ASYMMETRIC EJECTA DISTRIBUTION IN SN 1006

The combined XIS image of SN 1006 after subtraction of the non-X-ray background (NXB; xisnxbgen; Tawa et al. 2008) is shown in Figure 1. First, we divided the entire SNR into "non-thermal" and "thermal" regions and extracted spectra using all the five-pointing observations. The former consists of the NE and SW rims confined with the dotted lines in Figure 1, while the latter is the whole remaining region. We subtracted the cosmic X-ray background (CXB) from each spectrum by applying a power-law model with a photon index of 1.4 (Kushino et al. 2002).

The left panel of Figure 2 shows the NXB/CXB-subtracted spectra of the non-thermal and thermal regions. K-shell lines of O, Ne, Mg, Si, S, Ar, Ca, and Fe were clearly detected in the thermal spectrum. To see more detail around the Fe-K-shell line, we show the 5–10 keV band spectrum in the right panel of Figure 2. Besides the Fe-K line, we see line-like structure at ∼5.4 keV.

Zoom In Zoom Out Reset image size

We obtained the Galactic X-ray background (BGD) data from near-sky three-pointing observations of SN 1006 (BGD; Table 1). The combined BGD spectrum after the subtraction of the NXB and CXB is shown in Figure 2. Although the BGD regions (and SN 1006) are located far off from the Galactic plane, we found clear excess X-rays. The excess X-rays were already found with the Tenma satellite (Koyama et al. 1987). Although its origin (thermal or non-thermal) was unclear, it was well described by a power-law model with a photon index of 2.1 ± 0.1. Ozaki et al. (1994) confirmed the excess X-rays with the Ginga satellite and represented the spectrum with a thermal bremsstrahlung model of 7 keV temperature. Soft X-rays from an evolved SNR, the Lupus Loop (Winkler et al. 1979), are widely spread over the full area of SN 1006, and hence would contaminate the soft X-ray band. In addition, possible contribution of the soft background (SB), which was often observed from off-plane sky regions such as the North Polar Spur (Miller et al. 2008) and the Milky Way halo (MWH) should not be ignored.

Since the observation dates of the SNR Center (AO2) and the BGD regions (PV phase) are largely separated, the difference in the energy resolution and event-detection efficiency among these observations cannot be ignored. We made, therefore, a model of the BGD spectrum, and added it to the source spectrum for the following fitting procedures, instead of the direct BGD subtraction.

As we noted, the BGD model should be composed of a power-law plus several thermal components with different electron temperatures. We fit the BGD spectrum with a model of these components. The best-fit results are shown in Figure 3 and Table 2. The photon index of the local excess was obtained to be 1.9 ± 0.1, consistent with the result of Koyama et al. (1987). We also found that the Lupus Loop has two temperature spectra with ∼0.22 keV and ∼0.98 keV.

Zoom In Zoom Out Reset image size

We then renormalized the flux of the BGD model by the effective area and added it to the source spectrum as the background components. We fixed all the BGD parameters determined in this way. To confirm the reliability of the background model, we fitted the spectrum of the non-thermal rims (NE and SW) with a power-law plus the BGD model, and obtained the best-fit photon index of ∼2.8. This value is consistent with a typical index of the synchrotron radiation in the X-ray band from the non-thermal rims of SN 1006 (e.g., Koyama et al. 1995).

Yamaguchi et al. (2008) found that the Si and S-Kα lines in the spectrum of the SE region are significantly broadened compared with those expected from a single non-equilibrium ionization (NEI) plasma model. This broadening was interpreted as a superposition of multiple ejecta components with different ionization timescales. We found similar line broadening in the entire thermal spectrum, and hence applied two NEI plasma components with variable abundances (VNEI, NEIvers2.0+⁴; Borkowski et al. 2001) to represent high- and low-ionization ejecta of SN 1006 (hereafter Ejecta1 and Ejecta2, respectively). Free parameters were electron temperature kT_e, ionization timescale n_et, emission measure EM = ∫n_en_HdV, and column density N_H for interstellar absorption. Here, n_e, n_H, t, and V are the number densities of electrons and protons, elapsed time and the X-ray-emitting volume, respectively.

Since Type Ia SN ejecta should have a pure-metal composition with little contribution from H and He, we fixed the abundance of oxygen (Anders & Grevesse 1989) to be a sufficiently high value (1 × 10⁴ solar), so that the bremsstrahlung from these elements is negligible compared with those originating from the heavy element ions. Abundances of C and N were fixed to 0, while those of Ne, Mg, Si, S, Ar, and Fe were allowed to vary freely, and that of Ni was linked to Fe. For Ejecta2, the Ca abundance was also linked to Fe, but that in Ejecta1 was a free parameter because L-shell lines of Ca may contribute to the spectrum in the soft X-ray band (<0.5 keV; see the Ejecta1 component in Figure 4).

Zoom In Zoom Out Reset image size

Since the ionization timescale for Fe is very low (Yamaguchi et al. 2008), Ca would also emit K-shell lines from low-ionization states (Ejecta2). However, current NEI model does not include such emission lines. We therefore added a Gaussian line at 3.69 keV to represent the low-ionization Ca-K-shell line in Ejecta2. We further added an NEI model with the solar abundances and a power law to represent the swept-up ISM and the non-thermal emission, respectively. The energy band around the neutral Si-K edge (1.7–1.8 keV) was ignored, because the current response function is not accurate in this limited energy band.⁵ We also eliminated the 0.5–0.63 keV band, since the calibration of the contamination on the optical blocking filter was problematic.⁶

In this initial fit, we found a "shoulder"-shaped residual at around 0.7–0.8 keV. This feature was already noticed by Yamaguchi et al. (2008) and interpreted to be higher transition series of the O vii-K-shell lines (i.e., Kδ, K, etc., which are not incorporated in public NEI models). Since our NEI model also lacks the O vii-K-shell lines higher than Kδ, we added two Gaussians at 723 eV and 730 eV to represent K and Kζ, respectively. The intensity ratio of Kζ/K was assumed to be 0.5 (Yamaguchi et al. 2008) or 0.75 (Broersen et al. 2013). Although this fit was significantly improved (χ²/dof was reduced from 1300/839 to 1225/838) in the both cases, we need an unreasonably high intensity ratio of K and Kζ compared with those of lower-excitation level (e.g., K/Kδ ∼ 3.4). This may require a further additional line for the excess around 0.7 keV.

Usually, this energy band is dominated by the Fe xvii-L-shell emission of the 3s→2p transition with comparable flux to that of the 3d→2p transition at ∼0.8 keV (Foster et al. 2012). However, in an extremely low-ionized plasma, L-shell emissions from even lower charged Fe are also expected to occur following inner L-shell ionization. Although they can contribute to a spectrum around 0.7 keV, the current NEI code does not include any emission from these ions. In fact, our initial best-fit kT_e ∼ 1.7 keV and n_et ∼ 1 × 10⁹ cm⁻³ s (for Ejecta2) predict that the most dominant charge state of Fe is less than Fe^{16 +}. Another young Type Ia SNR E0509−67.5, which also has an extremely low-ionization timescale of Fe (n_et < 2 × 10⁹ cm⁻³; Kosenko et al. 2008), shows a similar excess around 0.7 keV (Warren & Hughes 2004). The weakness of the O vii and O viii emissions in SNR E0509−67.5 (Warren & Hughes 2004; Badenes et al. 2008), leads the excess to be mainly due to the Fe-L emission. Since the O vii-Kα line is strong in SN 1006 (see Figure 4 and Table 3), both the higher K-shell transitions of O vii and Fe-L emissions should be taken into account to the excess around 0.7 keV. For simplicity, we combined all the relevant lines (O vii-K, Kζ and Fe-L 3d→2p transitions) to a single Gaussian at around 0.7 keV. As we noted in the previous subsection, we found a line-like structure at ∼5.4 keV. We therefore added a Gaussian line in the final fitting. Then the center energy and flux were determined to be $5.42^{+0.20}_{-0.11}$ keV and 3  ±  2 photons cm⁻² s⁻¹, respectively.

Table 3. Best-fit Parameters (See Figure 4)

Component	Parameter		Value
Absorption	NH (× 1020 cm−2)		6.80 ± 0.07
Ejecta1 (VNEI)	kTe (keV)		0.48 ± 0.01
	Abundance (104 solar)	C	0 (fixed)
		N	0 (fixed)
		O	1.0 (fixed)
		Ne	0.53 ± 0.02
		Mg	3.71 ± 0.07
		Si	13.7 ± 0.3
		S	30 ± 1
		Ar	27 ± 1
		Ca	160 ± 8
		Fe	0.56 ± 0.01
		Ni	(=Fe)
	net (cm−3 s)		4.39 ± 0.05 × 1010
	Fluxa (erg cm−2 s−1)		7.85 ± 0.01 × 10−11
Ejecta2 (VNEI)	kTe (keV)		1.73 ± 0.03
	Abundance (104 solar)	C	0 (fixed)
		N	0 (fixed)
		O	1.0 (fixed)
		Ne	0.53 ± 0.01
		Mg	2.3 ± 0.1
		Si	23.9 ± 0.4
		S	25.2 ± 0.8
		Ar	26 ± 3
		Ca	(=Fe)
		Fe	43 ± 3
		Ni	(=Fe)
	net (cm−3 s)		1.40 ± 0.01 × 109
	Fluxa (erg cm−2 s−1)		8.66 ± 0.01 × 10−11
ISM (NEI)	kTe (keV)		0.4 ± 0.1
	net (cm−3 s)		$5.1^{+0.8}_{-0.6}\times 10^{9}$
	Fluxa (erg cm−2 s−1)		4.60 ± 0.01 × 10−11
Power law	Γ		3.11 ± 0.02
	Fluxa (photons cm−2 s−1)		3.64 ± 0.03 × 10−2
Emission line	Center energy (keV)	Normalization (photons cm−2 s−1)	1σ width (eV)
Fe-L + O vii-K	0.73 ± 0.01	3.92 ± 0.08 × 10−3	0 (fixed)
Ca-K	3.69 (fixed)	1.8 ± 0.4 × 10−5	0 (fixed)
Cr-K	$5.42^{+0.20}_{-0.11}$	3 ± 2 × 10−6	0 (fixed)
Fe-K	6.45 ± 0.02	2.6 ± 0.4 × 10−5	$101^{+33}_{-27}$
χ2/dof			1225/838 = 1.46

Note. ^aFlux in the 0.2–10.0 keV band.

Download table as:  ASCIITypeset image

The best-fit results of this final model are given in Figure 4 and Table 3. The results are basically the same as those of Yamaguchi et al. (2008); the spectrum consists of a power law, and plasmas of ISM and two ejecta components with the different ionization timescales, the abundances of the both ejecta components are consistent with each other for Ne through Ar within a factor of ∼2, whereas the Fe abundance in Ejecta2 is more than one order of magnitude larger than that in Ejecta1.

In Figure 4, we see excess structures at the both sides of the line at ∼6.4 keV. This feature is likely due to broadening of the Fe-Kα line. We therefore fit the line with a Gaussian model and found the line energy and width to be 6.45  ±  0.02 keV and $101^{+33}_{-27}$ eV, respectively (Table 3). Although Ejecta1 has an extremely large abundance of Ca (Table 3), this value should not be taken seriously, because the Ca abundance is mainly determined by the L-shell lines, in which uncertainty in the atomic data is large.

We found that the contribution of the ISM component is relatively lower than the result of SE (Figure 8 in Yamaguchi et al. 2008), but is consistent with the interpretation by Cassam-Chenaï et al. (2008); the O emission in the entire SNR predominantly originates from the shock heated ejecta rather than the ISM.

Figure 5 shows the vignetting-corrected narrowband images of the Kα lines from O, Ne, Mg, Si, and S after subtraction of the underlying continuum levels estimated by the interpolation of the adjacent band fluxes. The images show rim-brightening morphology, particularly in the light elements such as O, Ne, and Mg. In the interior regions, O, Ne, and Mg are rather uniformly distributed. In contrast, the distributions of Si and S are more asymmetric in the inner region; the SE rim is much brighter than the other regions, and the "second shell" can be seen at the middle between the outermost shell and the SNR's center. We also present the ratio maps of Si/O, S/O, Si/Ne, and Si/Mg in Figure 6. All the ratios show clear increases toward the SE rim, suggesting asymmetric ejecta distribution for the heavier elements (i.e., Si and S).

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

To investigate more quantitatively, we divided the entire SNR into three regions: a center circle of the 8'radius (C), and half-annuluses of northwest (N) and southeast (S) (the thick white lines in Figure 6). The geometric center of the remnant was defined to be (α, δ)_J2000.0 = (225.7371, −41.9336). The spectra of these three regions are shown in Figure 7. We fitted the spectra with the same four-component model (and the same assumptions) applied for the entire thermal spectrum. The best-fit results are given in Table 4. We found that the heavier elements are indeed more abundant in the region S than in the region N.

Zoom In Zoom Out Reset image size

Table 4. Best-fit Parameters (See Figure 7)

Component	Parameter		Value
			Region S	Region C	Region N
Absorption	NH (× 1020 cm−2)		6.8 ± 0.1	6.8 ± 0.1	6.8 ± 0.2
Ejecta1 (VNEI)	kTe (keV)		0.95 ± 0.01	0.95 ± 0.01	$1.02^{+0.04}_{-0.03}$
	Abundance (104 solar)	C	0 (fixed)	0 (fixed)	0 (fixed)
		N	0 (fixed)	0 (fixed)	0 (fixed)
		O	1.0 (fixed)	1.0 (fixed)	1.0 (fixed)
		Ne	0.45 ± 0.03	0.54 ± 0.02	0.45 ± 0.07
		Mg	3.0 ± 0.1	2.17 ± 0.05	2.4 ± 0.2
		Si	6.8 ± 0.2	6.7 ± 0.2	2.7 ± 0.4
		S	7.3 ± 0.8	8.8 ± 0.7	<2
		Ar	28 ± 2	14 ± 1	35 ± 5
		Ca	28 ± 2	34 ± 12	110 ± 50
		Fe	0.65 ± 0.02	0.40 ± 0.01	0.46 ± 0.05
		Ni	(=Fe)	(=Fe)	(=Fe)
	net (cm−3 s)		2.66 ± 0.09 × 1010	1.91 ± 0.05 × 1010	1.8 ± 0.2 × 1010
	Fluxa (erg cm−2 s−1)		1.02 ± 0.01 × 10−11	1.73 ± 0.01 × 10−11	0.32 ± 0.01 × 10−11
Ejecta2 (VNEI)	kTe (keV)		$1.83^{+0.05}_{-0.04}$	2.50 ± 0.09	$2.74^{+0.18}_{-0.14}$
	Abundance (104 solar)	C	0 (fixed)	0 (fixed)	0 (fixed)
		N	0 (fixed)	0 (fixed)	0 (fixed)
		O	1.0 (fixed)	1.0 (fixed)	1.0 (fixed)
		Ne	<0.02	0.41 ± 0.06	0.58 ± 0.03
		Mg	2.1 ± 0.2	3.3 ± 0.6	2.5 ± 0.2
		Si	24 ± 1	37 ± 2	12.9 ± 0.7
		S	30 ± 2	34 ± 3	10 ± 1
		Ar	14 ± 7	32 ± 10	<5
		Ca	(=Fe)	(=Fe)	(=Fe)
		Fe	20 ± 5	17 ± 5	3 ± 2
		Ni	(=Fe)	(=Fe)	(=Fe)
	net (cm−3 s)		1.58 ± 0.02 × 109	1.31 ± 0.04 × 109	1.41 ± 0.02 × 109
	Fluxa (erg cm−2 s−1)		2.08 ± 0.04 × 10−11	1.29 ± 0.01 × 10−11	1.47 ± 0.01 × 10−11
ISM (NEI)	kTe (keV)		0.4 (fixed)	0.4 (fixed)	0.4 (fixed)
	net (cm−3 s)		5.1 × 109 (fixed)	5.1 × 109 (fixed)	5.1 × 109 (fixed)
	Fluxa (erg cm−2 s−1)		2.22 ± 0.01 × 10−11	1.76 ± 0.01 × 10−11	0.91 ± 0.01 × 10−11
Power law	Γ		2.98 ± 0.03	3.16 ± 0.04	3.27 ± 0.03
	Fluxa (photons cm−2 s−1)		0.68 ± 0.01 × 10−2	0.64 ± 0.01 × 10−2	0.63 ± 0.01 × 10−2
χ2/dof			670/604 = 1.11	751/569 = 1.32	560/511 = 1.10

Note. ^aFlux calculated 0.2–10.0 keV.

Download table as:  ASCIITypeset image

We further divided the whole SNR into 16 regions: eight annuli with the width of 2' and each divided into half-rings of the northwest (NW) and SE part (solid lines in Figure 6). Then we fitted each spectra with the same four-component model. Since the statistics of the Fe line was insufficient, we fixed its abundance to the best-fit value in Table 4. The best-fit EMs for each element in the ejecta (Electa1+ Ejecta2) are given in Figure 8.

Zoom In Zoom Out Reset image size

In Section 3.2, we successfully modeled the spectrum of the entire thermal region with the components applied previously by Yamaguchi et al. (2008). The emission lines of the heavy elements exhibit clear evidence for overabundance relative to the solar values, indicating the SN ejecta origin. The difference in the ionization age between the two components suggests that Ejecta2 was heated more recently than Ejecta1. The suppressed relative abundance of Fe in Ejecta1 (Table 3) implies absence of Fe in the outer ejecta layer, consistent to the previous claim by Yamaguchi et al. (2008).

We find a possible line feature at ∼5.4 keV. This centroid energy suggests its origin to be Cr-Kα emission. To examine this possibility, we plot in Figure 9 the Cr-Kα and Fe-Kα centroids observed in other SNRs (e.g., Yang et al. 2013) and compare with our results. We find the values from SN 1006 are reasonably close to the Cr-Kα–Fe-Kα centroid correlation, supporting our claim of the first detection of Cr from this remnant.

Zoom In Zoom Out Reset image size

Since atomic data for Cr are not available in the current NEI model, we estimate the emissivity ε_Cr following the method of the ASCA measurement for W49B (Hwang et al. 2000). Using the best-fit temperature and the ionization timescale for the Ejecta2 component, we calculate emissivities for Si, S, Ar, and Fe. The results are given in Figure 10. Interpolating these values to that of Cr, we estimated the Cr/Fe abundance ratio Z_Cr/Z_Fe to be 2.5 ± 1.9 solar. Here, we assumed that Cr is mainly contained in Ejecta2 as the case of Fe.

Zoom In Zoom Out Reset image size

As we noted, the NEI model we used also lacks emission data of Ca in the low-ionization state (Ejecta2). Therefore, we estimated Z_Ca/Z_Fe with the same method to be 0.8 ± 0.2 in Ejecta2. This result is consistent with our initial assumption that the abundance of Ca is nearly equal to Fe in Ejecta2.

We compare the abundances of Ejecta2 (relative to O) with the nucleosynthesis yields predicted by several theoretical models (Figure 11). We employ the Type Ia deflagration or delayed-detonation models (Iwamoto et al. 1999) and the core-collapse SN models with various progenitor masses (15 M_☉, 20 M_☉, 30 M_☉, and 40 M_☉; Woosley & Weaver 1995) The observed abundance pattern is confirmed to be broadly consistent with the standard Type Ia models. The abundance ratios of Si/O, S/O, Ar/O, and Ca/O are enhanced compared with those predicted by the W7 deflagration model, and slightly closer to the yields of the delayed-detonation explosion (CDD1).

Zoom In Zoom Out Reset image size

The spherical outer shell with the rim brightening in O, Ne, and Mg (Figure 5) would be mainly due to the ISM components, because the ejecta distributions in Figure 8 show no enhancement in these regions. On the other hand, the interior is likely to be dominated by the ejecta (see, e.g., Figure 8). Then the ejecta distributions are broadly separated into two patterns: the uniform interior for the lighter elements (O, Ne, and Mg) and more asymmetric distribution of the heavier elements (Si and S) with offset peak positions by 5' (∼3.2 pc). The off-center distribution of the heavier elements is also confirmed in Figure 6. Notably, these two groups (O–Ne–Mg and Si–S) are synthesized in different nuclear burning processes in Type Ia SN explosion: C burning for O–Ne–Mg, whereas O burning and incomplete Si burning for Si–S. We thus presume that the latter products distribute more asymmetrically than the former.

The asymmetry of the ejecta should be formed either at the time of the SN explosion or in the process of the following interaction with the ISM. If an ambient density in the SE is higher than those in the other regions, a reverse shock might heat up the ejecta earlier, which may explain the relative enhancement of the EM in the SE. In order to estimate the ambient densities in the SE and the NW rims, we made spectra from an annulus at the outer edges of the SE and NW. We fitted the spectra and found the best-fit normalizations (EM) of the ISM components 2.46 ± 0.29 × 10⁻⁴ cm⁻⁵ and 1.93  ±  0.22 × 10⁻⁴ cm⁻⁵ for the SE and NW, respectively. Given that SN 1006 is a sphere with a radius of 10.2 pc, the volume of the annulus is estimated to be 35 pc³. Then the ISM densities are n_H = 0.11 ± 0.01 cm⁻³ and 0.09 ± 0.01 cm⁻³, hence the ambient densities (n_H/4) are ∼0.03 cm⁻³ and ∼0.02 cm⁻³, for SE and NW, respectively. Previous observations (e.g., Heng et al. 2007) indicate that the ambient density in the NW is 0.15–0.3 cm⁻³, significantly higher than our result. Since the best-fit normalization (EM) was coupled with that of n_et and anti-correlated to kT_e, a higher value of n_H/4 (up to ∼0.1 cm⁻³) is statistically acceptable in the NW. The annulus we used for the spectrum is broader than the outermost edge of the recent shock encounter. On the other hand, the forward shock in the NW interacted with the denser region fairly recently (Katsuda et al. 2013), which may explain the discrepancy between our result and the previous studies.

Accordingly, within possible statistical and systematic uncertainty, we can conclude that the ambient density in the SE is lower than 0.1 cm⁻³ and does not particularly exceed those in the other rims. Furthermore, if an inhomogeneous ambient density has affected the large-scale variation of the ejecta, the lighter elements (O, Ne, and Mg) should similarly distribute asymmetrically. As shown in Figure 6, the actual distributions clearly refute this hypothesis. Thus, it is more plausible that the ejecta asymmetry was made by an asymmetric SN explosion.

We found the "second shell" in Si and S at the middle between the outermost shell and the SNR's center (∼6' from the center). The "second shell" is most likely from the ejecta, because we see a peak of Si and S at this position in the ejecta abundance (Figure 5). Since the foreground absorption is negligible in the energies above ∼1.5 keV, this second-shell structure is not due to absorption effect, but is likely to originate from the ejecta rim, reminiscent of similar double-shell morphology in G299.2-2.9 (Park et al. 2007; although this remnant is dominated by swept-up ISM, unlike SN 1006).

We also found that the EM of Fe increases from the NW to SE, as Yamaguchi et al. (2008) presumed that the Fe-rich core spreads out to SE. These results indicate that the ejecta, particularly heavier elements in more inner region, expand preferentially toward the SE. Based on the broad UV absorption lines of Si and Fe of several background objects, Hamilton et al. (1997) estimated that the diameter toward the far side of SN 1006 is roughly 20% larger than that of the rest of the remnant (see Figure 7 of Hamilton et al. 1997). Winkler et al. (2005) also confirmed the ejecta structure of Si and Fe by the similar analysis that the shell of SN 1006 is almost spherical (see Figure 8 of Winkler et al. 2005). We thus propose that the ejecta expanded not only transverse to but also along the line of sight; the ejecta were displaced toward the SE and far side of the center (Figure 12). If the reverse shock heated up a part of the ejecta components, then the centrally peaked profile is easy understood (Figure 12). Also, if the reverse shock recently reached the surface of the Fe-rich core (light-gray region in Figure 12), the EM should be higher at the shifted direction (lower right in Figure 12).

Zoom In Zoom Out Reset image size

Including the "second shell," the peak positions of the heavy elements (Si, S, and Ar) are ∼5' = 3.2 pc away from the geometric center to SE (Figure 8). Assuming the distance of 2.2 kpc and the SNR age of 1000 yr, we estimate that the velocity difference (asymmetric explosion) between the SE and NW directions is ∼3100 km s⁻¹ in projection. We found that Fe-Kα is broadened by $101^{+33}_{-27}$ eV (1σ), for the first time from SN 1006. If this width is due to the Doppler broadening of the ejecta as is the case of Tycho (Hayato et al. 2010), the expansion velocity in the line of sight should be ∼3000 km s⁻¹. This value is consistent with those by Katsuda et al. (2013) in which they measured the proper motion of the NW rim in X-ray.