THE PLERIONIC SUPERNOVA REMNANT G21.5−0.9 POWERED BY PSR J1833−1034: NEW SPECTROSCOPIC AND IMAGING RESULTS REVEALED WITH THE CHANDRA X-RAY OBSERVATORY

The combined Chandra image in Figure 1 shows the structure of G21.5−0.9. Located at α(2000) = 18^h33^m3354, δ(2000) = −10°34'076 is a point source corresponding to the location of the pulsar PSR J1833−1034. The PWN is approximately 40'' in radius and is seen to have indentations in the northwest and southeast. As well, many filamentary structures can be seen in the PWN. The X-ray halo extends to a radius of 153'' with limb brightening observed along the eastern edge. The foreground source SS 397 in the southwest portion of the halo was removed from the data prior to any spectral analysis. The northern portion of the halo is dominated by bright knots which appear to merge with the limb in the northeast. The brightest knot is located north and slightly to the west of the PWN. As mentioned earlier, this will be referred to as the "northern knot" throughout the paper and corresponds to the "North Spur" studied by Bocchino et al. (2005). The open questions on these regions are studied further in Section 3.

Figure 2 zooms on the PWN and compares the Chandra data with radio data. Despite targeted searches for the SNR shell in the radio, the X-ray limb and halo have not yet been detected at radio wavelengths, except for the northern knot (Bietenholz et al. 2010). In Figure 2(a), the contours from the X-ray image are overlaid on radio data taken with the Nobeyama Millimeter-Wave Array at 22.3 GHz (Fürst et al. 1988). Figure 2(b) combines the 0.2–10.0 ACIS X-ray data (blue) with the 22.3 GHz radio data (red) to show the similarity of the structure at both wavelengths. In Figure 2(c), 4.75 GHz radio data from the VLA (see Bietenholz & Bartel 2008 for details) are colored red and the 0.2–10.0 keV X-ray data are again colored blue. The X-ray and radio structure are remarkably similar, with the radio following the shape of the X-ray PWN along the edges of the PWN, including indentations in the northwest and southeast. The PWN appears slightly larger at 4.75 GHz than at 22.3 GHz, but the images at 22.3 GHz and 0.2–10 keV are comparable in size, which indicates a small magnetic field in the PWN. Indeed, a low magnetic field estimate (B ∼ 25 μG; see also Section 5.2) has been inferred from modeling G21.5−0.9 (de Jager et al. 2008). The most prominent difference between the radio and X-ray emission is in the center of the PWN. The pulsar is seen in a location that peaks in X-rays but has a minimum in radio emission. The X-ray emission traces the particles freshly injected by the pulsar, whereas the radio emission traces the older population characterized by a much larger synchrotron lifetime. The location of minimum radio emission could be indicative of a magnetic field direction along the line of sight, since the synchrotron emissivity scales as ∝ B^α+1_⊥ (where B_⊥ is the magnetic field's component that is perpendicular to the line of sight and α is the spectral index). This interpretation is consistent with the radial magnetic field distribution inferred from polarization studies (Fürst et al. 1988).

Zoom In Zoom Out Reset image size

For each region shown in Figure 1, weighted spectra were extracted for every observation using the CIAO tools dmcopy and dmextract. The background chosen for each region is described in the section specific to that region.

To compensate for the effects of cosmic radiation damage, a CTI correction was applied to the ACIS data (CIAO CTI correction⁵ for ACIS-I, −120°C data, Penn State CTI correction⁶ (Townsley et al. 2000) for ACIS-I, −110°C and ACIS-S, −110°C and −120°C data). The −100°C data cannot be corrected for CTI and were therefore omitted from the spectral analysis.

For the observations that were CTI-corrected using the Penn State CTI corrector, RMF files were provided with the corrector. For the observations that were CTI corrected with CIAO (acis_process_events), the tool mkacisrmf was used to create weighted RMF files.

The regions used in the following spectral analysis are listed in Table 2 and shown in Figure 1. The spectra were individually binned using the FTOOL⁷ grppha to improve the signal-to-noise ratio (S/N). The minimum number of counts per bin for the various regions are shown in Tables 3 and 4. All spectral models used contain a component which accounts for absorption along the line of sight (wabs in XSPEC). A power law was used to model synchrotron emission from high energy electrons in a magnetic field. A blackbody model was used to study any thermal emission from the neutron star directly. A pshock model (a plane-parallel non-equilibrium ionization model with different ionization ages and a constant electron temperature; Borkowski et al. 2001) was used to search for thermal emission from shock-heated ejecta or interstellar matter. The pshock model is characterized by the ionization timescale, τ = n_et, where n_e is the post-shock electron density and t is the time since the passage of the shock. The vpshock model is a pshock model which also accounts for variable abundances of metals. The srcut model was used to model the non-thermal component of the SNR associated with electrons accelerated by the SNR shock (see, e.g., Reynolds & Keohane 1999). In order to use complete response information included with each observation, all spectra for a particular region were fit simultaneously. All errors are reported to the 90% confidence level.

Only observations with an off-axis angle less than 3' were used to study the compact source (Table 1). For an off-axis angle of 3', at 1.5 keV, 90% of the energy of a point source is contained within 15 (within 25 at 6.4 keV). Therefore, a circular region with 2'' radius, centered at α(2000) = 18^h33^m3354, δ(2000) = −10°34'076, was defined as the extraction region for each data set. The spectra were each grouped to have a minimum of 50 counts bin⁻¹ and cover the energy range 0.5–8.0 keV. To remove contamination from the PWN, the background chosen was an annulus centered on PSR J1833−1034 with radius 2''–4''.

The best fit using an absorbed power-law model yields a column density of N_H = 2.24^+0.09_−0.10 × 10²² cm⁻², a photon index of Γ = 1.47^+0.05_−0.06, and an observed flux in the 0.5–8.0 keV band of (3.2 ± 0.3) × 10⁻¹² erg cm⁻² s⁻¹ (χ² = 1287.9 and ν = 1047 degrees of freedom; see Table 3 for a list of parameters). This column density is consistent with the previous work on G21.5−0.9 and with our global fit to the PWN and will be used in the remainder of the spectral analysis. Fitting with an absorbed blackbody alone gives a low column density of N_H = 0.78^+0.05_−0.06 × 10²² cm⁻² and a temperature of kT_bb = 1.34^+0.03_−0.03 keV = 1.55^+0.03_−0.03 × 10⁷ K (reduced χ²_ν = 1.51 and ν = 1047). Freezing the column density to the acceptable value of 2.24 × 10²² cm⁻² gives an unacceptable fit with a temperature of kT_bb ∼ 1.0 keV and χ²_ν ∼ 2.6 for a single component blackbody. As expected, the blackbody fit alone is poor at low and high energies, indicating the need for a non-thermal component.

To test for a combination of thermal + magnetospheric emission, we freeze the hydrogen column density to N_H = 2.24 × 10²² cm⁻² and fit the pulsar's emission with an absorbed power-law+blackbody model (Figure 3). The best fit (χ² = 1271.6, ν = 1046) was obtained for a hard photon index of Γ = 1.14^+0.05_−0.07 and a temperature of kT_bb = 0.52^+0.03_−0.04 keV = 6.0^+0.3_−0.5 × 10⁶ K. This is an improved fit over the power-law fit alone, with an F-test probability of ∼2.6 × 10⁻⁴. Additional observations of the pulsar will help confirm or constrain this emission.

Zoom In Zoom Out Reset image size

The additional blackbody component suggests the first evidence for thermal emission from the pulsar, a result that is observed in other young neutron stars. The observed thermal flux ((2.6 ± 1.0) × 10⁻¹³ erg cm⁻² s⁻¹) was found to be ∼9% of the non-thermal flux ((3.0 ± 0.6) × 10⁻¹² erg cm⁻² s⁻¹) in the 0.5–8.0 keV range. Converting the unabsorbed thermal flux ((7.4 ± 3.0) × 10⁻¹³ erg cm⁻² s⁻¹, ∼16% of unabsorbed non-thermal flux) to luminosity and assuming a blackbody, we can derive the size of the emitting area using the blackbody formula: L = 4πR²σT⁴_bb = 4πD²F (where σ is the Stephan–Boltzmann constant and F is the observed flux). The radius of the emitting region is found to be R = 0.49 ± 0.15 km for a temperature of 0.52 keV. This is an unreasonably small radius for the size of a neutron star, suggesting that we may instead be observing emission from a small "hot spot" on the surface of the neutron star. Detecting the X-ray pulsations is needed to confirm this interpretation.⁸ Alternatively, constraining the radius of the emitting region to 12 km and assuming a distance of 5 kpc to G21.5−0.9, we use the bbodyrad model to then calculate the surface temperature of the neutron star, assuming the thermal component is from the entire surface. In this case, the power-law model parameters are the same as those for the power-law model alone (with a photon index Γ = 1.47) but we derive a lower temperature kT = 0.11 (<0.14) keV, corresponding to an effective temperature of 1.28 (<1.57) × 10⁶ K which is closer to what has been observed in other young neutron stars. This result is further discussed in Section 5.2.

Since studying the emission from the limb requires subtracting the emission from the halo (to remove the contamination by the dust scattering halo), here we briefly summarize our spectral fit to the X-ray halo.

The X-ray halo of G21.5−0.9 is visible in Figure 1 at a radius of 45''–153'' from the location of PSR J1833−1034. Since the northern half of the halo is dominated by emission from the knots, the southern half of the X-ray halo was selected for study. Emission from the eastern limb (defined between 125'' and 153'', see Section 3.4) and SS 397 was removed from the data prior to extracting the spectra. The background used was the southern half of an annulus with radius 153''–175''. The halo was fit with a power law with the column density fixed at the best-fit value from PSR J1833−1034 (N_H = 2.24^+0.09_−0.10 × 10²² cm⁻²). We find a photon index of Γ = 2.50^+0.05_−0.05, χ²_ν = 1.08, and ν = 865 (Figure 4). The fit is improved by the addition of a thermal pshock component with a temperature of kT = 0.33^+0.12_−0.08 keV and an ionization timescale of n_et = 1.0^+12.5_−1.0× 10⁸ cm⁻³ s (Γ = 2.25^+0.11_−0.18, χ²_ν = 1.04, ν = 862). This fit to the halo is used in Section 3.4 to subtract the halo component from the limb region.

Zoom In Zoom Out Reset image size

The eastern limb (Figure 1, radius = 125''–153''⁹) was studied to search for emission characteristic of an SNR shell. Due to the lower count rate in the limb, the background spectrum was extracted from a region outside the halo and therefore the data contain a component due to the halo. To correct for the halo emission, we add a model component equal to the best fit to the halo (Section 3.3), with the normalization scaled to the area of the limb. The parameters for this component are all frozen and the fits presented below are emission from the limb only.

The pshock model was first used to search for thermal emission from interstellar matter shock-heated by the forward shock. As shown in Table 4, the pshock model provides an adequate fit (χ²_ν = 0.538, ν = 909); however, with an unrealistically high temperature (7.5^+6.9_−2.5 keV), suggesting that the X-ray emission is likely non-thermal.

The limb of G21.5−0.9 cannot be explained by a dust-scattering halo nor by shock-heated ejecta (e.g., Bocchino et al. 2005) and must have another origin. Shocks in young SNRs have been known to accelerate electrons to TeV energies where they produce synchrotron X-rays (Reynolds 1998). To determine if the non-thermal component of the limb is due to this acceleration, we used the srcut model in XSPEC. Since the limb has not yet been observed in the radio (Bietenholz et al. 2010), we do not know the radio spectral index (α, where S ∝ ν^−α) for the limb and consider a range for α between 0.3 and 0.8, which covers the range observed for other SNRs (Green 2009). Similarly, without a radio observation of the shell, we do not know the 1 GHz radio flux density and so we leave it as a free parameter to find an estimate of the radio flux density. For α = 0.5, which is typical for SNR shells, and N_H = 2.24 × 10²² cm⁻², we find a rolloff frequency of ν_rolloff = 4.3^+9.2_−2.2 × 10¹⁷ Hz, and a 1 GHz flux density of 4.9^+0.7_−0.8 × 10⁻³ Jy, corresponding to a surface brightness of 2.3^+0.3_−0.4 × 10⁻²² W m⁻² Hz⁻¹ sr⁻¹ for a solid angle Ω = 2.1 × 10⁻⁷ sr (χ²_ν = 0.541, ν = 910; Figure 5). This estimate of the surface brightness is a factor of ∼ 17 times smaller than the upper limit obtained by Slane et al. (2000) and a factor of ∼3 times smaller than the most recent upper limit of 7 × 10⁻²² W m⁻² Hz⁻¹ sr⁻¹ obtained by Bietenholz et al. (2010) using a new sensitive 1.43 GHz observation with the VLA.

Zoom In Zoom Out Reset image size

For α = 0.3 and N_H = 2.24 × 10²² cm⁻² with the model srcut, we find a rolloff frequency of ν_rolloff = 2.2^+2.5_−1.0 × 10¹⁷ Hz and a 1 GHz flux density of 1.4^+0.2_−0.3 × 10⁻⁴ Jy, corresponding to a surface brightness of 6.7^+0.9_−1.4 × 10⁻²⁴ W m⁻² Hz⁻¹ sr⁻¹ (χ²_ν = 0.542, ν = 910), approximately two orders of magnitude smaller than the observed recent upper limit (Bietenholz et al. 2010). For α = 0.8 and N_H = 2.24 × 10²² cm⁻² with the model srcut, we find a rolloff frequency of ν_rolloff = 8.7⁺²⁴⁰⁰_−3.3 × 10¹⁷ Hz and a 1 GHz flux density of 1.4^+0.3_−0.2 Jy, corresponding to a surface brightness of 6.7^+1.4_−1.0 × 10⁻²⁰ W m⁻² Hz⁻¹ sr⁻¹ (χ²_ν = 0.542, ν = 910). This estimated surface brightness is 2 orders of magnitude higher than the recent upper limit by Bietenholz et al. (2010), suggesting that the spectral index is at the lower end of the 0.3–0.8 range.

The quality of fit for each of the power-law, pshock, and srcut models is similar, however, the temperature for the thermal model is too high, confirming that the limb is dominated by non-thermal emission and strengthening the previous suggestion that it is a possible site for cosmic-ray acceleration (Bocchino et al. 2005; see Section 5). A two-component (thermal + non-thermal) model does not give well-constrained parameters and so we have not included the results here.

Bocchino et al. (2005) examined the brightest knot of emission to the north of the PWN ("North Spur"). Using a two-component power law plus thermal component (with the latter modeled by the vnei model, a non-equilibrium ionization model with a single ionization age and temperature, and with variable metal abundances), they found two solutions for the metal abundances with a similar quality of fit. The first solution is characterized by enhanced abundances for Mg and Si, with a Mg abundance of 0.6–3 times solar, an Si abundance of 2–20 times solar, a temperature of kT ∼ 0.17 keV, and an ionization timescale of n_et ∼ 7 × 10¹¹ cm⁻³ s. The second solution is characterized by solar abundances, a temperature of kT ∼ 0.30 keV, and a lower ionization timescale of n_et ∼ 1 × 10¹⁰ cm⁻³ s. Here, we explore the same region with more data in an attempt to better constrain the parameters.

Before attempting two-component models, however, we first fit with single component models: a power-law or srcut model for a non-thermal interpretation and a pshock¹⁰ model for a thermal interpretation and found the following. An absorbed power-law fit to the brightest knot in the north (Figure 1, background inside the halo) produces an artificially low hydrogen column density (N_H = 0.98^+0.15_−0.15 × 10²² cm⁻²) with χ²_ν = 0.98 (ν = 605). Freezing N_H to the best-fit value for PSR J1833−1034 (2.24 × 10²² cm⁻²) gives a photon index of Γ = 2.72^+0.09_−0.09, an unabsorbed flux of 6.3 × 10⁻¹³ erg cm⁻² s⁻¹, and an estimated X-ray luminosity (0.5–8.0 keV) of 1.9 × 10³³ erg s⁻¹ (χ²_ν = 1.18, ν = 606, assumed D = 5 kpc) for the knot. While the fit is statistically acceptable, the residuals show evidence of line emission and the fit is improved by adding a thermal pshock component, as discussed below. Similarly, an srcut model alone is not favored based on the evidence of thermal emission lines in the spectrum.

As shown in Table 4, the pshock model alone with a column density N_H = 2.24 × 10²² cm⁻² gives a temperature kT = 4.9^+0.9_−0.8 keV and an ionization timescale n_et = 2.0^+0.6_−0.4 × 10¹⁰ cm⁻³ s with χ²_ν = 0.959 (ν = 605). Allowing the abundances of Mg, Si, and S to vary, we find that Mg = 1.08 (0.95–1.21), Si = 0.96 (0.79–1.13), and S = 0.52 (0.12–0.93); all consistent with solar abundances. While the fit is statistically acceptable, the shock temperature is unrealistically high, suggesting that the spectrum is dominated by non-thermal emission.

We therefore confirm the need for a two-component, thermal + non-thermal, model to fit the northern knot. As well we rule out an equilibrium ionization model (such as MEKAL used by Bocchino et al. 2005) for the thermal component, since the fit requires a low ionization timescale (<10¹² cm⁻³ s). Next, we explore the two solutions discussed in Bocchino et al. (2005).

Fitting a two-component model (power law+pshock) to the northern knot (N_H = 2.24 × 10²² cm⁻²), we find a photon index of Γ = 2.21^+0.15_−0.15, a temperature of kT = 0.20^+0.04_−0.06 keV, and an ionization timescale of n_et = 7.1^+7.1_−3.4 × 10⁹ cm⁻³ s (χ²_ν = 0.96, ν = 603, Figure 6). The observed thermal flux is (9.3 ± 3.2) × 10⁻¹⁵ erg cm⁻² s⁻¹, which is only ∼6% that of the non-thermal flux ((1.6 ± 0.3) × 10⁻¹³ erg cm⁻² s⁻¹) in the 0.5–8.0 keV range, again confirming that the spectrum of the northern knot is dominated by non-thermal emission. Figures 7 and 8 show the range of values allowed by the above fit for the parameters of the pshock component. Figure 7 shows the relationship between kT and n_et, and Figure 8 shows the relationship between the emission measure and n_et. Allowing the abundances of Mg, Si, and S to vary, we find that Mg = 0.72 (0.40–1.06), Si = 0.84 (0.32–1.35), and S = 107.1 (3.9–210.3). Figure 9 shows that Mg and Si are consistent with solar, in agreement with solution 2 of Bocchino et al. (2005). We do not observe a solution with overabundances of Mg and Si. That is we rule out solution 1 of Bocchino et al. (2005). However, we note that S (although poorly constrained) appears overabundant in our fits. While additional observations will help constrain the S abundance, we note that the apparent high S abundance might be an artifact of the model used, because the S lies in the energy range where X-ray spectra from the pshock and power-law components overlap.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Finally, a two-component srcut+pshock model was used to further study the non-thermal emission from the northern knot that could be due to particle acceleration at a shock. We fix the column density, N_H, to 2.24 × 10²² cm⁻² and the radio spectral index, α, to 0.5 (as for the limb, we also attempted other values for α in the range that brackets all possible spectral indices—see the results summarized in Table 4). We find a temperature of kT = 0.21^+0.04_−0.04 keV, an ionization timescale of n_et = 5.8^+7.1_−1.8 × 10⁹ cm⁻³ s, a rolloff frequency of ν_rolloff = 4.3^+4.1_−2.1 × 10¹⁸ Hz, and a 1 GHz flux density of 2.6^+1.5_−0.9 × 10⁻³ Jy (χ²_ν = 0.963, ν = 603). The thermal flux (1.0^+1.5_−0.5 × 10⁻¹⁴ erg cm⁻² s⁻¹) is again ∼7% of the non-thermal flux (1.5^+0.9_−0.5 × 10⁻¹³ erg cm⁻² s⁻¹) in the 0.5–8.0 keV range. Varying the abundances again yields Mg and Si abundances that are consistent with solar, but indicates enhanced (and poorly constrained) abundance for S; a result that is consistent with the power-law+pshock model above. We note that the derived 1 GHz flux density (for α = 0.5) is about an order of magnitude smaller than the flux density of 19 ± 7 mJy inferred from the radio detection of the northern knot with the VLA (Bietenholz et al. 2010). Freezing the srcut normalization to the measured value of ∼20 mJy while fitting for the corresponding spectral index yields α = 0.61^+0.03_−0.02, kT = 0.21^+0.03_−0.04 keV, n_et = 6.0^+12.2_−2.6 × 10⁹ cm⁻³ s, and a rolloff frequency ν_roll = 6.9^+11.3_−3.7 × 10¹⁷ Hz (χ²_ν = 0.962, ν = 603). Due to the large uncertainty in the VLA spectral index measurement, our fitted value for α could not be excluded by the radio study and remains to be confirmed.

To conclude, the X-ray emission from the bright northern knot is dominated by non-thermal emission, suggesting cosmic-ray acceleration at a shock. The thermal component does not require enhanced abundances of Mg and Si, suggesting that the second solution of Bocchino et al. (2005) is more reliable. The inferred low temperature, ionization timescale, and solar abundances have been interpreted as evidence for possible interaction between ejecta and the H envelope of a type IIP SN.

The absence of shells around Crab-like SNRs has been a mystery for decades. Recent deep or targeted searches have been however in some cases successful in revealing the long-sought SNR shells around PWNe. Aside from G21.5−0.9, shells have been found in 3C 58 (Gotthelf et al. 2007)¹¹ and most recently G54.1+0.3 (Bocchino et al. 2010).¹² A shell has been unsuccessfully searched for around the Crab Nebula (Frail et al. 1995; Seward et al. 2006). A deep search with Chandra revealed a dust-scattered halo with an intensity of 5% that of the Crab, out to a radial distance of 18'. However, there was no evidence of emission from shock-heated material in the form of a shell. The luminosities for the shells seen in 3C 58 and G21.5−0.9 are below the current upper limit of L_X ∼ 10³⁴ erg s⁻¹ for the Crab Nebula (Seward et al. 2006).¹³ The low luminosity of the shell in G21.5−0.9 (∼2 × 10³³ erg s⁻¹ in the non-thermal interpretation) could be attributed to the fact that the SNR is young and expanding in a low-density medium, and so the shell may only now be coming into view. Furthermore, the heavy absorption toward G21.5−0.9 (N_H = 2.2 × 10²² cm⁻²) explains the difficulty in detecting the SNR shell which became clearly visible (although only the eastern side) after accumulating ∼500 ks of Chandra time. Next, we will comment on the age of the SNR and the ambient density of the interstellar medium (ISM) in which it is expanding.

To estimate the density of the ISM in which the SNR is expanding, we use the pshock fit to the limb¹⁴ which yields an emission measure of EM = 3.7 × 10⁻⁴ cm⁻⁵. This implies that ∫n_en_HdV ∼ fn_en_HV = 10¹⁴(4πD²)(EM) = 1.1 × 10⁵⁶ cm⁻³ (assuming D = 5 kpc), where f is the volume filling factor, n_e is the post-shock electron density, and n_H ∼ n_e/1.2. Using a volume of 3.3 × 10⁵⁶ cm³ (assuming the limb is a shell and the region we studied is ∼10% of the shell volume), n_e ∼ 0.63f^−1/2 cm⁻³, which implies an upstream ambient density n₀ = n_e/4.8 = 0.13 f^−1/2 cm⁻³. For filling factors f ∼ 0.1–1.0, this corresponds to n₀ ∼ 0.13–0.42 cm⁻³ ((2.2–7.1) × 10⁻²⁵ g cm⁻³ when n₀ includes only hydrogen). This estimate of the density is consistent with the upper limit of 0.65 cm⁻³ derived by Bocchino et al. (2005) and confirms that G21.5−0.9 is expanding into a low-density medium.

We subsequently comment on the non-thermal interpretation of the limb as evidence of cosmic-ray acceleration to TeV energies. Young SNRs have been shown to accelerate cosmic rays to TeV energies at shock fronts (e.g., SN1006; Koyama et al. 1995). The energy at which the electron energy distribution steepens from its slope at radio-emitting energies is given by

$$\begin{equation} E_{\mathrm{max}} = 10 \left(\frac{10\,\mu G}{B}\right)^{1/2}\left(\frac{\nu _{\mathrm{rolloff}}}{0.5 \times 10^{16}}\right)^{1/2} \mathrm{TeV} \end{equation} \tag{ 1 }$$

(Reynolds & Keohane 1999). Fitting the limb with the srcut model with a range α = 0.3–0.8, we found that ν_rolloff ∼ 2–9 × 10¹⁷ Hz, which implies E_max(B/(10 μG))^1/2 ∼ 60–130 TeV. This range (60–130 TeV, for a magnetic field of 10 μG) implies that shells in plerionic SNRs could be important sites for cosmic-ray acceleration to TeV energies. We note that the lower values of α give an upper limit on the maximum electron energy which is within the range of values found by Reynolds & Keohane (1999) for young shell-type SNRs. The predicted E_max is dependent on ν_rolloff, which is very dependent on α that remains to be measured in the radio. The magnetic field estimate in the region surrounding the limb is also needed to refine the estimated maximum energies for the accelerated particles.

Finally, we note that the non-thermal component of the northern knot also implies cosmic-ray acceleration at a shock—possibly resulting from the interaction between ejecta and the H envelope of a type IIP SN (see Section 3.5). Using the srcut model component with α = 0.6 (the likely value based on its radio detection, see Section 3.5), the derived rolloff frequency is ν_roll = 6.9 (3.2–18.2) × 10¹⁷ Hz, which yields E_max = 117 (80–191) TeV (again assuming B = 10 μG).

Camilo et al. (2006) show that inside the elliptical emission region containing the pulsar, the compact source is offset to the northeast from the center of the ellipse. The HRC-I image shown in Figure 14 confirms this offset. The image is centered on PSR J1833−1034, smoothed by 2 pixels (02635), and is 9'' × 9''. Using the ACIS imaging data used to produce Figure 1, we measured the distance between the location of the pulsar and the center of the arc traced by the eastern limb. PSR J1833−1034 is found to be offset from the center of the shell of G21.5−0.9 by 75, which is equivalent to a 0.18 pc offset for a distance to G21.5−0.9 of 5 kpc. For a remnant age of 870⁺²⁰⁰₋₁₅₀, this corresponds to a two-dimensional pulsar velocity of 200 ± 40 km s⁻¹. This velocity is comparable with the average two-dimensional speed of 246 ± 22 km s⁻¹ derived for "normal" pulsars by Hobbs et al. (2005).

Zoom In Zoom Out Reset image size

The PWN of G21.5−0.9 demonstrates the type of variability previously observed in other PWNe, including bright knots which appear, move at speeds of ∼0.1c–0.3c, and fade with time. There is no evidence of expansion of the nebula, nor do we expect to observe any with this sequence of observations (Section 4).

Using B_eq ∼ 0.3 mG and $\dot{E} = 3.37 \times 10^{37}$ erg s⁻¹ for G21.5−0.9, Camilo et al. (2006) calculate the characteristic scale of the wind termination shock around PSR J1833−1034 to be ∼ 16, which corresponds to r_t = 1.2 × 10¹⁷ cm = 0.04 pc for a distance of 5 kpc to G21.5−0.9 (Section 1). For an off-axis angle of 3', at 1.5 keV, 90% of the encircled energy of a point source is contained within 25 (within 15 at 6.4 keV) with ACIS. For the HRC data, the 90% encircled energy radius is 09 at 1.5 keV and 20 at 6.4 keV (for an off-axis angle of 03). The estimated termination shock radius is then comparable to the point-spread function, making it difficult to distinguish the termination shock from the pulsar emission. Recent estimates of the magnetic field strength in the nebula predict a field strength which is an order of magnitude below equipartition. de Jager et al. (2008) estimate the current field strength as 25 μG and S. J. Tanaka & F. Takahara (2010, private communication) estimate 64 μG.

In Section 3.2, we presented the spectroscopic analysis of PSR J1833−1034, the pulsar powering G21.5−0.9. The X-ray spectrum is dominated by hard non-thermal emission, with evidence of a thermal blackbody component that represents ∼9% of the non-thermal flux observed from the pulsar in the 0.5–8.0 keV range. While this additional component needs to be confirmed/constrained with additional observations of the pulsar, we comment next on the implication of this detection. Tsuruta et al. (2002) used X-ray observations of neutron stars to compare neutron star cooling theories. They found that less massive stars appear to cool by standard cooling (conventional slower neutrino cooling mechanisms: modified Urca, plasmon neutrino, and bremsstrahlung), while more massive stars cool by non-standard cooling (exotic, extremely fast cooling processes: direct Urca processes involving nucleons, hyperons, pions, kaons, and quarks). For an age of 870⁺²⁰⁰₋₁₅₀ yr (Bietenholz & Bartel 2008) and a luminosity of L_∞ ∼ 1.3^+0.5_−0.5 × 10³³ erg s⁻¹ (blackbody component of the power-law+blackbody fit; see Section 3.2), PSR J1833−1034 falls between the non-standard and standard cooling curves. It is however closer to the non-standard curve than the standard curve, which implies luminosities of L_∞∼ 5 × 10³² erg s⁻¹ and ∼9 × 10³³ erg s⁻¹, respectively, for an 870 year old neutron star. Fixing the radius to 12 km, the luminosity of the bbodyrad component is ∼5.9 × 10³² erg s⁻¹, consistent with the non-standard cooling curve of Tsuruta et al. (2002). Other young neutron stars observed to have thermal components in their spectra also point to rapid cooling by non-standard processes (Slane et al. 2004). The blackbody fit presented here for PSR J1833−1034 yields a temperature that is higher or comparable to that inferred for other young pulsars (T_∞ ∼ 5 × 10⁶ K, reduced to ∼1 × 10⁶ K when fixing the size of the emitting region to that of the neutron star surface). For example, PSR J0205+6449 (∼830 year old based on its probable association with SN 1181) in 3C 58 has an upper limit on its temperature of 1.02 × 10⁶ K (Slane et al. 2004), while the highly magnetized and young pulsar PSR J1119−6127 (with a characteristic age of ∼1600 yr) in G292.2−0.5 has a temperature of (1.2–2.5) × 10⁶ K (Safi-Harb & Kumar 2008).