EVIDENCE FOR THERMAL X-RAY LINE EMISSION FROM THE SYNCHROTRON-DOMINATED SUPERNOVA REMNANT RX J1713.7-3946

The discovery of synchrotron X-ray emission from the shell of SN 1006 provided the first evidence for the diffusive shock acceleration of electrons to very high energies (up to 100 TeV) in supernova remnant (SNR) shocks, and cemented the long-suspected link between SNRs and cosmic rays (Koyama et al. 1995). It is now known that almost all young SNRs produce synchrotron radiation in X-rays. Of these, RX J1713.7-3946 is the brightest X-ray synchrotron SNR in our Galaxy, making it an exceptionally important target for the study of cosmic-ray acceleration.

RX J1713.7-3946 is a peculiar SNR in that its X-ray emission is dominated by synchrotron radiation; despite several studies, no thermal line emission has been detected (e.g., Koyama et al. 1997; Slane et al. 1999; Uchiyama et al. 2003; Cassam-Chenaï et al. 2004; Takahashi et al. 2008; Tanaka et al. 2008; Acero et al. 2009). While Pannuti et al. (2003) and Hiraga et al. (2005) claimed detection of thermal X-ray emission, no emission line features, the most direct evidence for thermal emission, were observed. Therefore, their claims have been debated.

The lack of thermal emission can help set an upper limit on the shocked ambient medium density, assuming a plasma geometry and an electron temperature. Deriving the ambient density is quite important, as it controls the level of the hadronic contribution to the gamma-ray emission. As a matter of fact, the origin of gamma-rays (leptonic or hadronic) in several SNRs including RX J1713.7-3946, has been a matter of considerable debate (e.g., Aharonian et al. 2004; Uchiyama et al. 2007; Yamazaki et al. 2009; Drury et al. 2009; Abdo et al. 2011; Ellison et al. 2012; Fukui et al. 2012; Inoue et al. 2012; Lee et al. 2012; Gabici & Aharonian 2014; Acero et al. 2015). Slane et al. (1999) analyzed ASCA data, and estimated the upper limit of the ambient density to be 0.03–0.69 ${D}_{1}^{-0.5}$ cm⁻³, where D₁ is the distance in units of 1 kpc; the distance of the SNR is well established to be 1 kpc by the radial velocity of the interacting molecular gas (Fukui et al. 2003), which was subsequently confirmed by Moriguchi et al. (2005) and Sano et al. (2010, 2013, 2015) from more detailed CO and X-ray studies. The loose density constraint is mainly due to the unknown plasma temperature, for which Slane et al. (1999) assumed a range of 0.1–2.5 keV, which would be reasonable given the complicated electron heating mechanism at collisionless shocks (e.g., Ghavamian et al. 2007; Drury et al. 2009). Later, Cassam-Chenaï et al. (2004) analyzed XMM-Newton data, and placed a much tighter upper limit of 0.02 ${D}_{1}^{-0.5}$ cm⁻³ at a measured temperature of ∼0.5–0.6 keV. However, this result is highly uncertain, since the introduction of a thermal component improved the fit only slightly (${\rm{\Delta }}\chi \sim$ 10–30). In fact, the authors did not claim detection of thermal emission. Therefore, non-detection of thermal emission is not sufficient to constrain the ambient density as the upper limit depends strongly on the assumed plasma temperature: a detection is necessary.

Meanwhile, thermal emission could arise from SN ejecta as well. In this case, the metal abundances allow us to probe the progenitor star, e.g., its mass and explosive nucleosynthesis within the SN. The ejecta distribution also provides important clues about the explosion mechanism. In addition, ejecta dynamics and/or the plasma states can reveal the evolutionary state of an SNR. Thus, the detection of thermal emission from RX J1713.7-3946 (from either shocked ambient medium or from shocked ejecta) has been sought since its discovery.

Here, we report on the first detection of X-ray line emission from RX J1713.7-3946, based on our new analysis of XMM-Newton and Suzaku data. The line emission is detected using both observatories and can be best interpreted as a thermal component from SN ejecta. Using the inferred abundances and other observational information, we discuss both the progenitor system and the SN type.

We analyzed archival data acquired by XMM-Newton and Suzaku, as summarized in Table 1. The fields of view are shown in Figure 1. We reprocessed the raw data, using version 14.0.0 of XMM-Newton Science Analysis Software and version 20 of the Suzaku software, together with the latest versions of the calibration files at the analysis phase. We only used data from the front-illuminated CCDs, since they have slightly better spectral resolution than back-illuminated CCDs: the MOS camera of XMM-Newton's European Photon Imaging Camera (Strüder et al. 2001; Turner et al. 2001) and sensors 0, 2, and 3 of Suzaku's X-ray Imaging Spectrometer (XIS: Koyama et al. 2007). We do not use PN data from the deep XMM-Newton pointings, since the data were taken in the timing mode to study the central compact object (CCO) 1WGA J1713.4-3946. Some of the XMM-Newton data are affected by background (BG) flares. We thus removed those periods, resulting in the net exposure times given in Table 1.

Zoom In Zoom Out Reset image size

Table 1.  X-Ray Data Used for Our Spectral Analysis

Instrument	Obs. ID (Position)	Obs. Date	Position (α, δ)${}_{{\rm{J2000}}}$	Effective Exposurea (ks)
XMM-Newton MOS1+2b	0093670101 (1)	2001 Sep 05	17:14:11.0, −39:25:00	17.4
XMM-Newton MOS1+2b,c	0093670501 (2)	2001 Sep 05	17:14:11.0, −39:25:44	13.5
XMM-Newton MOS1+2b	0093670301 (3)	2001 Sep 08	17:11:53.0, −39:55:59	15.6
XMM-Newton MOS1+2b	0093670401 (4)	2002 Mar 14	17:15:30.0, −40:00:57	12.1
XMM-Newton MOS1+2b	0203470401 (5)	2004 Mar 25	17:14:11.0, −39:25:44	16.7
XMM-Newton MOS1+2b	0203470501 (6)	2004 Mar 25	17:11:59.0, −39:31:14	14.4
XMM-Newton MOS1+2b,c	0207300201 (7)	2004 Feb 22	17:13:23.5, −39:48:29	16.2
XMM-Newton MOS1+2b	0502080101 (8)	2007 Sep 15	17:15:22.7, −39:39:38	9.9
XMM-Newton MOS1+2b	0502080301 (9)	2007 Oct 03	17:11:00.9, −39:44:26	5.4
XMM-Newton MOS1+2b	0551030101 (10)	2008 Sep 27	17:13:32.1, −40:11:59	24.6
XMM-Newton MOS1+2c	0722190101 (11)	2013 Aug 24	17:13:28.3, −39:49:53	129.6
XMM-Newton MOS1+2c	0740830201 (12)	2014 Mar 02	17:13:28.3, −39:49:53	106.4
Suzaku XIS0+2+3b	501064010 (13)	2006 Sep 11	17:12:39.2, −39:43:41	21.3
Suzaku XIS0+1+2+3c	100026020 (OFF1)	2005 Sep 25	17:09:31.9, −38:49:24	34.9
Suzaku XIS0+1+2+3c	100026030 (OFF2)	2005 Sep 28	17:09:05.1, −41:02:07	37.5

Notes.

^aAveraged among the sensors. ^bUsed for image analysis. ^cUsed for spectral analysis.

Download table as:  ASCIITypeset image

To take into account the non-X-ray BG (NXB), we used the xisnxbgen software (Tawa et al. 2008) for the XIS data, while for XMM-Newton we selected Filter-Wheel-Closed event lists¹⁶ obtained close in time to our observations. The total exposure time for the Suzaku NXB data is ten times that of the exposure time for the observation; for XMM-Newton the exposure time for the NXB is comparable to that of the observation. The shorter NXB exposure time for XMM-Newton is due to the fact that NXB data are relatively scarce. For our spectral analysis of the central region of the SNR (Section 3), we integrated all available data sets to improve statistics, resulting in merged MOS(1+2) and XIS(0+2+3) spectra with exposure times of 272.2 ks (plus 259.2 ks for an eastern part of the central region) and 63.9 ks, respectively. The exposure time is at least five times longer than those of previous CCD-based searches for thermal X-rays from the central portion of this remnant. Each spectrum was grouped into bins with at least 20 counts in order to allow us to use ${\chi }^{2}$ statistics. To fit the spectra, we used the version 12.8.2k of the XSPEC package (Arnaud 1996).

We first generated a softness-ratio map from the XMM-Newton data, to identify possible locations where thermal X-ray emission might be detectable. Figure 2 shows a band-ratio map of 0.5–1.5 to 1.5–8.0 keV; the low energy band was selected to include the Ne and Mg K-shell lines which are good tracers of thermal emission. We can see a clear spatial variation: the interior faint region is softer than the surrounding bright shell region. This contrast is likely caused by variations of the interstellar absorption, given that the soft region corresponds to a visual extinction hole (Sano et al. 2015). The soft central region is an ideal location for searching for thermal X-ray emission because (1) less absorption tends to enhance thermal X-rays which are usually softer than synchrotron X-rays, and (2) the synchrotron X-ray emission is weaker in the interior than in the surrounding shell, making it easier to detect faint thermal line emission in the presence of the dominant synchrotron continuum. Therefore, we focused our spectral analysis on a circular region enclosing the softest emission, which is shown as a black circle with a radius of 53 in Figure 2. The actual spectral extraction regions for the individual XMM-Newton and Suzaku observations are shown in Figure 3. Note that we excluded the region contaminated by the CCO, and that the extraction regions of the Obs.IDs 0722190101 (MOS1+2) and 501064010 (XIS0+2+3) do not exactly match the black circle due to the field of view of those particular observations. Another (technically) important reason to focus on this area is the existence of a substantial integrated exposure time from the archival XMM-Newton data, thanks to a recent deep observation of the CCO.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

We constructed a sky BG model by fitting NXB-subtracted XIS spectra taken from the two nearby off-source regions, OFF1 and OFF2, which are the only dedicated Suzaku and XMM-Newton observations to generate sky BGs of RX J1713.7-3946 (Takahashi et al. 2008; Tanaka et al. 2008). While there seem to be source-free regions in part of the XMM-Newton fields of view in Figure 1, we refrain from using them to avoid possible contamination of the faint diffuse emission associated with RX J1713.7-3946. Our model consists of an unabsorbed apec model to account for the local BG components (local hot bubble and SWCX emission), and an absorbed (the TBabs model; Wilms et al. 2000) apec + apec + powerlaw for the remote BG components (Galactic ridge X-ray emission and cosmic X-ray BG). This model successfully reproduces both sets as shown in Figure 4, where individual components are separately shown with labels of Local BG, Remote BG1 (the cooler apec), Remote BG2 (the hotter apec), and CXB (powerlaw). The best-fit parameters are given in Table 2. We used this model to represent the sky BG.

Zoom In Zoom Out Reset image size

Table 2.  Spectral-fit Parameters of Suzaku's Off-source Regions

Parameter	OFF1	OFF2
Local BG
${{kT}}_{{\rm{e}}}$ (keV)	0.19 ± 0.01	0.19 ± 0.01
$\int {n}_{{\rm{e}}}{n}_{{\rm{H}}}d{\ell }$ (1016 cm−5)	${3.54}_{-0.14}^{+0.16}$	2.80 ± 0.13
Remote BG
${N}_{{\rm{H}}}$ (10 ${}^{22}$ cm−2)	1.0 ± 0.1	${1.05}_{-0.02}^{+0.04}$
${{kT}}_{{\rm{e}}}1$ (keV)	0.73 ± 0.03	0.74 ± 0.02
$\int {n}_{{\rm{e}}}{n}_{{\rm{H}}}d{\ell }$ (1016 cm−5)	${6.69}_{-0.38}^{+0.31}$	${15.44}_{-0.54}^{+0.38}$
${{kT}}_{{\rm{e}}}2$ (keV)	${8.8}_{-1.7}^{+3.0}$	${7.4}_{-0.7}^{+1.0}$
$\int {n}_{{\rm{e}}}{n}_{{\rm{H}}}d{\ell }$ (1016 cm−5)	${4.52}_{-0.58}^{+2.88}$	${10.53}_{-0.43}^{+1.83}$
Γ	${1.25}_{-0.08}^{+0.05}$	${1.81}_{-0.06}^{+0.07}$
Norm (10−5 ph keV−1 cm−2 s−1 arcmin−2 at 1 keV)	${1.36}_{-0.31}^{+0.07}$	${2.95}_{-0.38}^{+0.30}$
${\chi }^{2}/$dof	744.3/577	934.7/811

Note. The abundances are set to the solar values.

Download table as:  ASCIITypeset image

Figure 5 shows the NXB-subtracted XMM-Newton MOS and Suzaku XIS spectra extracted from the regions in Figure 3. The three data sets are a merged, Obs.IDs 0093670501 + 0207300201 + 0740830201, MOS1+2 spectrum in black (XMM-Newton-1 in the upper panel), a MOS1+2 spectrum from Obs.ID 0722190101 in red (XMM-Newton-2 in the upper panel), and an XIS0+2+3 spectrum in green. We simultaneously fit these spectra with a power-law model (Model A) plus the sky BG model whose parameters are listed in Table 2. Since the three spectra are taken from different sized regions as shown in Figure 3, we allowed their model intensities to vary freely. For the power-law component, we let the photon index and normalization be free parameters. For the BG components, we fixed all the parameters at the values in Table 2, except for the absorbing column density, which we reduced from ${N}_{{\rm{H}}}\sim 1\times {10}^{22}$ cm⁻² (Table 3) to ${N}_{{\rm{H}}}=0.7\times {10}^{22}$ cm⁻². This is because our spectral extraction region is located within the absorption hole, where the visual extinction A_V is ∼1.5 mag less than that in the surrounding region (Sano et al. 2015), which is equivalent to ${\rm{\Delta }}{N}_{{\rm{H}}}$ = $3\times {10}^{21}$ cm⁻², according to the ${N}_{{\rm{H}}}$–A_V relation, ${N}_{{\rm{H}}}=2.2\times {10}^{21}$ cm⁻²A_V mag (Gorenstein 1975; Watson 2011). The ${N}_{{\rm{H}}}$ value we employed is consistent with that measured for the CCO, $\sim 0.7\times {10}^{22}$ cm⁻²(Cassam-Chenaï et al. 2004). We allowed the overall normalizations of the local and remote BG components to vary freely, while keeping fixed the relative values of the remote BG components.

Zoom In Zoom Out Reset image size

Table 3.  Spectral-fit Parameters

Parameter	Model A		Model B
BG Region	OFF1	OFF2	OFF1	OFF2
${N}_{{\rm{H}}}$ (1022 cm−2)	${0.63}_{-0.02}^{+0.04}$	0.58 ± 0.02	0.56 ± 0.02	${0.53}_{-0.01}^{+0.03}$
Γ	${2.40}_{-0.03}^{+0.06}$	${2.27}_{-0.02}^{+0.03}$	2.23 ± 0.03	${2.14}_{-0.01}^{+0.02}$
Norm (10−5 ph keV−1 cm−2 s−1 arcmin−2 at 1 keV)	${4.23}_{-0.06}^{+0.05}$	${3.97}_{-0.06}^{+0.09}$	${3.78}_{-0.07}^{+0.02}$	${3.62}_{-0.07}^{+0.03}$
Initial ${{kT}}_{{\rm{e}}}$ (keV)	⋯	⋯	0.1a	0.1a
${{kT}}_{{\rm{e}}}$ (keV)	⋯	⋯	${0.57}_{-0.06}^{+0.27}$	${0.58}_{-0.07}^{+0.10}$
log(${n}_{{\rm{e}}}t/{\mathrm{cm}}^{-3}\;{\rm{s}}$)	⋯	⋯	${11.61}_{-0.15}^{+0.16}$	${11.60}_{-0.18}^{+0.15}$
Abundance (solar) Ne	⋯	⋯	${3.3}_{-0.7}^{+4.3}$	${2.0}_{-2.1}^{+4.0}$
Mg	⋯	⋯	${7.0}_{-4.4}^{+11.0}$	${4.3}_{-2.1}^{+4.3}$
Si	⋯	⋯	${5.4}_{-1.6}^{+47.7}$	${3.2}_{-0.9}^{+5.1}$
Fe	⋯	⋯	$\lt 0.5$	$\lt 0.4$
$\int {n}_{{\rm{e}}}{n}_{{\rm{H}}}d{\ell }$ (1016 cm−5)	⋯	⋯	${2.39}_{-0.86}^{+1.42}$	${3.93}_{-2.28}^{+7.83}$
${n}_{{\rm{e}}}$ (cm−3)	⋯	⋯	0.10	0.13
Mass (${M}_{\odot }$)	⋯	⋯	0.019	0.024
Local background
Constant factor	${0.78}_{-0.07}^{+0.06}$	${0.89}_{-0.10}^{+0.11}$	${0.66}_{-0.08}^{+0.12}$	${0.75}_{-0.07}^{+0.45}$
Remote background
${N}_{{\rm{H}}}$ (1022 cm−2)	0.7a	0.7a	0.7a	0.7a
Constant factor	${2.04}_{-0.21}^{+0.26}$	${0.83}_{-0.11}^{+0.07}$	${1.40}_{-0.18}^{+0.14}$	${0.57}_{-0.16}^{+0.09}$
${\chi }^{2}/$dof	1312.4/849	1330.5/849	1148.7/842	1154.5/842

Note.

^aFixed values. The abundances are set to the solar values unless otherwise stated. As for BGs, the hidden parameters are the same as those in Table 2.

Download table as:  ASCIITypeset image

The fit results are shown in Table 3 (Model A) and Figure 5. As is emphasized in the middle and lower panels in Figure 5, the model does not reproduce the data well, leaving significant residuals around 1, 1.35, and 1.85 keV for the MOS spectra. Hints of the same residuals can be found in the XIS spectra as well. Since the residuals appear to be due to the presence of line emission, we refit the data by allowing the abundances of Ne, Mg, and Si in the remote BG low-temperature component to vary. Allowing these parameters to vary significantly improved the fits. However, the resulting BG spectral shape is quite different from those in the surrounding off-source regions, and requires unusually high metal abundances. This result leads us to consider that the thermal emission arises from RX J1713.7-3946 itself.

To test this possibility, we added a thermal non-equilibrium ionization componetnt (vvrnei in the XSPEC) to the power-law model. This model (Model B) significantly improves the fit, as shown in Figure 6 and Table 3 where the fitting details can be found. The line features at 1 keV, 1.35 keV, and 1.85 keV most likely represent Ne Lyα, Mg Heα, and Si Heα, respectively. Note that a bump near 1.5 keV seen in the XMM-Newton-1 spectrum would probably be caused by imperfect (over) subtraction of the instrumental BG, specifically the Al–K line at 1.5 keV. The statistical significance of adding the thermal component is estimated to be ≫99%, based on the F-test; ${\chi }^{2}/\mathrm{dof}$ is reduced from 1312.4/849 to 1148.7/842 when using the OFF1 BG, and from 1330.5/849 to 1154.5/842 for OFF2. We have verified that the fit results do not change significantly if we use the vnei model, which is an older and simpler version of the vvrnei model.

Zoom In Zoom Out Reset image size

The fact that the line emission is detected not only with the MOS but probably also with the XIS supports that these lines are not related to an instrumental effect. In addition, the lines do not fall near known instrumental lines such as Al K at 1.5 keV and Si K at 1.7 keV in the MOS, ruling out instrumental effects. On the other hand, the solar wind charge exchange (SWCX) can produce Ne Lyα and Mg Heα. Carter & Sembay (2008) analyzed ∼170 XMM-Newton observations to establish a method for identifying SWCX contamination. They claimed that SWCX enhancement is characterized by enhanced lines especially from O vii and O viii. Since our target (RX J1713.7-3946) is heavy absorbed by the intervening material, emission below 0.6 keV is dominated by the local BG, i.e., the local hot bubble and/or SWCX. Therefore, we are able to check whether SWCX enhancement is present in the data using light curves in the 0.5–0.7 keV energy range that includes O Heα and O Lyα. We find no SWCX enhancement in any of the data used in our spectral analysis; all of the light curves for the O K lines are stable except for time periods affected by BG flares due to soft protons. In addition, the flux below 0.6 keV (which is dominated by the local BG) is roughly the same between the SNR source spectrum and the BG regions taken outside the SNR (at different epochs), which also suggests no SWCX contamination in our data.

The abundance pattern of over-solar Ne, Mg, and Si suggests that the emitting plasma originates from SN ejecta rather than shocked interstellar medium (ISM). We note that the absolute abundances (X/H), especially for Fe, in Table 3 are poorly constrained. This is because the thermal continuum from H is hard to measure in an X-ray spectrum dominated by featureless synchrotron radiation, making it difficult to determine absolute abundances. On the other hand, the relative abundances of metals with prominent emission lines can be constrained more tightly. Therefore, we created confidence contour plots for Ne versus Mg, Ne versus Si, and Ne versus Fe as shown in Figure 7. These are calculated for BG-OFF2, but similar results are found using BG-OFF1. From these plots, the Mg/Ne ratio is 2.0–2.6 times the solar value with 1-sigma confidence, Si/Ne is 1.5–2.0, and Fe/Ne is ≲0.05. These ratios confirm the existence of an Fe abundance deficit, supporting our interpretation that the thermal X-ray emission arises from SN ejecta. We note that no Fe L line emission associated with this SNR is evident in our spectra; in particular, the 0.7–0.9 keV band, where prominent 3s $\to$ 2p and 3d $\to$ 2p lines often dominate SNR spectra, is dominated by both synchrotron radiation and the remote BGs. This is why we can calculate an upper limit on the Fe/Ne ratio. Also, the abundance pattern is in agreement with what is expected from a core-collapse (CC) SN explosion, which is evidenced by the presence of the CCO for RX J1713.7-3946.

Zoom In Zoom Out Reset image size

We estimate the electron density to be ∼0.10–0.13 cm⁻³ and the ejecta mass to be ∼0.019–0.024 ${M}_{\odot },$ assuming a line of sight plasma column depth ($d{\ell }$) of 3' (i.e., 0.87 pc at a distance of 1 kpc) which corresponds to ∼10% of the shock radius. Since the area of our spectral extraction region, 84 arcmin², corresponds to only ∼3% of the entire remnant, the total ejecta mass in this SNR should be larger than this. If we simply multiply by the area ratio, then we obtain the total ejecta mass to be ∼0.63–0.8 ${M}_{\odot }.$

4.1. The Nature of Thermal X-Ray Emission

4.2. Inferring a Progenitor Star and the SN Type

4.3. Future Prospects

By analyzing deep XMM-Newton and relatively shallow Suzaku observations of the central region of RX J1713.7-3946, we detected clear X-ray line emission, including Ne Lyα and Mg Heα, for the first time from this SNR. Our spectral analysis showed that these lines are likely arising from SN ejecta. The relative abundances among Ne, Mg, Si, and Fe suggest that the progenitor of this remnant was a relatively low-mass star, ≲20 M_⊙. On the other hand, the blastwave speed is comparatively fast among similarly aged CC SNRs, which suggests a relatively small amount of SN ejecta, in agreement with what is observed. This fact, combined with the relatively low-mass progenitor, led us to propose that RX J1713.7-3946 is the result of SN Ib/c from a close binary system in which binary interactions removed a massive hydrogen envelope.

We thank fruitful discussions with Dr. Masaomi Tanaka on supernova types and progenitor stars. We also thank the anonymous referee for very fast and constructive comments that improved the presentation of the results. This work is supported by Japan Society for the Promotion of Science KAKENHI Grant Numbers 25800119 (S. Katsuda), 24540229 (K. Koyama), 23000004 (H. Tsunemi).