paper, we present a detailed analysis of this LAT source with much longer accumulation time of about 17 months. This paper is organized as follows. In Section 2, the observation and the data reduction are summarized. The analysis results for the LAT source in the direction of SNR W49B are reported using 17 months of the LAT data in Section 3. In Section 4, we discuss whether the gamma rays come from the SNR shell or a pulsar and study the cosmic-ray acceleration using multi-wavelength data.
2. OBSERVATION AND DATA REDUCTION
The Fermi Gamma-ray Space Telescope was launched on 2008 June 11. The LAT on board Fermi is composed of electron–positron pair trackers, featuring solid state silicon trackers and cesium iodide calorimeters, sensitive to photons in a very broad energy band (from 0.02 to >300 GeV). The LAT has a large effective area (∼8000 cm2 above 1 GeV if on-axis), viewing ∼2.4 sr of the full sky with a good angular resolution (68% containment radius better than ∼1° above 1 GeV). The tracker of the LAT is divided into front and back sections. The front section (first 12 planes) has thin converters to improve the point-spread function (PSF), while the back section (four planes after the front section) has thicker converters to enlarge the effective area. The angular resolution of the back events is a factor of 2 worse than that of the front events at 1 GeV.
The LAT data used here were collected for about 17 months from 2008 August 4 to 2009 December 26. The diffuse event class was chosen and photons beyond the earth zenith angle of 105° were excluded to minimize Earth albedo gamma rays.
Among the standard science analysis tools,54 we utilized gtlike for spectral fits and gtfindsrc to find a point source location. With gtlike, an unbinned maximum likelihood fit is performed on the spatial and spectral distributions of observed gamma rays to optimize spectral parameters of the input model taking into account the energy dependence of the PSF. On the other hand, gtfindsrc optimizes a point source location by finding the best likelihood for different positions around an initial guess until the convergence tolerance for a positional fit is reached. The P6_V3 instrument response functions were used for the analyses in this paper. Details of the LAT instrument and data reduction are described in Atwood et al. (2009).
3. ANALYSIS AND RESULTS
3.1. Detection and Source Localization
The LAT observation revealed significant (38σ) gamma-ray emission from the direction of SNR W49B with 17 months of data. Figure 1 shows LAT count maps in the vicinity of SNR W49B in the 2–6 GeV and 6–30 GeV bands. Only front events are used in the count map to achieve better angular resolution. The effective LAT PSF is constructed using a spectral shape obtained through a maximum likelihood fit (gtlike) in the corresponding energy band for each count map (see Section 3.3). The statistical and systematic uncertainties in the spectral shape do not noticeably affect the PSF shape. A Spitzer near-infrared (5.8 μm) map, which traces ionic shocks in the SNR, is overlaid on the count maps.55 Both count maps clearly suggest that gamma-ray emission comes predominantly from the SNR W49B region, not from a nearby star-forming region, W49A. Comparisons between gamma-ray distributions and LAT PSFs in both energy bands indicate that the observed gamma-ray emission could be consistent with a point source.
Figure 1. Fermi-LAT count map in the vicinity of SNR W49B in units of counts per pixel. The pixel size is 001. The LAT localization is represented by a black circle with a radius of 0024 (95% confidence level) centered at (α, δ) = (287756, 9096). Cyan circles represent radii of the effective LAT PSF at 75%, 50%, and 25% of the peak. Magenta and green contours indicate W49B and W49A in the Spitzer IRAC 5.8 μm, respectively. Top: the count map in 2–6 GeV is smoothed by a Gaussian kernel of σ = 02. Bottom: the count map in 6–30 GeV is smoothed by a Gaussian kernel of σ = 01.
Download figure:
Standard image High-resolution imageIn order to confirm the consistency with a point source, a radial profile of the gamma rays from the above source location is compared with that expected for a point source for front events in 2–30 GeV band as shown in Figure 2. The background, which is composed mainly of the Galactic diffuse emission, is subtracted. No sign of spatial extension can be seen in Figure 2.
Figure 2. Radial profile of the LAT count map with front data in 2–30 GeV in units of counts per pixel. The pixel size is 001. The origin of the profile is the LAT localization of the W49B source at (α, δ) = (287756, 9096). The histogram shows the profile of the effective LAT PSF.
Download figure:
Standard image High-resolution imageTo evaluate the consistency with a point source quantitatively, we compared the likelihood of the spectral fit for a point source and an elliptical shape (3' × 4' in size; compatible with the extent of the IR image as shown in Figure 1) with a uniform surface brightness. Here, we assumed a broken power-law function to model the source spectrum in the fit (see Section 3.3 for details). The resulting likelihood was almost the same for both cases (the difference of log likelihood was ∼3), which means that the source emission is consistent with that from a point source. Therefore, to simplify the analyses, the gamma-ray source in the SNR W49B region is analyzed as a point source in this paper. Assuming a point source, the gamma-ray source position was found to be (α, δ) = (287756, 9096) with an error radius of 0024 at 95% confidence level using gtfindsrc, as indicated by the black circle in Figure 1.
3.2. Evaluation of Galactic Diffuse Model
Since uncertainties associated with the underlying Galactic diffuse emission are expected to be the largest systematic effects for spectral analyses of the W49B source, those effects should be carefully evaluated. The uncertainties of the Galactic diffuse emission are primarily due to the imperfection of the Galactic diffuse model and/or the contributions from unresolved point sources. As a first step of the evaluation process, the position and energy dependences of the discrepancies between the observed gamma-ray distributions and the Galactic diffuse model are studied in the regions where the Galactic diffuse emission is considered to be dominant around the W49B source. The normalization of the Galactic diffuse model is determined by running gtlike for a circular region with a radius of 10° centered on the W49B source in the energy range of 0.2–200 GeV. The position of the W49B source is fixed at (α, δ) = (287756, 9096) determined by gtfindsrc (see Section 3.1). The positions and spectral shapes of all other sources are fixed at the value in the 1FGL catalog, while the flux is allowed to vary, except for PSR J1907+06 (Abdo et al. 2010b), SNR W51C (Abdo et al. 2009b), and SNR W44 (Abdo et al. 2010c) which are 3°, 6°, and 9° away from W49B, respectively. Since these sources around SNR W49B are very bright as evident in Figure 3, we carefully evaluated spectral models for these sources. For this study, we modeled W44 as two point sources at (α, δ) = (28389, 156) and (28410, 115), to approximately account for its angular extent, while the positions of the other sources are fixed at the values determined by the catalog. The spectral shape of these four bright sources is assumed to be a broken power law since likelihood tests between a power-law function and a broken power-law function favored a broken power-law hypothesis at >10σ (PSR J1907+06), >7σ (SNR W51C), and >15σ (SNR W44) confidence levels. All spectral parameters (flux, spectral break, and spectral indices at low and high energy) are allowed to vary in the fitting since the spectral model is different from that reported in the 1FGL catalog. The Galactic diffuse emission is modeled using "gll_iem_v02.fit." An isotropic component (isotropic_iem_v02.txt) is also included to account for instrumental and extragalactic diffuse backgrounds. Both background models are the standard diffuse emission models released by the LAT team.54 The normalization factors of the Galactic diffuse and the isotropic models are allowed to vary.
Figure 3. LAT count map above 1 GeV around SNR W49B in units of counts per pixel. The pixel size is 01, and Gaussian smoothing is applied with a kernel size of σ = 03. The W49B source, which is located at the center, is clearly visible. The green boxes (10 × 16) represent the regions used for the evaluation of spatial dispersion of the difference between the Galactic diffuse model and the observed distribution. The cyan and white circles represent the regions where the flux of the Galactic diffuse model was varied to evaluate effects of spatial dispersion of the model.
Download figure:
Standard image High-resolution imageIn order to evaluate the validity of spectra for background models, we compared two counts spectra in the 0.2–10 GeV band: the spectrum expected from the models obtained by the above procedure and the observed spectrum. This comparison is performed in a nearby circular region with a radius of 05 centered on Δl ∼ +2° and Δb ∼ 0° from SNR W49B where the Galactic diffuse component is dominant. Figure 4 shows resulting fractional residuals, namely (observed-model)/model, as a function of energy. We fit the residuals with a cubic function as shown in Figure 4, which will be used to estimate the systematic error in flux due to uncertainties of the Galactic diffuse model as discussed in Section 3.3.
Figure 4. Fractional residuals at eight energy bins in 0.2–10 GeV between the observed LAT data and the best-fit Galactic diffuse emission model in the nearby circle centered at (α, δ) = (2884, 102) with radius of 05. The fluxes of all sources included in the fit model except for the Galactic diffuse component are subtracted from the observed data. The blue line shows a cubic function fitted to the residual data.
Download figure:
Standard image High-resolution imageUncertainties of the spatial distribution of the Galactic diffuse emission are evaluated by measuring the dispersion of the fractional residuals in 14 regions, where the Galactic diffuse component is dominant (Figure 3). The regions around four very bright sources, the W49B source, PSR J1907+06, SNR W51C, and SNR W44, are excluded. The fractional residual for each region is calculated in five energy bands: 0.20–0.32 GeV, 0.32–0.50 GeV, 0.50–0.80 GeV, 0.80–1.3, and 1.3–10 GeV. Figure 5 shows the resulting distribution of the fractional residuals for 14 regions in five energy bands. The figure shows that 68% and 90% of the fractional residual are within 4% and 6%, respectively. To be conservative, the fractional residual of 6% will be used to estimate the systematic error in flux due to uncertainties of the Galactic diffuse model below.
Figure 5. Histogram of the fractional residuals between the observed LAT data and the Galactic diffuse model fixed at the best-fit parameters determined by gtlike. The fluxes of all sources included in the fit model except for the Galactic diffuse component are subtracted from the observed data. The residual was calculated in five energy bands (0.20–0.32 GeV, 0.32–0.50 GeV, 0.50–0.80 GeV, 0.80–1.3, 1.3–10 GeV) for each region as shown in Figure 3.
Download figure:
Standard image High-resolution image3.3. Gamma-ray Spectrum of W49B
The spectral energy distribution (SED) of the source associated with W49B is evaluated by dividing the 0.2–200 GeV energy band into 11 energy bins, extracting data inside a circular region with a radius of 10° centered on the W49B source and by using gtlike to obtain a flux value at the center of each bin. In each gtlike run, the W49B source, the other 1FGL sources, Galactic diffuse, and isotropic backgrounds are fitted with their normalization free. The W49B source is fitted with a simple power-law function in each energy bin with its spectral index fixed at 2.2 below 5 GeV and 2.9 above 5 GeV using the fitting result in 0.2–200 GeV (see below), while the indices of the other sources are fixed at the values in the 1 FGL catalog. Note that the obtained flux of the W49B source is insensitive to the choice of the index, if it is fixed in a reasonable range (say, 2–3). Figure 6 shows the resulting SED for the W49B source.
Figure 6. SED of the W49B source measured with the Fermi-LAT. The vertical red and black lines represent statistical errors (1σ) and systematic errors, respectively. The blue line represents the best-fit broken power law from an unbinned likelihood fit in 0.2–200 GeV.
Download figure:
Standard image High-resolution imageIn order to evaluate systematic effects on the SED due to uncertainties of the Galactic diffuse model, we varied the Galactic diffuse model used in the fit. Systematic errors due to uncertainties in the energy spectrum of the Galactic diffuse model are estimated by comparing the fit with and without the modification of the energy distributions of the Galactic diffuse model according to the curve in Figure 4. We did not modify the shape of the energy distribution above 10 GeV. Though the fractional residual intensities in Figure 4 are within our current understanding of the systematic uncertainties in the effective area, these residuals were used conservatively as uncertainties of the Galactic diffuse model. We obtain an estimate of uncertainties as ⩽30% for below 1 GeV, ⩽20% in 1–2 GeV, and ⩽10% above 3 GeV. Systematic errors due to uncertainties of the spatial distribution of the Galactic diffuse model as shown in Figure 5 are estimated, using two modified Galactic diffuse models in which fluxes are varied by 6% in all energy bins for one of two regions with a 3° radius, a disk centered on W49B or an offset disk at (α, δ) = (28518, 451). The resulting systematic errors are estimated to be 45% at 300 MeV, decreasing to 12% at 700 MeV, and ⩽6% above 1 GeV for both cases. We adopt the maximum value among these errors at each energy bin as the systematic error due to the Galactic diffuse model. Other systematic errors include uncertainties of the effective area which are 10% at 100 MeV, decreasing to 5% at 560 MeV, and increasing to 20% at 10 GeV and above. Total systematic errors are set by adding in quadrature the uncertainties due to the Galactic diffuse model and the effective area. The total systematic errors in each energy bin are indicated by black error bars in Figure 6, while statistical errors are indicated by red error bars.
Inspection of Figure 6 suggests a spectrum steepening above a few GeV. We performed a likelihood-ratio test between a power-law (the null hypothesis) and a smoothly broken power-law functions (the alternative hypothesis) for 0.2–200 GeV data inside a circular region with a radius of 10° centered on the W49B source. The smoothly broken power-law function is described as
where photon indices Γ1 below the break, Γ2 above the break, a break energy Ebreak and a normalization factor K are free parameters. The parameter β is fixed at 0.05. The simple broken power-law function is not adopted here, since the function cannot be differentiated at the break energy resulting in unstable fit results and inaccurate error estimates. We obtained a test statistics of TSBPL = −2ln(LPL/LBPL) = 22.9, which means a simple power law can be rejected at a significance of 4.4σ. The parameters obtained with the broken power-law model are photon indices Γ1 = 2.18 ± 0.04, Γ2 = 2.9 ± 0.2, and Ebreak = 4.8 ± 1.6 GeV, with an integrated flux in 0.2–200 GeV of (1.74 ± 0.06) × 10−7 photon cm−2 s−1, while the photon index obtained with the simple power law is 2.29 ± 0.02. The gamma-ray luminosity in 0.2–200 GeV is calculated as 1.5 × 1036(D/8 kpc)2 erg s−1. Figure 7 shows the resulting fit with a broken power-law spectrum to the count spectrum within a radius of 05 around the W49B source location. This underscores the importance of understanding the Galactic diffuse emission for the spectral analyses of the W49B source. We checked if the significance of the spectral break changes for different Galactic diffuse models. We found that TSBPL is 20.0 with the Galactic diffuse models used for evaluating the spatial distribution uncertainties, corresponding to a significance of 4.1σ. TSBPL is 11.8 for the Galactic diffuse model used for evaluating uncertainties of the energy spectrum, corresponding to a significance of 3.0σ. Depending on the chosen Galactic diffuse model, the significance of the break ranges between 3 and 4.4σ.
Figure 7. Count spectrum within a radius of 05 around the W49B source location at (α, δ) = (287756, 9096). The blue, red, green, and black lines show best-fit model curves for the W49B source, the Galactic diffuse emission model, the sum of the other sources (all sources except for the W49B source and the Galactic diffuse model), and the sum of all sources included in the fit model, respectively.
Download figure:
Standard image High-resolution image3.4. Upper Limit on W49A
W49A (G43.0+0.0) is one of the most active and luminous star-forming regions (∼107L☉) in the Galaxy (Conti & Blum 2002), located 021 to the west of SNR W49B as shown in Figure 1. Its distance is estimated to be 11.4 ± 1.2 kpc (Gwinn et al. 1992).
In this analysis, we find no gamma-ray counterpart for W49A. An upper limit to the GeV flux from W49A is determined by performing gtlike analysis. The model used for the fit includes W49A, the W49B source, all other 1FGL sources, Galactic diffuse, and isotropic backgrounds. The W49A source is assumed to have uniform surface brightness inside a circle with radius 5'. A simple power-law function with its photon index fixed at 2.0 or 2.5 is used to model the W49A spectrum. The upper limits on the flux (0.2–200 GeV) obtained from the fits with the indices fixed at 2.0 and 2.5 are 9.5 × 10−9 photon cm−2 s−1 and 3.4 × 10−8 photon cm−2 s−1 at 95% confidence level, corresponding to luminosity limits of <3 × 1035(D/11.4 kpc)2 erg s−1 and <4.9 × 1035(D/11.4 kpc)2 erg s−1, respectively. The uncertainties due to the Galactic diffuse model as discussed in Section 3.2 have little effect on the upper limit in the case of the photon index 2.0, while those increase the upper limit to <7.8 × 1035(D/11.4 kpc)2 erg s−1 in the case of the photon index 2.5.
4. DISCUSSION
4.1. Pulsar?
The gamma-ray emission positionally coincident with SNR W49B is unresolved with the LAT. This is reasonable given the fact that the angular extent of SNR W49B is somewhat smaller than the effective LAT PSF. The extent of GeV gamma-ray emission from middle-aged SNRs W51C (Abdo et al. 2009b) and W44 (Abdo et al. 2010c) made it possible to attribute the observed gamma-ray signals to the shells of these SNRs. Since this is not possible with W49B, we will examine a possibility that a pulsar's magnetosphere is responsible for the observed gamma-ray emission even though no pulsed emission has been detected with the LAT. In addition, no radio pulsars are found within a radius of 04 around the LAT position of the W49B source in the ATNF catalog, while the LAT position is determined with 0024 at 95% confidence level. Note that we do not consider a PWN here, since the observed gamma-ray flux is very difficult to be accounted for by a radio-quiet PWN.
To compare the spectral shape of the W49B source with that of typical LAT pulsars in the first pulsar catalog (Abdo et al. 2010e), we fit the LAT spectrum of the W49B source by a power law with an exponential cutoff:
where photon index Γ, a cutoff energy Ecutoff, and a normalization factor K are free parameters. The parameters of the W49B source obtained by gtlike are Γ = 2.10 ± 0.02 and Ecutoff = 15 ± 1 GeV. We performed a likelihood-ratio test between a power law (the null hypothesis) and a cutoff power law (the alternative hypothesis) and obtained test statistics of TScutoff=-2ln(LPL/Lcutoff) = 27, which means that we can reject a simple power law at a significance of ∼5σ. About 90% of the 46 LAT pulsars in the catalog (Abdo et al. 2010e) have Γ < 1.9 and Ecutoff < 5.0 GeV. No pulsar exhibits Ecutoff>6.5 GeV among the LAT pulsars that have an error on Ecutoff less than 4 GeV. The LAT spectrum of W49B is different from what has been obtained for almost all gamma-ray pulsars so far.
A pulsar may have eluded detection in X-rays due to the presence of bright X-ray emission from shock-heated plasmas. Using 55 ks of Chandra data (PI: S. S. Holt) we put an upper limit on the X-ray flux of a possible hidden pulsar of FX (2–10 keV) <6.5 × 10−14 erg s−1 cm−2 on the assumption that the pulsar spectrum is a power law with a photon index of 2.0. The foreground column density NH used here is 6 × 1022 cm−2. This corresponds to an upper limit on the X-ray luminosity of LX (2–10 keV) <5 × 1032(D/8 kpc)2 erg s−1. The empirical correlation of the X-ray and spin-down luminosity of rotation-powered pulsars can be written as
where LX and Lsd are the X-ray luminosity in the 2–10 keV and the spin-down energy loss in units of erg s−1, respectively (Possenti et al. 2002). This relation constrains the spin down luminosity of any undetected pulsars in W49B to be Lsd < 1 × 1036(D/8 kpc)2 erg s-1. However, the gamma-ray luminosity (0.2–200 GeV) of the W49B source is 1.5 × 1036(D/8 kpc)2 erg s-1, which exceeds Lsd. Together with the spectral argument, we conclude that the gamma-ray emission in the direction of W49B is unlikely to come from a pulsar.
4.2. Gamma Rays from the SNR Shell
Here we consider a scenario in which the gamma-ray source originates in the radio-emitting shell of SNR W49B. This scenario is supported by the best-fit LAT position being coincident with the brightest part of synchrotron radio emission as shown in Figure 8. The near-infrared [Fe ii] emission, arising from warm ionized gas with a density of order 1000 cm−3, correlates well with the synchrotron map (Keohane et al. 2007).
Figure 8. LAT source position at a 95% confidence level (a magenta circle) is superposed on the Spitzer IRAC 5.8 μm image. Contours show 20 cm radio intensity obtained from MAGPIS (Helfand et al. 2006).
Download figure:
Standard image High-resolution imageWe assume that the particles responsible for the LAT flux are distributed in a radio-emitting zone which can be characterized by a constant hydrogen density nH and magnetic field strength B. The volume of the emission zone is written as V = f(4π/3)R3, where f ⩽ 1 denotes a filling factor and R = 4.4 pc is the radius of the remnant. The radio-emitting material would originate in swept-up stellar wind and/or interstellar gas. We adopt the total mass contained in the zone as MH = 50 M☉, which would be valid within a factor of few. We then consider three cases: (1) nH = 10 cm−3 and f = 0.6; (2) nH = 100 cm−3 and f = 0.06; (3) nH = 1000 cm−3 and f = 0.006. Note that the constant product of fnH implies the fixed mass in the gamma-ray-emitting region. Case (1) is considered for a reference purpose, even though it would hardly explain the similarity between the synchrotron and the [Fe ii] images. This set of parameters is more appropriate for the X-ray-emitting gas, whose density is estimated as n ∼ 5–8 cm−3 (Miceli et al. 2006).
We adopt the following form as injection distributions of protons and electrons (Abdo et al. 2009b):
where p0 = 1 GeV c−1. The indices and the break momentum are set to be common between electrons and protons. The radio synchrotron index α = 0.48 (Green 1988) corresponds to s ≃ 2. The kinetic equation for the momentum distribution of high-energy particles in the shell can be written as
where be,p = −dp/dt is the momentum loss rate, and Qe,p(p) (assumed to be time-independent) is the particle injection rate. To obtain the radiation spectra from the remnant, Ne,p(p, T0) is numerically calculated for T0 = 2000 yr. Note that energy loss processes such as ionization/Coulomb and synchrotron losses are generally not fast enough to modify the gamma-ray spectrum in the LAT band. The gamma-ray emission mechanisms include the π0-decay gamma rays due to high-energy protons, and bremsstrahlung and IC scattering processes by high-energy electrons. Calculations of the gamma-ray emission were done using the method described in Abdo et al. (2009b). The large gamma-ray luminosity of Lγ ∼ 1 × 1036 erg s−1 precludes IC scattering as a dominant contributor to the gamma-ray emission as discussed in Abdo et al. (2009b). Specifically, the total energy required in electrons would be unrealistically large We = ∫(γ-1)mec2Nedp ∼ 1 × 1051 erg. We shall consider the π0-decay and electron bremsstrahlung models to account for the observed gamma-ray spectrum.
The SED of SNR W49B in the radio and gamma-ray bands is shown in Figure 9, together with the π0-decay emission models. The radio data are modeled by the synchrotron radiation. We construct the π0-decay emission models for the different values of nH = 10, 100, and1000 cm−3 (Table 1). Leptonic components (synchrotron, bremsstrahlung, and IC) are calculated assuming ae/ap = 0.01, a value similar to what is observed for cosmic rays at GeV energies. Note that contributions of the secondary electrons and positrons produced in pp collisions are small for the sets of parameters that we adopted (Table1; see also Abdo et al. 2010c). The secondary synchrotron spectrum is shown in Figure 9 (a3), where its flux is about 10% of the total synchrotron flux at 1 GHz for nH = 1000 cm−3. The contribution of the secondaries to the gamma-ray emission is also small, about 10% of the electron bremsstrahlung components for nH = 1000 cm−3.
Figure 9. SEDs of W49B with model curves for three cases. Cases (a1), (a2), and (a3) represent nH = 10, 100, and1000 cm−3, respectively (see Table 1). The gamma-ray emission is assumed to be dominated by π0-decay. The radio emission (Moffett & Reynolds 1994) is explained by synchrotron radiation from primary and secondary electrons. The dashed line in the radio band represents the synchrotron emission from the secondary electrons. The gamma-ray emission is modeled with a combination of π0-decay (dashed line), bremsstrahlung (dot-dashed line), and IC scattering (dotted line).
Download figure:
Standard image High-resolution imageTable 1. Parameters of Multiwavelength Models
| Model | Parameters | Energetics | |||||||
|---|---|---|---|---|---|---|---|---|---|
| ae/ap | Δs | pbr | B | nH | f | (a)Wp or (b)We | (a)Up or (b)Ue | UB | |
| (GeV c−1) | (μG) | (cm−3) | (1050 erg) | (eV cm−3) | (eV cm−3) | ||||
| (Case a1) π0-decay | 0.01 | 0.7 | 4 | 15 | 10 | 0.6 | 11 | 1.1 × 105 | 5.6 |
| (Case a2) π0-decay | 0.01 | 0.7 | 4 | 60 | 100 | 0.06 | 1.1 | 1.1 × 105 | 90 |
| (Case a3) π0-decay | 0.01 | 0.7 | 4 | 240 | 1000 | 0.006 | 0.10 | 1.0 × 105 | 1400 |
| (Case b1) Bremsstrahlung | 1.0 | 1.0 | 4 | 5 | 10 | 0.6 | 2.6 | 2.6 × 104 | 0.62 |
| (Case b2) Bremsstrahlung | 1.0 | 1.0 | 4 | 20 | 100 | 0.06 | 0.23 | 2.3 × 104 | 10 |
| (Case b3) Bremsstrahlung | 1.0 | 1.0 | 4 | 80 | 1000 | 0.006 | 0.016 | 1.6 × 104 | 160 |
Notes. Seed photons for IC include IR (kTIR = 3 × 10−3 eV, UIR = 1 eV cm−3), optical (kTopt = 0.25 eV, Uopt = 1 eV cm−3), and the CMB. The total energy, We,p and energy density, Ue,p of radiating particles are calculated for p>10 MeV c−1.
Download table as: ASCIITypeset image
The product of nH and Wp remains almost constant irrespective of nH: nHWp ≃ 10 × 1051 erg cm−3. We obtain B ≃ 240 μG in the case of nH = 1000 cm−3. The SED itself can be formally explained in all the cases. The energy density of relativistic protons amounts to Up ≃ 1 × 105 eVcm−3. This value is much higher than Up ∼ 100 eVcm−3 calculated for π0-decay-dominant modeling of middle-aged SNR W51C (Abdo et al. 2009b).
In Figure 10, the gamma-ray spectrum is modeled formally by relativistic bremsstrahlung of electrons. The less luminous π0-decay component is also plotted using ae/ap = 1. It is shown that nH ≳ 100 cm−3 is required to reproduce the radio spectrum. If relativistic bremsstrahlung is responsible for the gamma rays, the ratio of the energy density of relativistic electrons to that of magnetic fields becomes very high, Ue/UB ≳ 100 (Table 1). The energy density required (≃2 × 104 eVcm−3) is also much higher than Ue ∼ 20 eVcm−3 calculated for bremsstrahlung dominant modeling of W51C.
Figure 10. Same as Figure 9 but the gamma-ray emission is assumed to be dominated by electron bremsstrahlung. Cases (b1), (b2), and (b3) represent nH = 10, 100, and1000 cm−3, respectively (see Table 1).
Download figure:
Standard image High-resolution image
5. CONCLUSIONS
We have studied gamma-ray emission in the direction of SNR W49B using about 17 months of data accumulated by the Fermi-LAT. The observed energy spectrum in 0.2–200 GeV exhibits steepening toward high energies, although a simple power-law function cannot be completely ruled out given the uncertainties of the Galactic diffuse model. The luminosity is estimated to be 1.5 × 1036(D/8 kpc)2 erg s−1, which makes this source one of the most luminous gamma-ray sources in the Galaxy.
The gamma-ray source is unresolved by the LAT, which is consistent with the angular size of SNR W49B (∼4' in diameter) taking into account the effective LAT PSF. Assuming a point source, the source position is found to be (α, δ) = (287756, 9096) with an error radius of 0024 at 95% confidence level. This result clearly shows that the gamma-ray emission comes predominantly from the SNR W49B region, not from a nearby star-forming region, W49A. We put an upper limit on the gamma-ray luminosity of W49A as <3 × 1035(D/11.4 kpc)2 erg s−1 at 95% confidence level. The gamma-ray emission in the direction of SNR W49B is unlikely to come from a pulsar. The gamma-ray energy distribution is different from that observed for other pulsars with the LAT. In addition, no pulsed emission has been detected with the LAT nor are any radio pulsars known in this direction.
A good match between the best-fit LAT position and the brightest part of synchrotron radio emission suggests that the gamma-ray source originates in the shell of SNR W49B. The LAT spectrum can be formally explained either by π0-decay gamma rays or by electron bremsstrahlung. For both cases, the calculated energy density of relativistic particles is evaluated to be very high, Ue,p>104 eV cm−3.
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.
Footnotes
- 53
- 54
Available at the Fermi Science Support Center (http://fermi.gsfc.nasa.gov/ssc).
- 55
The IR data are available from NASA/IPAC Infrared Science Archive (http://irsa.ipac.caltech.edu/data/SPITZER/GLIMPSE).