A CHANDRA OBSERVATION OF SUPERNOVA REMNANT G350.1−0.3 AND ITS CENTRAL COMPACT OBJECT

A Chandra image of G350.1−0.3 is shown in Figure 1(a). We filtered the data on the 0.5–8.0 keV energy band and smoothed the image by applying a Gaussian kernel with a σ of 3 pixels (1.48 arcsec). The source is dominated by an extended bright clump of material in the east with fainter emission extending far to the west. The dashed contour, shown in the figure, surrounds the SNR at a level of 3σ above the background and indicates the apparent extent of the emission. Throughout our analysis, we adopt a source radius of R = 3.3 d_4.5 pc (2.5 arcmin) and assume, following the argument of G08, that d_4.5 = 1. Although the data reveal an elliptical shell-like structure, a large region in the southwest is markedly fainter than the adjacent emission with a flux only a factor of ∼1.5 above the background. The unresolved X-ray source XMMU J172054.5−372652 (indicated by the arrow in Figure 1(a)) is seen far to the west of the bright emission, significantly displaced from the apparent center of the SNR (marked by the white cross in Figure 1(a)). Figure 1(b) shows the detailed structure of the ejecta region and reveals previously unresolved filamentary structures running across its center. We find that the six regions enumerated in Figures 1(a) and (b) represent the range of spectral characteristics found in this remnant. The spectra and the corresponding response files were produced using the specextract script in Ciao 4.2. We subtracted the local background spectrum from a source-free region of the detector, partly shown as the dashed region in the southwest corner of Figure 1(a), and grouped the data to a minimum of 15 counts per bin. We fit each spectrum by an absorbed nonequilibrium ionization (NEI) plane-parallel shock model with variable abundances (XSPEC model "VNEI"; Borkowski et al. 2001). Elements below Mg and above Fe were fixed at solar abundances since their contributions in the fitted bandpass are negligible. The uncertainties were calculated at 90% confidence (1.6σ). Although fixing the absorption at an average global value of ∼4.0 × 10²² cm^-2 produced acceptable fits for regions 1a, 1b, 2, 3, and 4 (the exception is region 1c and is discussed below), we allow N_H to vary in order to account for possible small-scale variations in the absorbing column. The spectra and fitted models for regions 1a, 1b, and 1c are shown in Figure 2 with parameters (for regions 1a and 1b) listed in Table 1. The fits for regions 2–4 are shown in Figure 3, with parameters listed in Table 1.

Zoom In Zoom Out Reset image size

The spectrum of region 1a (extracted from one of the bright filaments in Figure 1(b)) is roughly representative of that found throughout the bright eastern emission. It can be fit by an absorbed VNEI model with column density N_H ≈ 3.8 × 10²² cm^-2, electron temperature kT_e ≈ 1.4 keV, ionization timescale τ ≈ 2.2 × 10¹¹ s cm^-3, and large overabundances of Mg, Si, S, Ca, and Fe. These characteristics unambiguously demonstrate the presence of hot, metal-rich ejecta. The spectrum of region 1b (extracted from the other bright filament) is similar to that of region 1a although the Mg and Si lines clearly indicate a somewhat lower ionization timescale. Although morphologically well defined, we find that the spectral characteristics of the filaments are basically consistent with the surrounding regions.

Although G08 found evidence for variations in elemental abundances between nearby regions within the bright eastern emission, our spectral fits of small regions in this vicinity show no significant variations within uncertainties. In order to compare our results with those of the XMM-Newton observation, we attempted to fit the entire ejecta-dominated emission (region 1c: the same area as fitted by G08) using a VNEI model. As can be seen from the high residuals, this model does not provide a good fit, and hence, we do not display the fit parameters. However, the best-fit parameters are roughly consistent with those of regions 1a and 1b, indicating that the extremely large number of X-ray photon counts (∼100,000) and small error bars likely magnify small deficiencies in the model and result in the poor χ² statistics. In addition, it is possible that small variations in temperature and elemental abundances exist, but are not discernible with the statistics in our spectra from small-scale, spatially resolved regions. Using the two-component model VPSHOCK + RAYMOND (the same as that used by G08—their fit results are summarized in Section 1) for region 1c also fails to provide a good fit. The inconsistencies between our results and those of G08 are likely explained simply by the fact that the long exposure in the Chandra observation exposes weaknesses in the models that were previously hidden due to the fact that the XMM-Newton spectrum has fewer photon counts.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Region 3 is largely representative of the spectral features found along the perimeter of the ejecta region and throughout the remnant's western half. The spectrum can be described by a VNEI model with absorbing column N_H ≈ 4.4 × 10²² cm^-2, electron temperature kT_e ≈ 0.85 keV, ionization timescale τ ≈ 2.7 × 10¹¹ s cm^-3, and near-solar abundances of all metals. These results suggest an interstellar/circumstellar origin. Although otherwise similar, the fit for region 4 indicates a somewhat lower temperature and an ionization timescale of τ ≈ 1.3 × 10¹⁰ s cm^-3, lower than that of region 3 by a factor of 20. The centroid of the Si line in the region 4 spectrum is at a slightly lower energy than that of region 3 and since the goodness of fit statistics are dominated by the Si line (the error bars are smallest in this region), it is this feature that causes the drastic difference in the ionization timescale. Region 2, adjacent to the bright eastern emission, has somewhat enhanced abundances of Mg, Si, and S and a highly enhanced abundance of Ca. Although similar in temperature to region 3, the overabundance of metals suggests the possible presence of ejecta.

The model histogram for the spectrum of region 4 slightly misses the centroid of the S line complex, suggesting that the several bright patches within this region likely vary in their ionization states. Fitting the individual clumps, however, leads to a similar result with the added handicap of poor statistics in the spectra. We thus present the fit to the composite region with the understanding that a deeper exposure is necessary to characterize the spectrum more precisely. The spectral fits to regions 1b, 2, and 4 do not formally correspond to good fits based on the χ² statistics. In the case of region 1b, this deficiency is due to the fact that the model histogram misses the minimum of the Mg line and is too low on the low-energy shoulder of the Si line. In the region 2 spectrum, the fit is too low on the maximum of the Si line, as can be seen from the high residuals in this area. We acknowledge these weaknesses in the models and use them to simply make qualitative statements about the approximate distribution of elemental abundances, temperatures, and ionization states, rather than using them to derive precise numerical results.

A 24 μm infrared image of G350.1−0.3, overlaid with the contours from the X-ray data, is shown in Figure 4(a). The two bright clumps in the east and the south correspond to regions 1c and 4, respectively. Most notably, the upper filament from the X-ray image (region 1b) appears to outline precisely the eastern edge of the infrared emission while the lower filament (region 1a) is notably absent at 24 μm. An enlarged 24 μm image of the ejecta-dominated region, overlaid with the X-ray contours, is shown in Figure 4(b). While the angular resolution at 24 μm is much coarser than that provided by Chandra, these images are sufficiently sensitive to identify two distinct regions of emission with clear morphological similarities to the X-ray emission from the SNR (regions 1c and 4 in Figure 1). Figures 4(c) and (d) show IRAC 8 μm and MIPS 70 μm images of the spatial region corresponding to G350.1−0.3. We do not detect any significant emission from G350.1−0.3 at 8 μm and although we see 70 μm emission in regions corresponding to the SNR, the emission does not differ significantly from regions in the surrounding medium. With the current resolution we cannot determine whether or not some of the 70 μm emission is associated with the SNR. We searched the 3.6 μm, 4.5 μm, and 5.8 μm IRAC bands but found no apparent emission from G350.1−0.3 or XMMU J172054.5−372652.

Zoom In Zoom Out Reset image size

We analyze the dynamics and evolution of G350.1−0.3 using the numerical study of nonradiative SNRs by Truelove & McKee (1999, hereafter TM99). We fit the X-ray spectrum of the SNR's entire western half by an absorbed VNEI model in order to derive the average post-shock electron density and the mass of the swept-up material. The parameters of the model are roughly intermediate to those of regions 3 and 4. We assume a spherical half-shell with thickness R/12 for the geometry and a value of 4 for the compression ratio of post-shock and pre-shock densities. We also take the ratio between the electron and atomic hydrogen densities to be n_e/n_H = 1.2, valid for cosmic abundances. With the normalization parameter from the fit (∼0.056), we use the relation

$$\begin{equation} \frac{n_e n_{\rm H} V}{4 \pi d^2} = 0.056\times 10^{14} {\rm \ cm}^{-5}, \end{equation} \tag{ 1 }$$

where V is the estimated volume and d is the distance, to derive an approximate post-shock electron density n_e ≈ 5.6 d^−1/2_4.5 cm^-3 and a corresponding swept-up mass of ∼2 d^−3/2_4.5 M_☉, with the obvious caveat that the uncertainties are unquantifiable since the visible emission—and hence the observed radius—is only a lower constraint on the true size of the SNR. G08 derived a substantially higher density through two independent methods. However, the first calculation assumes ionization equilibrium, inferred from using a two-component model (which we were unable to fit, as discussed in Section 3.1). The second calculation assumes a Sedov solution—an assumption that is likely not valid for reasons discussed below. If the molecular cloud interaction scenario (as proposed by G08) is correct, the density throughout the ejecta-dominated half of the SNR is likely to be higher than that in the western half. Hence, extrapolating our estimate to the eastern half of the SNR (for a total of ∼4 d^−3/2_4.5 M_☉) is likely an underestimate of the true swept-up mass, although this figure can be used as a rough lower bound.

For a progenitor with an ejecta density profile ∝r⁻⁷ (typical of core-collapse supernovae), TM99 demonstrate that the radius and age at which a free-expanding SNR enters the Sedov–Taylor (ST) phase are given by

$$\begin{equation} R_{{\rm ST}} = 0.881 M_{{\rm ej}}^{1/3} \rho _{0}^{-1/3} \end{equation} \tag{ 2 }$$

and

$$\begin{equation} t_{{\rm ST}} = 0.732 E_{{\rm SN}}^{-1/2} M_{{\rm ej}}^{5/6} \rho _{0}^{-1/3}, \end{equation} \tag{ 3 }$$

where M_ej is the ejecta mass and E_SN is the explosion energy. Here, ρ₀ is the pre-shock mass density and is given by

$$\begin{equation} \rho _{0} = {\displaystyle \frac{1.4}{(1.2)(4)}} m_{\rm H} n_{e} = 0.30 m_{\rm H} n_{e}, \end{equation} \tag{ 4 }$$

where m_H is the mass of the hydrogen atom, 1.4 is the equivalent molecular weight of the hydrogen and helium mixture (assuming cosmic abundances), and the factors 1.2 and 4 are the electron/hydrogen density ratio and post-shock/pre-shock compression ratio, respectively. Using typical values of M_ej = 4–14 M_☉ for the ejecta mass consistent with the formation of a neutron star (Woosley et al. 2002), we calculate R_ST to be 4.1–6.2 pc, somewhat larger than the observed radius for G350.1−0.3 (R = 3.3 d_4.5 pc). Assuming that the observed radius of the SNR is approximately equal to its true radius (in Section 4 we discuss the scenario in which this assumption is not valid), this result suggests that G350.1−0.3 is likely in the free expansion (ejecta-dominated) stage and has not yet entered the ST phase.

If we make an assumption about the explosion energy, we can determine the age using the relation

$$\begin{equation} t = 0.988 t_{{\rm ST}} \left({\displaystyle \frac{R}{R_{{\rm ST}}}}\right)^{7/4}, \end{equation} \tag{ 5 }$$

as derived from TM99. We explore the parameter space for M_ej = 4–14 M_☉ and E_SN = (0.5–1.0) × 10⁵¹ erg and derive an age of t = 600–1200 years. The shock velocity and proton temperature follow from⁸

$$\begin{equation} v_{b} = {\displaystyle \frac{\it dR}{dt}} = 0.576 \left({\displaystyle \frac{R_{{\rm ST}}}{t_{{\rm ST}}}}\right)\left({\displaystyle \frac{t}{t_{{\rm ST}}}}\right)^{-3/7} \end{equation} \tag{ 6 }$$

and

$$\begin{equation} kT_{p} = 0.11 m_{\rm H} v_{b}^{2}. \end{equation} \tag{ 7 }$$

The results of these calculations are summarized in Table 2. Using kT_e ≈ 0.7 keV for the electron temperature of the ISM component, the average value from our fits, the inferred electron-to-proton temperature ratio is 0.07–0.27. This is in excellent agreement with measured values of this ratio for SNR shocks in the inferred velocity range (Ghavamian et al. 2007).

Using the derived age for the remnant, we calculate the projected velocity of XMMU J172054.5−372652 to be 1400–2600 km s^-1, assuming a displacement of 1.7 d_4.5 pc (1.3 arcmin) from the adopted center of the SNR. We generated a light curve of XMMU J172054.5−372652 but found no short-term flux variation within the Chandra exposure. We applied barycenter corrections to the photon arrival times and searched for periodicity using the Z²-test (Buccheri et al. 1983). No pulsations with period longer than 6.4 s and pulsed fraction larger than 16% were detected at 99% confidence. The Chandra image in Figure 1(a) shows no extended emission surrounding the CCO and a generated radial profile (Figure 5) of the source confirms that it is fully consistent with a model point-spread function (PSF). As a note, photon pile-up is negligible with a level of less than 5%.

Zoom In Zoom Out Reset image size

We extracted the ∼4000 CCO counts from a 2 arcsec radius aperture and grouped the source spectrum to a minimum of 25 counts per bin. Using the SHERPA environment, we fit the spectrum to simple absorbed BB and PL models. Although the PL fit is slightly better than the BB, the large photon index (Γ = 5.5) suggests a thermal origin for the emission. Both fits are consistent with those discussed in G08. Since CCOs are often characterized by two-component models (BB+PL and BB+BB), we investigated those as well. The results are summarized in Table 3. The best-fit absorption column density is ∼4 × 10²² cm² and agrees with that of the SNR. While the CCO spectrum is dominated by a BB with kT ≈ 0.4 keV and a small emission radius of ∼3 km, the high-energy component is not very well constrained. Our results suggest that it can be modeled by either an additional BB component with a very high temperature (0.9 keV), or a PL component. Adding a second component improves the fit although not at a statistically significant level. Unfortunately, the current data do not allow us to distinguish between these two scenarios and a deeper exposure is required. We present the spectrum of XMMU J172054.5−372652 overlaid with the BB+PL model in Figure 6.

Zoom In Zoom Out Reset image size