DISCOVERY OF X-RAY-EMITTING O–Ne–Mg-RICH EJECTA IN THE GALACTIC SUPERNOVA REMNANT PUPPIS A

X-ray emission from evolved supernova remnants (SNRs) is dominated by the interstellar medium (ISM) swept up by expanding SN ejecta. Therefore, evolved SNRs had been considered to be more suitable for studies of high-Mach number (>10) shock physics, ISM/shock interactions, or the ISM itself, and less suitable for studies of the SN explosion/nucleosynthesis. However, over the last two decades, the ASCA, ROSAT, Chandra, and XMM-Newton X-ray Observatories have uncovered a number of ejecta features in evolved SNRs such as the Vela SNR (e.g., Aschenbach et al. 1995; Tsunemi et al. 1999), the Cygnus Loop (e.g., Miyata et al. 1998; Katsuda et al. 2008b), and several Large Magellanic Cloud (e.g., Hughes et al. 2003; Borkowski et al. 2006) and Small Magellanic Cloud SNRs (e.g., Park et al. 2003; Hendrick et al. 2005).

Puppis A is one of the brightest SNRs in the X-ray sky and shows both ISM/shock interactions and SN ejecta. The large extent (∼50' in diameter) as well as the high surface brightness allows us to study detailed structures in the remnant. It has been suggested that the expanding shell of Puppis A is interacting with dense H i and CO clouds from the eastern (E) rim to the northern (N) rim, based on its asymmetric X-ray surface brightness (e.g., Petre et al. 1982) and the alignment of these dense clouds with the edge of the remnant (e.g., Dubner & Arnal 1988). Chandra observations of the most prominent cloud–shock interaction at the E rim can be compared directly to scaled laboratory simulations to infer a mature interaction of age 2000–4000 years (Hwang et al. 2005; Klein et al. 2003). Thus, Puppis A is an excellent astrophysical laboratory where we can study detailed structures of cloud–shock interactions.

On the other hand, evidence of SN ejecta has been detected in Puppis A. From optical observations, Winkler & Kirshner (1985) discovered an O-rich fast-moving filament (named Ω filament) in the northeastern (NE) portion of the remnant. Subsequently, several O-rich optical knots have been found near the Ω filament, with proper-motion vectors suggesting constant expansion from a common center with a dynamical age of 3700 ± 300 years (Winkler et al. 1988). Recently, signatures of ejecta have been found in X-ray observations as well. Hwang et al. (2008) noticed Si-rich ejecta localized in the NE quadrant, based on a Suzaku survey of this SNR (see also Hwang et al. 2008 for a review of metal-rich ejecta indicated by earlier X-ray observations). Katsuda et al. (2008a) also reported the discovery of fast-moving metal-rich ejecta knots with blueshifted lines in the NE portion of the remnant, based on XMM-Newton observations. So far, the composition of all these ejecta features indicates that Puppis A originated from a core-collapse SN. This is consistent with the presence of a central compact object (CCO; Petre et al. 1996).

Here, we report the discovery of an O–Ne–Mg-rich ejecta filament in the NE quadrant of the remnant from Chandra and XMM-Newton observations. We also discuss the origin of the filament.

Puppis A has been observed several times by the CCD cameras onboard Chandra and XMM-Newton, with objectives such as the cloud–shock interactions, the CCO, or possible SN ejecta. Together, these observations provide almost complete coverage of this large X-ray remnant, as shown in Figure 1.

Zoom In Zoom Out Reset image size

All the raw data from XMM-Newton were processed using version 8.0.0 of the XMM Science Analysis Software (SAS). We select X-ray events corresponding to patterns 0–12.⁶ We further clean the data by rejecting high background (BG) intervals and removing all the events in bad columns listed in Kirsch (2006). After the filtering, the data were vignetting-corrected using the SAS task evigweight. As for Chandra, we reprocessed all the level-1 event data, applying the standard data reduction⁷ using CIAO version 4.1 and CALDB version 3.5.3. There is no high BG intervals during the Chandra observations. These observations are summarized in Table 1.

The moderate spectral resolution of the CCD cameras together with good spatial resolution of the X-ray telescopes onboard Chandra and XMM-Newton allows us to investigate the detailed spectral variation within Puppis A. Figure 2 shows an exposure- and vignetting-corrected three-color image of merged XMM-Newton and Chandra data. Red, green, and blue correspond to 0.5–0.7 keV (mostly O K-shell lines), 0.7–1.2 keV (mostly Ne K-shell lines), and 1.2–5.0 keV bands, respectively. We see clear color variations such as a blue (hard emission) belt crossing the remnant from the NE to the southwest (SW) as was previously noticed with ROSAT and ASCA observations (Aschenbach 1993; Tamura 1995; Hwang et al. 2005; Hui & Becker 2006), and the western (W) limb that is enhanced in red (soft emission). We will briefly discuss their origins based on our spectral analyses later in this paper.

Zoom In Zoom Out Reset image size

Since the CCD spectra show strong lines especially from Ne as shown in Figure 3(a) which is a spatially integrated MOS1 spectrum, we generate a line-ratio map of H-like Lyα line to Heα line blend for Ne in Figure 3(b), covering the same region as the three-color image (Figure 2). The energy bands used to generate the line-ratio map are indicated in Figure 3(a): 0.97–1.1 keV for H-like Lyα and 0.86–0.97 keV for the Heα blend. We focus on a filamentary feature enclosed in the white box in Figure 3(b). This feature (hereafter, the NE filament) is clearly seen in Figure 2 as a bright filament with different colors from those in its vicinity, indicating that it is spectrally distinct. In the NE filament, the surface brightness and line ratio seem anticorrelated, with the low line ratio indicating less emission from H-like ions. Another feature showing such an anticorrelation is marked by a black arrow in Figure 3(b). This feature is the previously noted O–Ne–Mg-rich ejecta knot (Katsuda et al. 2008a) which is coincident with the optical Ω filament (Winkler & Kirshner 1985). The spectral similarity between the NE filament and the Ω filament leads us to speculate that the NE filament may also be an ejecta feature.

Zoom In Zoom Out Reset image size

4.1. Analysis of the NE Filament

Figure 4 shows a close-up Chandra three-color image of the NE filament and its surroundings. Based on the morphology, we divide the feature into four regions; two filaments and two knotty features. These regions are marked by white ellipses labeled as Fil-1/2 and Knot-1/2 in Figure 4. The sizes of these regions range from 20'' (the diameter for Knot-1's circle) to 120'' (longer diameters for Fil-1/2's ellipses), corresponding to ∼0.21–1.26 pc at a distance of 2.2 kpc (Reynoso et al. 2003).

Zoom In Zoom Out Reset image size

Two Chandra observations (ObsIDs 5564 and 6371) cover Fil-1; we extract spectra from each and combine them to improve signal-to-noise ratio. The other regions are covered by only one Chandra observation (ObsID 6371). We subtract local BGs estimated from the surroundings using the box and pie-shaped regions shown in Figure 4. The numbers of counts after the local-BG subtraction are 7287, 6281, 2641, and 11722 for Fil-1, 2, Knot-1, and 2, respectively. To perform a χ² test, each spectrum is grouped into bins with at least 20 counts. Figure 5 shows the Fil-1 and Knot-1 spectra in black together with their BG spectra in red. The source spectrum below ∼1.5 keV shows excess emission which is dominated by O and Ne K-shell lines for Fil-1, and O, Ne, and Mg K-shell lines for Knot-1.

Zoom In Zoom Out Reset image size

Based on the spectral similarity, we divide the four spectra into two groups: Fil-1/2 and Knot-1/2. We simultaneously fit the two spectra in each group (either Fil-1/2 or Knot-1/2). We employ an absorbed, single-component, plane-parallel shock model (the Tbabs) and the vpshock model (NEI version 2.0, e.g., Borkowski et al. 2001) in XSPEC version 12.5.1). We freely vary the electron temperature, kT_e, the ionization timescale, n_et, and the volume emission measure (VEM =∫n_en_HdV, where n_e and n_H are number densities of electrons and protons, respectively, and V is the X-ray-emitting volume) for individual spectra. Above, n_et is the electron density times the elapsed time after shock heating, and we take a range from zero up to a fitted maximum value. The relative abundances of each group are tied together. The abundances of several elements, whose emission lines are prominent in the spectrum, are treated as free parameters: O, Ne, Mg, and Fe (=Ni) for Fil-1/2, and O, Ne, Mg, Si (=S), and Fe (=Ni) for Knot-1/2. The other elements are fixed to the solar values. Throughout this paper, we employ solar abundances of Anders & Grevesse (1989). We find that the abundances for Fil-1/2 are super-solar (>1000 times the solar values). In such a metal-rich plasma, it is difficult to precisely determine the absolute abundances from the X-ray spectra since H does not contribute to the continuum emission in this band. We thus set the O abundance to be 2000 times the solar value as we did for our spectral analysis of the Ω filament (Katsuda et al. 2008a; see also Winkler & Kirshner 1985).

We initially treated the hydrogen column density, N_H, as a free parameter, and obtained the best-fit value of ∼5.5 × 10²¹ cm⁻². This value is quite large compared with a typical value of 3 × 10²¹ cm⁻² for Puppis A derived from recent Suzaku observations (Hwang et al. 2008) as well as our spectral analysis shown in the following section. Given that the color of Fil-1/2 is red in Figure 4 (which means that Fil-1/2 has soft spectra), the large absorption would be strange. Furthermore, it is not reasonable that only the NE filament shows such a large absorption. We thus fix N_H to the typical value of 3 × 10²¹ cm⁻² in the analysis of Fil-1/2 and Knot-1/2. Formally, the fit for Fil-1/2 is far from acceptable (χ²/degrees of freedom (dof) = 480/248), but this is likely due to the imperfect subtraction of the local BGs. While the count rates after the local-BG subtraction are negative in some energy bands, separate fits for Fil-1 and Fil-2 do not yield a better fit for Fil-2 (while the abundances obtained for the two regions are consistent with each other). In this context, we introduce systematic errors on the model, with the error for each spectral bin being defined as $\sqrt{C_\mathrm{data}+(C_\mathrm{model}\times \alpha)}$, where C_data, C_model, and α denote data counts, model counts, and the factor of the systematic error, respectively. We set the values of α, so that we can obtain null hypothesis probabilities of the best-fit models to be ∼5%.

Figure 6 shows the local-BG-subtracted spectra along with the best-fit models. The best-fit parameters and fit statistics are listed in Table 2. The elemental abundances are measured relative to O as well as H. The super-solar abundances of O, Ne, and Mg and relative abundances of heavy elements such as Si and Fe in Fil-1/2 suggest that the origin of the filament is O–Ne–Mg-rich ejecta. On the other hand, Knot-1/2 shows slight enhancements of O, Ne, and Mg and sub-solar abundances of Si and Fe, suggesting that the knots are mixtures of the O–Ne–Mg-rich ejecta and the ISM. In addition to the abundances, the best-fit electron temperatures are also different between the two groups. These spectral differences, together with the different appearance between the two groups, indicate that they could be located at different positions along the line of sight. We note that the low temperatures and ionization timescales found in Fil-1/2 result in the low line ratios of H-like to He-like ions seen in Figure 3(b).

Zoom In Zoom Out Reset image size

Table 2. Spectral Fit Parameters

Parameter	Fil-1	Fil-2	Knot-1	Knot-2
NH (×1021 cm−2)	3 (fixed)
kTe (keV)	0.43 (0.28–0.71)	0.33 (0.26–0.54)	0.79 (0.55–1.26)	0.80 (0.60–1.06)
log(net/cm−3 s)	10.20 (9.82–10.74)	10.24 (9.88–10.61)	10.45 (10.25–10.68)	10.59 (10.44–10.82)
O (O☉)	2000 (fixed)		1.8 (0.9–3.2)
Ne (Ne☉)	3400 (2800–4500)		3.7 (1.9–9.8)
Mg (Mg☉)	3400 (1800–8000)		3.5 (1.8–6.5))
Si (Si☉)	1 (fixed)		0.8 (0.3–2.0)
Fe (Fe☉)	0 (<470)		0.9 (0.4–2.5)
Ne/O (Ne☉/O☉)	1.75 (1.45–2.15)		2.05 (1.93–2.18)
Mg/O (Mg☉/O☉)	1.85 (1.30–2.65)		2.00 (1.84–2.17)
Si/O (Si☉/O☉)	–		0.50 (0.15–0.85)
Fe/O (Fe☉/O☉)	0 (<0.25)		0.49 (0.41–0.58)
∫nenHdla (×1017 cm−5)	0.21 (0.14–0.35)	0.62 (0.39–1.39)	360 (170–900)	420 (200–880)
∫nenOdla (×1017 cm−5)	0.36 (0.24–0.60)	1.06 (0.66–2.37)	0.55 (0.26–1.38)	0.64 (0.31–1.35)
χ2/dof	287/250		204/186

Notes. ^aVEM normalized by the region area, dl is the plasma depth, and n_O is the number density of O. Values in parentheses indicate the 90% confidence ranges (calculated by including the systematic errors of 40% and 10% to the data for Fil-1/2 and Knot-1/2, respectively). The errors of relative abundances are calculated after fixing N_H, kT_e, and log(n_et) at the best-fit values. Other metals are fixed to the solar values (Anders & Grevesse 1989).

Download table as:  ASCIITypeset image

4.2. Radially Resolved Spectral Analysis

We have performed spectral analyses of the NE filament including both the filamentary and knotty features, and found them to be rich in O, Ne, and Mg. If other ejecta regions are found, then it is possible that the contact discontinuity and the reverse shock might be identified within the remnant based on the extent of the X-ray-emitting ejecta. To search for these features, we perform radially resolved spectral analysis of pie-shaped areas from the NE rim to the W rim shown in Figure 7. The center of the pie-shaped area is the inferred expansion center of the O-rich optical knots (Winkler et al. 1988). We divide the area into thin annular regions such that each region contains ∼10,000 photons. The thickness of these regions ranges from 3'' to 11'' for Chandra covering part of the NE portion, and from 13'' to 180'' for XMM-Newton covering the entire area. As BG, we use regions outside Puppis A in the same observations for Chandra. As for XMM-Newton, we use the data set accumulated from blank sky observations prepared by Read & Ponman (2003) because it provides much better statistics than the BG available from Puppis A observations. We have checked for a few regions that different BGs (either the blank sky or the source-free region around Puppis A) do not affect the spectral parameters within 90% confidence limits.

Zoom In Zoom Out Reset image size

We fit the spectra using an absorbed, single-component vpshock model. Free parameters are N_H, kT_e, n_et, VEM, and abundances of O, Ne, Mg, Si, S, and Fe (=Ni). Abundances of other elements are fixed to the solar values. For some regions in the W rim, the source emission above 2 keV is so weak that we tie the abundance of S to that of Si. Figure 8 depicts example spectra from three regions whose radial positions are indicated by arrows with letters A, B, and C in Figure 7. The fit levels for these regions are typical, with reduced χ²-values ranging from 1.02 (A) to 1.56 (C). These examples show good matches between the data and the model. The best-fit parameters for all regions are plotted as a function of radius in Figure 9. The black and red data are derived from XMM-Newton and Chandra, respectively. Note that in these plots we exclude the region of the NE filament (−810'' < R < −720''), since we already know that at least two spectral components (i.e., the O–Ne–Mg-rich ejecta and its surroundings) are required to reproduce spectra there. As we can see in the Figure 9 bottom right panel (i.e., reduced χ²), this model represents all the spectra fairly well. The general consistency of the results between XMM-Newton and Chandra validates our analysis.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

The fitted abundances are all lower than the solar values. The abundances relative to O are consistent with the solar ratios within a factor of 2, showing that there is no strong evidence of ejecta in this region except for the O–Ne–Mg-rich NE filament. The electron temperature is almost constant at 0.7 keV in the E portion of the SNR, while it gradually decreases toward the W rim to ∼0.4 keV. This temperature decrease presumably makes the W limb appear reddish in the three-color image of the SNR (see, Figure 2, left). The gradual decrease of the ionization timescale from the E rim to the W rim may reflect the gradual increase of electron density toward E (Petre et al. 1982). The slight increase of N_H toward the NE rim might reflect the extinction by the dense cloud sitting in front of Puppis A (Reynoso et al. 1995). With the exception of the NE rim, the value of N_H is almost constant at 3 × 10²¹ cm⁻², consistent with recent Suzaku observations (Hwang et al. 2008). Given that the area we investigate lies both inside and outside the hard emission (blue color in Figure 2, left) belt across Puppis A (e.g., Aschenbach 1993; see Section 3), the blue belt-like feature is more likely caused by variations in the temperature and the ionization timescale than in column density.

The X-ray surface brightness map and the Ne K line-ratio map reveal a distinct filamentary feature (i.e., the NE filament) and two knots in the NE portion of Puppis A. Our spectral analysis shows that these features are rich in O, Ne, and Mg, with abundance ratios of Ne/O, Mg/O twice solar, and that of Fe/O less than solar. The super-solar (>1000) fitted metal abundances of the NE filament clearly show that its origin is likely O–Ne–Mg-rich ejecta. The relative abundances of the O–Ne–Mg-rich ejecta can then be used to estimate the progenitor mass. We compare the relative abundances measured in the filament with those predicted by the theoretical calculations (Rauscher et al. 2002) for different progenitor masses of 15 M_☉, 25 M_☉, 30 M_☉, and 35 M_☉. We find that the data agree with those calculations for the O–Ne–Mg-rich layers in progenitor stars of ≲ 25 M_☉ masses. We note that the same conclusion can be derived by using the relative abundances of the Ω filament (Katsuda et al. 2008a), since the relative abundances of Fil-1/2 are consistent with those of the Ω filament. This is also consistent with the progenitor-mass range of 15–25 M_☉ inferred from comparisons of relative abundances of metal-rich (especially Si-rich) ejecta in the NE portion with several nucleosynthesis models (Hwang et al. 2008). We caution however that we only measured abundances of just a small fraction of the total ejecta, and so the estimate of the progenitor mass is still not conclusive.

In order to try to identify the contact discontinuity and the reverse shock, we performed radially resolved spectral analysis from the NE rim (including the O–Ne–Mg-rich NE filament) to the W rim. The spectral analysis revealed that all the spectra except for the NE filament show sub-solar metal abundances (∼0.5 times the solar values) whose relative abundances are consistent with the solar values within a factor of 2—consistent with those from Suzaku observations covering the same region (Hwang et al. 2008). This result suggests that the X-ray emission is dominated by the ISM everywhere except for the NE filament. Interpretations of the results could be (1) the outer ejecta heated by the reverse shock have already cooled, while the inner ejecta have not yet been heated by the reverse shock, (2) metal-rich ejecta could not be detected (except for the NE filament) because of their faintness compared with the ISM, or (3) most of the ejecta has not yet been heated by the reverse shock, although this last case seems unlikely for a remnant of several thousand years' age such as Puppis A. With the current data, this remains an open question.

In the area we investigated, the only region that shows evidence of SN ejecta is the small region where the O–Ne–Mg-rich NE filament and knots are located. This result may indicate that they are peculiar fragments of ejecta inside and/or outside the SNR. We then note that the NE filament is coincident with three⁸ optical O-rich fast-moving knots (Winkler et al. 1988). Figure 10 shows the proper-motion vectors of the O-rich knots on the X-ray image, in which we see that all the vectors, ∼018 yr⁻¹ (∼1900 km s⁻¹ at a distance of 2.2 kpc) toward the NE, are similar to each other. If the NE filament is physically associated with these optical knots, we expect the same proper motions for the NE filament as well.

Zoom In Zoom Out Reset image size

On the other hand, it is interesting to note that the NE filament runs parallel to the NE edge of the remnant (see, Figures 2 and 3(b)). This suggests that the morphology of the filament is related to the NE edge of the SNR where the forward shock is suggested to be interacting with dense clouds (e.g., Reynoso et al. 1995). Then, we can speculate that the filament traces ejecta heated by either a strong reflected shock off the dense ambient clouds in the NE rim or a reverse shock strongly developed in this portion of the SNR. Given that we see evidence of ejecta only in the NE filament, we may interpret the inner side of the filament to be the reflection shock and the outer side to be a boundary between the O–Ne–Mg-rich ejecta, and either the ISM or ejecta rich in lighter elements. A problem for this scenario is that the expected temperature signature at the reflected shock is not seen (Figure 9); the temperature should be higher in the twice-shocked outer regions than in the singly shocked inner regions (e.g., Hester et al. 1994). It is also uncomfortable that the filament spans only a portion of the NE edge, whereas it appears that the entire NE edge is interacting with dense clouds (e.g., Reynoso et al. 1995). However, these signatures might be less clear due to projection effects and/or possible density inhomogeneities of the ejecta; the denser ejecta will tend to have the lower temperature, as the temperature is partly determined by the condition that the ram pressure, ρv², be approximately constant. One way to test the reflection shock scenario would be to measure the proper motion of the filament, as it should then be slower than the freely expanding ejecta.

We tried measuring proper motions of the NE filament, based on X-ray images of Einstein in 1979, ROSAT in 1992, 1993, and 1994, and Chandra in 2006. Although we indeed found a possible motion, it is difficult to derive conclusive results, because there is no point-like source in the Einstein and ROSAT images to be used for image registrations, and the brightness of the NE filament seems to change during these observations. In this respect, future Chandra observations of the NE filament are desired to firmly detect the possible proper motion and brightness change, and to conclude the origin of the NE filament.

We are grateful to J. Garber and collaborators for fruitful discussions about the optical O-rich knots. We thank K. Borkowski for discussions about Fe L line emissivities. S.K. is supported by a JSPS Research Fellowship for Young Scientists and in part by NASA grant NNG06EO90A.