An X-Ray Shell Reveals the Supernova Explosion for Galactic Microquasar SS 433

We selected the 0.4–1.2 keV and 2.0–5.0 keV bands for imaging analysis, where the softer band traces thermal emission from hot plasma and the nonthermal synchrotron emission dominates the harder X-ray band. The intermediate energy band was excluded to avoid the strong instrumental Al K (1.49 keV for MOS and pn) and Si K (1.75 keV for MOS) lines. For each observation, mos-spectra and pn-spectra produce images and exposure maps in the two selected bands for each camera. The instrumental background of each image, or rather, the quiescent particle background (QPB), is modeled through mos_back or pn_back and projected onto the sky by rot-im-det-sky. All the images, exposure maps, and QPB backgrounds were mosaicked using merge_comp_xmm. To underline the faint structure of the diffuse emission, we applied the task cheese to detect and mask pointlike sources in each band and subsequently adaptively binned the image based on the weighted Voronoi tessellation algorithm (M. Cappellari & Y. Copin 2003; S. Diehl & T. S. Statler 2006) to ensure an average signal-to-noise ratio of 10 bin^–1. The dual-band false-color image (Figure 1) shows the heated gas (red) and synchrotron emission of the jet (cyan).

A shell-like structure appears northeast of SS 433 in the 0.4–1.2 keV band (Figures 1 and 2(a)). Although fainter than in the jet-impact areas in the east and west, the X-ray structure is clearly detected and spatially correlated with the radio shell in the north. The emission of this structure above 2 keV is too weak to be detected in that band, indicative of a soft X-ray spectrum. Figure 2 provides a close-up image of the northern X-ray shell (a) in comparison with the radio emission (b), polarized radio signal (c), and H i gas (d). The shell has a curvature radius of $\sim 20^{\prime}$ and is roughly centered at the position of SS 433. An X-ray-enhanced clump appears in the north, where the radio emission is locally enhanced (G. M. Dubner et al. 1998) and highly polarized (X. Y. Gao et al. 2011; J. S. Farnes et al. 2017). Meanwhile, the X-ray clump is spatially associated with an H i cloud at a distance of ∼5 kpc (Y. Su et al. 2018), probably indicative of a density enhancement due to an interaction with the H i cloud.

Zoom In Zoom Out Reset image size

We analyze the spectra of the whole northern X-ray shell ($\sim 9\buildrel{\,\prime}\over{.} 4\times 4\buildrel{\,\prime}\over{.} 9$, defined as "Shell") and the substructure clump ($\sim 3\buildrel{\,\prime}\over{.} 9\times 2\buildrel{\,\prime}\over{.} 0$, defined as "Clump"). The source regions are shown in Figure 2, and the sky background is selected outside the radio boundary of W50. The instrumental background matters, especially when analyzing weak diffuse emission and when the instrumental background is nonuniform across the field of view. Therefore, instead of directly subtracting the sky background, we obtained the sky background model by fitting the QPB-subtracted background spectra and applied the model to the source spectra by multiplying an area ratio between the source and background regions. The task mos-spectra and pn-spectra extracted spectra and generated response files, while the QPB spectra of each region are modeled by mos_back and pn_back. The spectra were grouped with ftgrouppha to guarantee at least 50 counts bin^–1 in favor of chi-square statistics. XSPEC (version 12.13.0 c; K. A. Arnaud 1996) with AtomDB (version 3.0.9) was used for spectral analysis. The fitting parameters are listed in Table 1, and the spectra with folded models are shown in Appendix B.

Table 1. XMM-Newton X-Ray Spectral Fitting Results

	Absorption	Plasma								χ2/d.o.f.
Region	NH	Model	kT	τ	Norm.	O	Ne	Mg	Si
	1021 cm−2		keV	1010 s cm−3	10−4 cm−5	Z/Z⊙	Z/Z⊙	Z/Z⊙	Z/Z⊙
Shell	${6.61}_{-0.81}^{+0.90}$	vnei	${1.23}_{-0.33}^{+0.58}$	${2.6}_{-0.9}^{+1.9}$	${1.27}_{-0.31}^{+0.51}$	⋯	⋯	${1.22}_{-0.28}^{+0.33}$	${2.64}_{-0.60}^{+0.77}$	546.8/467
	${6.9}_{-2.7}^{+1.9}$	vpshock	${1.1}_{-0.3}^{+1.0}$	${5.2}_{-3.2}^{+10.0}$	${1.63}_{-0.81}^{+0.98}$	${0.84}_{-0.58}^{+1.02}$	⋯	${1.21}_{-0.34}^{+0.49}$	${3.03}_{-0.84}^{+0.65}$	547.7/466
	${4.87}_{-0.89}^{+0.74}$	vapec	${0.77}_{-0.03}^{+0.04}$	⋯	${1.62}_{-0.28}^{+0.31}$	${9.4}_{-2.7}^{+3.1}$	⋯	${3.3}_{-1.1}^{+1.4}$	${3.5}_{-1.0}^{+1.1}$	547.4/467
	${5.1}_{-1.9}^{+1.2}$	vapec*	${0.74}_{-0.04}^{+0.05}$	⋯	${2.07}_{-0.74}^{+0.70}$	⋯	⋯	${3.4}_{-1.1}^{+2.1}$	${3.5}_{-1.0}^{+1.8}$	561.7/468
	⋯	Null	⋯	⋯	⋯	⋯	⋯	⋯	⋯	763.9/472
Clump	${10.3}_{-1.6}^{+1.9}$	vnei	${0.8}_{-0.3}^{+1.1}$	${1.0}_{-0.3}^{+4.8}$	${1.4}_{-1.1}^{+2.9}$	⋯	⋯	${1.07}_{-0.33}^{+0.39}$	${2.2}_{-1.3}^{+2.4}$	195.5/162
	${8.3}_{-2.8}^{+8.1}$	vpshock	${0.92}_{-0.70}^{+0.78}$	${3.1}_{-1.7}^{+9.7}$	${0.8}_{-0.5}^{+2.5}$	${0.6}_{-0.4}^{+1.0}$	⋯	${1.25}_{-0.67}^{+0.64}$	${1.8}_{-1.0}^{+2.7}$	194.9/161
	${4.9}_{-1.7}^{+1.6}$	vapec	${0.74}_{-0.07}^{+0.07}$	⋯	${0.55}_{-0.19}^{+0.30}$	${5.9}_{-3.6}^{+4.6}$	<4.5	${3.8}_{-1.7}^{+2.8}$	${1.8}_{-1.4}^{+2.0}$	200.4/161
	${6.3}_{-2.1}^{+1.6}$	vapec*	${0.71}_{-0.06}^{+0.07}$	⋯	${0.86}_{-0.33}^{+0.41}$	⋯	⋯	${2.9}_{-1.2}^{+2.2}$	${1.5}_{-0.9}^{+1.3}$	202.5/163
	⋯	Null	⋯	⋯	⋯	⋯	⋯	⋯	⋯	294.9/167

Note. The errors correspond to 1σ uncertainties.

Download table as:  ASCIITypeset image

When fitting the sky background, we considered the X-ray emission from the local hot bubble (apec), Galactic interstellar medium (ISM; vapec), and background active galactic nucleus (powerlaw with a photon index of 1.4; L.-W. Chen et al. 1997) (S. L. Snowden et al. 2004; K. D. Kuntz & S. L. Snowden 2008). To further mimic the spectra, we let the abundances of vapec vary and got a reduced chi-square (${\chi }_{r}^{2}$) of 0.98 with 191 degrees of freedom. For source spectra, apart from the sky background, we tested four different models, including powerlaw (for nonthermal emission), vapec (plasma in collision ionization equilibrium, CIE; R. K. Smith et al. 2001), vnei (nonequilibrium ionization, NEI, plasma), and vpshock (plane-parallel shocked nonequilibrium plasma; K. J. Borkowski et al. 2001). The absorption component of all these models is TBabs (J. Wilms et al. 2000).

The nonthermal model gives a high or steep photon index Γ of ${5.71}_{-0.55}^{+0.66}$ for the Shell at a confidence level of 1σ, much higher than the synchrotron emission detected in the lobes (Γ ≈ 1.6; W. Brinkmann et al. 2007; S. Safi-Harb et al. 2022) and a typical value of <3 for nonthermal objects. In contrast, such a photon index suggests a soft spectrum probably dominated by thermal X-ray emission. Meanwhile, the reduced chi-square is 1.31 with 470 degrees of freedom, significantly poorer than any other model. The NEI models, vnei and vpshock, are widely used to describe X-ray plasmas in SNRs (K. J. Borkowski et al. 2001). Limited by the count rate, we are unable to distinguish between these two NEI models statistically with current data. The fitting result generally supports a hot plasma with a temperature kT ∼ 1 keV and an ionization timescale n_e t ∼ 10¹⁰ s cm⁻³ for both regions. The abundance of Si is over twice the solar abundance (J. Wilms et al. 2000) for the best-fit model of the Shell, while those of O and Mg are near solar. For the Clump, the abundance seems to be similar but cannot be well constrained. A model describing plasma in CIE is also tested for these two regions. The best-fit CIE model gives similar fitting statistics to the NEI models but needs a very high metal abundance, especially for O and Mg, which often appears in the metal-rich ejecta of young or middle-aged SNRs (see J. Vink 2020, and references therein). As O lines (0.5–0.7 keV) are dominated by astrophysical background compared to Mg (∼1.3 keV) and Si (∼1.8 keV) lines (see spectra in Appendix B), we also fix O to the solar abundance in CIE, named vapec* in Table 1, and use an F-test to test the significance of the underionization state (i.e., vnei versus vapec). We noticed that a deviation from CIE is significant for the Shell with a null hypothesis probability of 4 × 10⁻⁴, while for the Clump, this improvement is less evident (probability ∼0.017). Therefore, we are in favor of the NEI model here, but the CIE model cannot be completely excluded from fitting statistics, as the ionization timescale and metal abundances need further constraints in a future study. ¹⁰ Meanwhile, as the diffuse emission has a low surface brightness, we also consider the "null hypothesis," where the emission is dominated by background fluctuation (i.e., fix the sky background and free its normalization instead of weighting by the sky area) as listed in Table 1. This along with further analysis of the lower limit of the norm parameter indicates a robust detection of the diffuse emission (see Appendix C for details).

The newly discovered X-ray shell north of SS 433 is not only much dimmer than the X-ray lobes but is also composed of hot and underionized plasma, while the thermal emission of the lobe is much cooler (<0.3 keV) and close to CIE (S. Safi-Harb et al. 2022). Thus, we tend to exclude the possibility that the X-ray shell was heated by the past relativistic jet with an extreme precession angle up to ∼90°, as we would expect a cooler and CIE plasma as found in bipolar X-ray lobes (S. Safi-Harb et al. 2022). Moreover, the radio polarization measurement of this region reveals tangential magnetic fields in the northern shell and a high polarization degree up to ∼50% (X. Y. Gao et al. 2011; J. S. Farnes et al. 2017), while magnetic fields in the jet region are found to align parallel to the jet direction (P. Kaaret et al. 2024). As observed in several old SNRs, highly ordered magnetic fields along the shell can be produced by compression of the ambient gas by shocks (X. Y. Gao et al. 2011; G. Dubner & E. Giacani 2015). Considering the spectroscopy and polarization properties, we suggest that the X-ray shell has a different physical origin from the elongated jet-blown lobes but shows properties consistent with an SNR.

Based on the spectral fitting of the X-ray shell with an NEI model, we estimate its average density to be ∼0.03 cm⁻³. This is lower than that of the terminal shock of the eastern lobe (∼1 cm⁻³; S. Safi-Harb et al. 2022), showing that the gas density in the direction along and perpendicular to the jets is different. The high temperature, low density, and faint optical emission (P. Boumis et al. 2007; J. S. Farnes et al. 2017) here suggest that the remnant is still in a Sedov–Taylor phase (G. Taylor 1950; L. I. Sedov 1959). Under this assumption, we deduce that its age is ∼17–29 kyr and the explosion energy is ∼0.4–1.4 × 10⁵¹ erg (see details in Appendix D), a typical value for a core-collapse SN. The supersolar abundance of Si has previously been found in the outflowing matter of SS 433 (W. Brinkmann et al. 2005), further supporting the existence of chemical enrichment near SS 433 by an SNR.

Another potential physical origin of the X-ray shell could be the equatorial wind from the supercritical accretion disk (N. I. Shakura & R. A. Sunyaev 1973; E. M. Churazov et al. 2024), where fast winds have been observed with optical or radio observations (Z. Paragi et al. 1999; I. Waisberg et al. 2019). We estimate the power of the equatorial wind as ∼3 × 10³⁷ erg s⁻¹, based on a mass-loss rate of the gas flow of 10⁻⁴ M_⊙ yr⁻¹ (I. S. Shkovskii 1981; E. P. J. van den Heuvel 1981; Y. Fuchs et al. 2006) and an average wind terminal velocity of 1000 km s⁻¹ (S. N. Fabrika 1997). This power is optimistically high since the wind velocity drops dramatically to ∼100 km s⁻¹ in areas perpendicular to the jet (S. Fabrika 2004). Nevertheless, the wind power is much smaller than ∼10³⁹ erg s⁻¹, the "average power" of a ∼20 kyr SNR. Based on a self-similar solution of the wind bubble (R. Weaver et al. 1977; M.-M. Mac Low & R. McCray 1988), the plasma would be much cooler (∼0.2 keV) than the fitting result. To further assess this wind scenario, we simulate the evolution of an isotropic wind driven from SS 433 into a uniform diffuse medium with a density of 0.01 cm⁻³ and also consider a dense cloud at a distance of 40 pc to mimic the dense gas conditions near the Clump. The snapshots of the distribution of density, temperature, and estimated surface brightness at the ages of 20, 95, and 150 kyr are shown in Figure 3. For the estimation of the power of the wind and SN along with the details of the simulation, see Appendix E.

Zoom In Zoom Out Reset image size

Our simulation shows that it would take ∼95 kyr for the winds to blow out a cavity with a radius of the minor axis of W50. The challenge is that the wind is not strong enough for the observed X-ray flux and the gas temperature. The brightest emission appears where the winds interact with the cloud, but the surface brightness is still only half of the observed value of the Clump, especially when projection effects are taken into consideration (see Appendix E for more details). Without a wind–cloud interaction, other parts of the wind bubble are significantly fainter than the region Shell. Besides, we found that the temperature of the wind bubble is inconsistent with the observation. At around 95 kyr, the bubble shell has a temperature of ≲0.5 keV (see Figure 3), significantly cooler than that obtained from the observation. Theoretically, the temperature of the wind shell declines with age due to adiabatical cooling (see Equation (E4)). The shell should thus be hotter at an earlier age, and the simulation gives an effective temperature (emissivity-weighted) of ∼0.6 keV for the shell at 20 kyr. At the wind–cloud interaction region where X-rays are enhanced, the effective temperature is less than 0.6 keV, either in the surface brightness peak (∼90 kyr) or later evolution (150 kyr as an example). Moreover, a longer evolution timescale of the wind compared to the SNR scenario likely suggests a larger ionization timescale so that the plasma could be in or close to the CIE state. Therefore, we conclude that the equatorial disk wind cannot describe the shell-like X-ray emission north of SS 433.

As shown above, an SNR provides the best explanation for the newly discovered X-ray structure. Our X-ray study and the previously measured black hole mass of 5–15 M_⊙ in SS 433 (D. R. Gies et al. 2002; M. G. Bowler 2018; A. M. Cherepashchuk et al. 2019) provide crucial information about the initial stellar mass and the SN properties for creating a black hole. Here we discuss the progenitor mass of the black hole in SS 433 using the pre-SN core mass as predicted by the stellar evolution models (S. E. Woosley & A. Heger 2007, 2015; T. Sukhbold & S. E. Woosley 2014; T. Sukhbold et al. 2016) along with core-collapse SN explosion models (W. Zhang et al. 2008; T. Sukhbold et al. 2016; C. L. Fryer et al. 2018). Figure 4 shows the pre-SN He core masses and the compact remnant mass as a function of the zero-age main-sequence (ZAMS) stellar mass for stars with solar metallicity.

Zoom In Zoom Out Reset image size

Previous studies suggest that the He core mass sets the upper limit of the compact object mass (W. Zhang et al. 2008; T. Sukhbold et al. 2016). For stars with initial mass ranging from 15 to 30 M_⊙, we fitted the data given by T. Sukhbold et al. (2016) and found that there exists an approximate linear relation between stellar initial mass M_ZAMS and the final mass of He core M_He:

$$\begin{eqnarray}&&{M}_{\mathrm{ZAMS}}\approx 2.44{M}_{\mathrm{He}}+4.75\,{M}_{\odot }.\end{eqnarray} \tag{ 1 }$$

Above ∼40 M_⊙, the massive star strips most of the envelope mass with strong stellar winds and ends with an intermediate-mass core. To create a 5 M_⊙ compact object, one needs a progenitor star more massive than ∼17 M_⊙, while a 10 M_⊙ compact object infers the progenitor mass range of ∼30–50 M_⊙. Nevertheless, this estimate can be oversimplified, as it does not consider the metal masses taken away by SN ejecta or the explosion parameters that significantly influence the final remnant mass.

Theoretical studies predict a wide diversity of SNe from these very massive stars, from failed SNe (stars that quietly collapse to black holes; C. S. Kochanek 2014) to "hypernovae" with ${E}_{\mathrm{SN}}\gtrsim {10}^{52}$ erg (K. Nomoto et al. 2013). To explore the dependence of the compact remnant mass on the explosion energy, we adopt the predicted results from three SN explosion models for stars with solar metallicities (W. Zhang et al. 2008; T. Sukhbold et al. 2016; C. L. Fryer et al. 2018). Two of these models are taken from W. Zhang et al. (2008) for explosion energies of ∼1.2 × 10⁵¹ erg and ∼2.4 × 10⁵¹ erg. Four models from T. Sukhbold et al. (2016) applied different central engines (N20, K. Nomoto & M. Hashimoto 1988; W15, S. E. Woosley et al. 2002; S19.8, S. E. Woosley et al. 2002; and W18, T. Sukhbold et al. 2016) and provided a range of explosion energies (we only take those producing a remnant mass over 2.5 M_⊙ and with SN explosions). We also used models from C. L. Fryer et al. (2018) that calculate the remnant masses and explosion energies for stars with initial masses of 15, 20, and 25 M_⊙.

Figure 4(a) illustrates the modeled remnant mass with the ZAMS stellar mass. The remnant masses from the SN models are much smaller than the He-core-confined mass, showing that a substantial amount of material is ejected in the SN explosion. Despite some differences, all these explosion models found that ≳5 M_⊙ compact objects are made by stars ≳25 M_⊙. As shown in Figure 4(b), explosion energy strongly influences the compact remnant mass, with a heavier remnant favoring a less energetic explosion.

Therefore, the normal SN explosion energy (0.7/1.1 ×10⁵¹ erg) for W50 and the 5–15 M_⊙ compact object in SS 433 (S. Fabrika 2004; P. T. Goodall et al. 2011; J. S. Farnes et al. 2017) suggests that this binary system contained a progenitor star with a mass ≳25 M_⊙. It is noteworthy that the mass of the black hole predicted by SN explosion models does not exceed 8 M_⊙, although the upper limit of the black hole mass can be increased for failed SN cases. These models would be challenged if the mass of SS 433 is proven heavier than 8 M_⊙. Hence, a more accurate mass measurement of the compact object in SS 433 is of great value.

The existence of an SNR in W50 provides a stricter constraint on the accretion timescale and an alternative origin of accreted materials. It is thought that supercritical mass transfer of SS 433 occurs due to the accretion from the donor star. To maintain the supercritical accretion, the donor star must be highly evolved and overfill the Roche lobe (S. Fabrika 2004) within 20–30 kyr, the age of the SNR. Because of the short timescale of the post-main-sequence evolution, these two stars in the binary system would share almost the same mass, which happens with a low probability (E. P. J. van den Heuvel 1981).

Alternatively, the accreted material may not purely originate from the companion star, and the SN fallback materials should also be considered. It has already been proposed that some of the young ultraluminous X-ray sources may be black holes accreting from their fallback disks (X.-D. Li 2003), but this hypothesis is waiting for an observational test. Here, we consider that a portion of metal-rich ejecta falls back to SS 433 and provides materials for long-term accretion (X.-D. Li 2003; Q. Han & X.-D. Li 2020), while similar phenomena are observed around some neutron stars but on a smaller scale (e.g., 4U 0142+61, Z. Wang et al. 2006; 1E 161348−5055 in SNR RCW 103, X.-D. Li 2007). On the assumption of a uniform distribution of 1 M_⊙ ejecta, we estimate the accretion rate of the fallback disk to be (S. Mineshige et al. 1997; X.-D. Li 2003)

$$\begin{eqnarray}\begin{array}{rcl}\dot{M} & \approx & (5.1\times {10}^{-5}\,{M}_{\odot }\,{\mathrm{yr}}^{-1})\\ & & \times \left(\displaystyle \frac{{M}_{c}}{8\,{M}_{\odot }}\right)\left(\displaystyle \frac{{M}_{\mathrm{acc}}}{1\,{M}_{\odot }}\right){\left(\displaystyle \frac{\alpha }{0.01}\displaystyle \frac{t}{{10}^{4}\,\mathrm{yr}}\right)}^{-1.35},\end{array}\end{eqnarray} \tag{ 2 }$$

where M_c is the mass of the compact object, M_acc is the mass of the fallback materials, α is the viscosity parameter, and t is the age. This is consistent with the observed mass-loss rate of 10⁻⁵−10⁻⁴ M_⊙ yr⁻¹ (I. S. Shkovskii 1981; E. P. J. van den Heuvel 1981; Y. Fuchs et al. 2006). This amount of fallback is possible for creating a black hole (T. Sukhbold et al. 2016; C. L. Fryer et al. 2018), but the exact mass of the fallback materials is still highly uncertain. Further supporting evidence is that the abundance of Ni exceeds the solar value by ∼10 and ∼2, respectively, in the jet and winds of SS 433 (W. Brinkmann et al. 2005; P. S. Medvedev et al. 2018). The Ni enhancement can be naturally explained as nucleosynthesis products released in an SN explosion rather than originating from the donor star's envelope.

W50 has been suggested to be a potential Galactic PeVatron, accelerating particles to hundreds of TeV energies (S. Safi-Harb & R. Petre 1999). After decades of searches for γ-rays from this system, it was finally detected recently in the GeV and high-energy TeV bands (A. U. Abeysekara et al. 2018; K. Fang et al. 2020; J. Li et al. 2020; Z. Cao et al. 2024; H.E.S.S. Collaboration et al. 2024), which has been attributed to the particle acceleration in the jets or winds. The identification of an SNR shell provides us with a new insight into the cosmic-ray acceleration processes occurring in the system. Generally, the γ-ray emission (see Appendix F for details) is extended along the luminous X-ray jet lobes, with the peak emission spatially correlated with nonthermal X-ray knots (A. U. Abeysekara et al. 2018; S. Safi-Harb et al. 2022; H.E.S.S. Collaboration et al. 2024; B. Mac Intyre et al. 2024, in preparation). It is also worth noticing that some portion of the TeV and GeV emission, despite being faint, appears beyond the jet precession cone of SS 433 (K. Fang et al. 2020; J. Li et al. 2020; H.E.S.S. Collaboration et al. 2024). To the west of SS 433, where our study is focused, the TeV γ-ray emission extends to the north in all three bands. The excess seems to have a shell-like morphology, especially above 10 TeV, and shares a similar angular distance to SS 433 with the newly discovered X-ray shell. Unfortunately, the significance of the TeV emission associated with the X-ray shell is ≤2σ based on the current H.E.S.S. data. Modeling of the TeV γ-ray emission suggests a particle acceleration timescale of 1−30 kyr (H.E.S.S. Collaboration et al. 2024), consistent with the SNR age. The interpretation of the recent H.E.S.S. observation suggests that the jets are decelerated to speeds of ∼0.08c by a shock at $\sim 21^{\prime}$ (30 pc at the distance of 5 kpc) away from SS 433 (H.E.S.S. Collaboration et al. 2024), which would coincide nicely with the location of the X-ray shell. Given that SNRs are in general known to be powerful particle accelerators, it will be crucial to examine the role of the SNR in γ-ray emission in W50 and study the possible interaction between the SNR and jets (S. Safi-Harb et al. 2022).