THE OUTER SHOCK OF THE OXYGEN-RICH SUPERNOVA REMNANT G292.0+1.8: EVIDENCE FOR THE INTERACTION WITH THE STELLAR WINDS FROM ITS MASSIVE PROGENITOR

The X-ray images of the deep Chandra observation clearly reveal the faint outer blast wave all around the remnant (Figure 1). While the northern and western outer boundaries are well defined and relatively bright, that of the southeast (SE) is faint. Although a slight limb brightening appears to be present in some parts of the northwestern boundary, in general, the surface brightness of the outer boundary does not show a significantly limb-brightened morphology. The radial brightness variation is rather flat and even increasing inward in some regions (e.g., SE). Given the core-collapse origin of the remnant, we consider this overall morphology in the framework of the SNR expanding in its circumstellar wind.

Zoom In Zoom Out Reset image size

For a quantitative analysis, we investigate the radial variation of the X-ray spectral characteristics near the outer boundary. This approach is motivated by the fact that the gas behind the forward shock preserves the information of the medium it has swept up. Chevalier (1982) found self-similar solutions for the interaction of expanding stellar ejecta (ρ ∝ r⁻ⁿ, n > 5) with an external medium (ρ ∝ r^−s): a remnant expanding in a uniform ambient medium (s = 0) would result in an inward-decreasing density and an inward-increasing temperature profile for the shocked ambient gas, while a remnant expanding in a medium with a wind profile (s = 2) would show an inward-increasing density and an inward-decreasing temperature. Therefore, by studying the current radial profile of the density and temperature, we can infer whether the SNR is currently expanding into a density gradient.

We selected two representative regions in the northwest (NW) and southeast. We examined various narrowband images and X-ray color images of the remnant and selected regions where the contamination by ejecta emission is negligible and the area is sufficiently large to study the radial spectral variation. Each region (NW and SE) is divided into several radial subregions to track the radial spectral variation (Figure 1). The photon statistics for each subregion are ∼3000–5000 counts in the 0.5–8 keV band. The extracted spectrum for each subregion is fitted with a single nonequilibrium ionization (NEI) plane-parallel shock model (vpshock in Xspec v12; Borkowski et al. 2001). The fits with variable metal abundances (O, Ne, Mg, Si, S, and Fe were fitted while the other species were fixed at their solar values; Anders & Grevesse 1989) are statistically better (χ²/ν ∼ 1.0) than those with metal abundances frozen (χ²/ν ∼ 1.2) at solar values. The fitted metal abundances are generally subsolar, confirming that there is no significant ejecta contamination in the selected regions. The derived spectral parameters from the individual subregional fitting are not well constrained due to the large number of free parameters involved in the fit. As the abundance pattern in each radial region is similar within statistical uncertainties, we simultaneously fit all subregions in each of the NW and SE regions with their metal abundances tied. The variation of the absorbing column density when fitted independently is statistically insignificant, and thus we tie the absorbing column densities among subregions, as well. We note that our fits result in a relatively higher N_H of ∼(5–6) × 10²¹ cm² than N_H ∼ 3.2 × 10²¹ cm² from the spectral analysis of the pulsar (Hughes et al. 2001). Fixing N_H at 3.2 × 10²¹ cm² increases the temperature (by ∼20%) and decreases the normalization (by ∼30%) in a systemic way and does not fundamentally affect our results. The above approach provides acceptable fits (example spectra and the best-fit models are shown in Figure 2) for individual subregions in each of the SE and NW regions with well-constrained spectral parameters. The fitted parameters are consistent with those from independent fits within statistical errors. We utilize these fit parameters for the following radial analysis. The fit results for regions NW and SE are summarized in Tables 2 and 3.

Zoom In Zoom Out Reset image size

In Figure 3, we plot the spectral variation as a function of the radial distance of each subregion from the geometric center of the remnant as determined by the radio continuum image (α₂₀₀₀, δ₂₀₀₀ = 11^h24^m348, −59°15'529; Gaensler & Wallace 2003), where the radial distance of each subregion is normalized by the distance to the forward shock in that region. As these regions are located in the outer parts of the remnant, the uncertainty of the remnant center does not make a significant effect on the results presented in this section. The radial variations of the spectral parameters are consistent between the NW and SE regions (Figure 3). The emission measure⁸ increases inward, while the electron temperature does not. The ionization timescale (n_et) shows a similar trend as the emission measure.

Zoom In Zoom Out Reset image size

We overlay model predicted radial distributions of the temperature and emission measure in Figure 3, where the radial profiles from Chevalier (1982) are projected assuming a spherical geometry. Chevalier (1982) provides models of density, pressure, and velocity of the gas as a function of the radial distance from the center. The model emission measure in Figure 3 is calculated by integrating the density square along the line of sight. The temperature is not explicitly provided in Chevalier (1982), and we have calculated it from the thermal pressure and gas density. The projected temperature in Figure 3 is taken as an average temperature along the line of sight weighted by density squared. While we ignore the variation in the intrinsic emissivity of the gas at different temperatures, the variation is not significant (<30%) in the temperature range of 0.6–0.9 keV. Models with four different sets of n and s are plotted. The solutions of s = 2 show distinguished radial variations from the solutions of s = 0, which reflect their different radial trends in density and temperature as briefly described in Section 3.1. The projected temperature of the shocked ambient medium of s = 2 solutions shows a decrease toward the contact discontinuity (CD), in contrast to the increase in s = 0 solutions. The emission measure of s = 2 solutions monotonically increases toward the CD. The emission measure of s = 0 solutions initially increases due to the geometrical effect, and then starts to decrease. These overall trends of the emission measure and temperature profiles are dominated by the ambient density structure (s = 0 versus 2) rather than by the ejecta profile. The primary effect by the power-law index n of the ejecta is on the distance between the forward shock front and the CD: e.g., a steeper ejecta profile makes the CD close to the forward shock. Figure 3 demonstrates that the general tendency of the observed temperature and emission measure are well described by models with s = 2 (the solid and dashed lines). The model of s = 2, n = 7, in particular, is in good agreement with the observations, where our fits show reduced χ² of 0.5–1 in both SE and NW regions. On the other hand, s = 0 models (the gray and the dotted lines) show significant discrepancies especially in the SE region with a reduced χ² of ∼3. The observed radial structures of the temperature and the emission measure are inconsistent with the remnant expanding in a uniform medium where the postshock inner region is expected to result in a lower density and a higher temperature than the region immediately behind the blast wave (Chevalier 1982).

We note that the temperature from the X-ray spectral fit is the electron temperature, while the temperature from the model is the ion temperature. These two can differ considerably in non-radiative shocks faster than 1000 km s⁻¹ (e.g., Ghavamian et al. 2007 and references therein). Assuming the SNR age of 3000 yr (Ghavamian et al. 2005; Winkler et al. 2009, this age is adopted throughout this paper), the average blast wave velocity would be 2500 km s⁻¹ for a remnant radius of ∼7.7 pc (see Section 5.2), and the blast wave should have been faster in the past. The observed ratio between the electron temperature (T_e) and the proton temperature (T_p) immediately behind the shock is known to vary with shock velocities (Ghavamian et al. 2007). But the detailed variation of T_e/T_p in the shocks of v_s ≳ 2500 km s⁻¹ is unclear and we assume a constant value of T_e/T_p across the radius of the postshock regions used in our analysis (the actual value does not matter for our analysis). We note that Coulomb equilibration in the downstream region will eventually equilibrate ions and electrons, and the electron temperature will increase. This would increase the downstream electron temperature. Qualitatively, the modeled radial profile of the electron temperature would thus be flatter (in case of the s = 2 models) than that of the ion temperature, which would still be consistent with our data.

The analytic model of Chevalier (1982) is a similarity solution assuming a power-law distribution of the ambient density and ejecta, which may not be entirely adequate for actual SNRs. For example, the s = 2 models have the initial wind density profile of infinite at r = 0 and predict an infinite density and zero temperature at the CD. However, we believe that this solution is a good approximation to the real case for the regions away from the CD. While the density distribution of the ejected material depends on the structure of the progenitor star, a power-law envelope surrounding a flat (or shallower) core is a reasonable approximation (Chevalier & Soker 1989; Matzner & McKee 1999). The assumption of the power-law density of ejecta will eventually break down as the reverse shock reaches the core. While this may significantly modify the structure of the shocked ejecta material, the effect on the structure of the shocked ambient gas would not be significant. The Chevalier (1982) models do not include the effects of cosmic ray acceleration in the shock. While the efficient cosmic ray acceleration may modify the underlying shock structure, the overall trend of the radial variation is likely to be still valid (e.g., Decourchelle et al. 2000). Therefore, we conclude that the observed trend of the decreasing temperature and increasing emission measure of the shocked ambient gas toward the CD is a robust indicator that the blast wave is currently expanding into a radially decreasing density profile (s = 2) rather than a constant density (s = 0). This result provides direct observational evidence that the blast wave of G292.0+1.8 is currently propagating through the stellar winds produced by its massive progenitor. On the other hand, our results may not constrain the power-law index of the ejecta (n). The apparent effects of n is the change in the relative distance from the CD to the outer boundary. However, this distance can be varied by other physical processes that are not properly included in the model, e.g., cosmic ray accelerations. Also, the propagation of the reverse shock into the shallow ejecta core may result in the decrease of the effective power-law index. Therefore, while the observation shows better agreement with the model of n = 7 than the model of n = 12, this may not be interpreted in favor of the n = 7 model over the n = 12 model.

The expansion of SN blast waves within circumstellar winds has been extensively studied with SNe observations. The shocked CSM emits in radio continuum and also in X-rays. The mass-loss rates of the progenitor stars, estimated by modeling their light curve, are ∼10⁻⁶ M_☉ yr⁻¹ for Type Ib/c and ∼10⁻⁴–10⁻⁵ M_☉ yr⁻¹ for Type II SNe (Sramek & Weiler 2003). The circumstellar interaction of SNRs, on the other hand, has not been well established, although there has been some circumstantial evidence (Chevalier 2005).

The radial profiles of the postshock temperature and density in G292.0+1.8 argue that the remnant is expanding within the circumstellar wind. A slight limb brightening is seen in a few regions in NW, and we attribute it to a local effect, e.g., the shock is probably encountering locally dense regions in the wind. The overall dynamics of the remnant would not be significantly altered by such a small local effect, and we consider that the SNR shock has not yet reached the termination shock of the wind. The radial profile of the wind density with a constant mass-loss rate of $\dot{M}$ and a wind velocity of v_w is given by $\rho = {\dot{M}} / {4 \pi r^2 v_w}$. Adopting the geometrical center of the remnant as the wind center, where we estimate an average shock radius of r_b = 7.7 d₆ pc, we estimate

$$\dot{M} = 2 \times 10^{-5} \left(\frac{v_w}{10\;\mathrm{km\;s}^{-1}}\right) \left(\frac{n_{\mathrm H}}{0.1\, \mathrm{cm}^{-3}}\right) \left(\frac{r}{7.7\,\mathrm{pc}}\right)^2\,{M}_{\odot }\, \mathrm{yr}^{-1}.$$

A slow wind velocity (v_w ∼ 10 km s⁻¹) appropriate for the RSG winds implies a mass-loss rate $\dot{M} = 2\hbox{--}5 \times 10^{-5}\,{M}_{\odot }\, \mathrm{yr}^{-1}$ for the observed density range (n_H = 0.1–0.3 cm⁻³), which is consistent with the observations of Type II SNe. On the other hand, assuming a fast (v_w > 1000 km s⁻¹) WR wind results in $\dot{M} > 10^{-3}\,{M}_{\odot }\, \mathrm{yr}^{-1}$, a few orders of magnitudes larger than what is inferred from Type Ib/c SNe observations.

The integrated mass of the wind swept up by the remnant, assuming the wind extends down to r = 0, is

$$M_w = 15 \left(\frac{n_{\mathrm H}}{0.1\, \mathrm{cm}^{-3}}\right) \left(\frac{r_b}{7.7\,\mathrm{pc}} \right)^3 \,{M}_{\odot }.$$

Accounting for the azimuthal density variation in the postshock regions, we estimate the mass range of M_w ∼ 15–40 M_☉. There are further uncertainties involved in this estimate due to the various assumptions we made. Our density measurement scales as inverse square root of the assumed distance to the remnant (ρ ∝ d^−1/2), thus M_w ∝ d^5/2. For example, a slightly smaller distance of 5 kpc (which is within uncertainties in the distance measurements to G292.0+1.8; Gaensler & Wallace 2003) will reduce the mass estimate by ∼40%. Another source of uncertainties could be our simple assumption on the remnant geometry, a sphere centered at the geometric center of the remnant. If the wind is clumped, a local density fluctuation may have led us to overestimate the average density, as the observed emission measure is likely biased by the high-density regions. The mass estimate will also be reduced if the shock compression ratio is higher than 4 due to efficient CR acceleration, while this effect may not be significant as there is no evidence for nonthermal X-ray emission in G292.0+1.8. We assume that the mass loss remained constant until the moment of the explosion. However, observations of some SNe suggest that their mass-loss rates have decreased with time before the explosion (e.g., van Dyk et al. 1994). Also, the progenitor star might have evolved through different stages (such as blue supergiant or WR star).

Theoretical studies of the stellar evolution suggest that stars of less massive than 35 M_☉ end their life during the RSG phase, while more massive stars evolve to WR phase with a possible earlier RSG phase. Given that the wind is likely a dense RSG wind, the progenitor star of G292.0+1.8 might have a initial mass less massive than 35 M_☉. In this regard, the wind mass as large as 40 M_☉ may not be realistic for stellar evolutionary models. However, we note that there could be uncertainties in the modeling of the late stage of the stellar evolution, which is very sensitive to mass-loss rate of the star, as the stellar mass-loss rate is rather poorly constrained by observations. On the other hand, previous nucleosynthesis studies of G292.0+1.8 using X-ray emission from metal-rich ejecta suggested a progenitor mass of 20–40 M_☉ (Hughes & Singh 1994; Gonzalez & Safi-Harb 2003; Park et al. 2004), while these progenitor mass estimates were limited by nondetection of the explosive nucleosynthesis products in this SNR (Park et al. 2004). Considering that there are large uncertainties on both estimates of the wind and the progenitor masses, our estimate of the wind mass is in plausible agreement with a general picture: the progenitor of G292.0+1.8 was a massive star which lost a significant amount of its mass during the pre-SN evolution.

Our azimuthal analysis in Section 4 suggests an azimuthal density variation. Since the outer boundary of the remnant has unlikely reached the termination shock of the RSG wind, the observed density variation would represent the variation in the wind density. The wind density in the SE boundary of the remnant is smaller than those in other regions by a factor of ∼2. While this could be due to some local effects, the wind might have been asymmetric. Alternatively, the center of the wind might have not been at the geometric center of the SNR. The asymmetric SN explosion might have caused the southeastern part of the remnant to expand more rapidly than in NW, and the azimuthal density variation could be due to an offset of the progenitor star from the SNR's geometric center. Recently, Winkler et al. (2009) studied proper motion of [O iii] ejecta and located the estimated expansion center of ejecta knots, which turned out to be consistent with the SNR's geometric center. However, we note that the observed proper motions of many individual optical knots show significant deviations from the best-fit model which indicated the explosion center at the geometric center of the SNR. Thus, the origin of the azimuthal density variation along the outer boundary of G292.0+1.8 is not conclusive.

The morphology of the remnant, especially when it is young, traces its explosion characteristics. The kinematic properties of G292.0+1.8, such as the age and the explosion energy, have been estimated in the previous works. However, G292.0+1.8 was previously assumed to be expanding in a uniform medium instead of the circumstellar wind (e.g., Gonzalez & Safi-Harb 2003; Gaensler & Wallace 2003), although G292.0+1.8 is believed to be a core-collapse SNR. In fact, detailed observational studies of young core-collapse SNRs accounting for the evolution in a circumstellar wind material have been conducted for only a limited number of remnants such as Cas A (Chevalier & Oishi 2003; Laming & Hwang 2003). Here, we estimate kinematic properties of G292.0+1.8, explicitly accounting for its evolution in the circumstellar wind.

For simplicity, we assume a spherically symmetric explosion. Like the SNR evolution in a homogeneous medium, the initial evolution of the remnant is dominated by the characteristics of the ejecta (the ejecta-dominated (ED) phase) and makes a transition to the Sedov–Taylor (ST) phase as the swept-up mass becomes larger than the ejecta mass. Truelove & McKee (1999, TM99 hereafter) discussed a unified solution which describes a spherically symmetric interaction of the ejected stellar material, which has a flat core surrounded by a power-law envelope, with an ambient medium. TM99 focused on the evolution in the homogeneous medium. Laming & Hwang (2003, LH03 hereafter) extended TM99's work to evolution in a stellar wind. The characteristic age (t₀) and size (x₀) of the remnant, which mark the approximate transition from the ED phase to the ST phase, are given by

$$t_0 = 5633 M_{ej}^{3/2} E_{51}^{-1/2} \big(n_{\mathrm H}R_b^2\big)^{-1} \,\mathrm{yr},$$

$$\hspace{-2.40pc}x_0 = 40.74 M_{ej} \big(n_{\mathrm H}R_b^2\big)^{-1}\,\mathrm{pc}$$

for the evolution of the remnant in the pre-SN wind (from Equations (A3) and (A4) of LH03); n_H is a hydrogen number density of the wind at the radius of the blast wave (R_b), E₅₁ ≡ E/10⁵¹ erg, E is the explosion energy and M_ej is the ejected mass from the SN, and M_ej, R_b, and n_H are in units of M_☉, pc, and cm⁻³, respectively. If we adopt an average radius of R_b = 7.7 pc from the geometrical center and the low density of n_H = 0.1 cm⁻³ that seems to be preferred by stellar evolution model, we obtain t₀ = 950M^3/2_ejE^−1/2₅₁ yr and x₀ = 6.9M_ej pc for G292.0+1.8. Assuming M_ej of several M_☉ as in Cas A and for a canonical explosion energy of E₅₁ = 1, we estimate t₀ ∼ 10⁴ yr, which is larger than the remnant age (3000 yr) (technically, the transition time to the ST phase (t_ST) can be smaller than t₀; Truelove & McKee 1999). Thus, G292.0+1.8 might still be in its ED phase or in its transition to the ST phase. And the ED phase evolution would be better fit to describe the current kinematics of the G292.0+1.8. The evolution of the SNR in the ED phase is largely determined by E₅₁, M_ej, and the power-law index n of the ejecta envelope. The radius of the blast wave (R_b) during the ED phase is described by (from Equation (A5) of LH03)

$$\begin{eqnarray} \frac{R_b}{x_0} & = \left\lbrace \frac{l_{\mathrm{ED}}^{(n-2)}}{\phi _{\mathrm{ED}}} \frac{3}{4\pi (n-3)n} \left[ \frac{10}{3} \frac{(n-5)}{(n-3)}\right]^{(n-3)/2} \right\rbrace ^{1/(n-2)} \nonumber\\ & \quad\times \left(\frac{t}{t_0} \right)^{(n-3)/(n-2)}, \end{eqnarray} \tag{ 1 }$$

where l_ED = 1.19 + 8/n² and ϕ_ED = 0.39–0.6exp(−n/4) (the physical meanings of l_ED and ϕ_ED are discussed in LH03 and TM99). Assuming E₅₁ = 1, R_b = 7.7 pc, n_H = 0.1 cm⁻³, and t = 3000 yr, the above solution gives a large ejecta mass of M_ej ∼ 20 M_☉ for the power-law index of the ejecta envelope n between 7 and 12. Adopting a higher wind density of n_H = 0.2 cm⁻³ reduces the ejecta mass estimate and gives M_ej ∼ 10–15 M_☉. The ejecta mass is also reduced if we assume a less energetic explosion. With E₅₁ = 0.5 and n_H = 0.1 cm⁻³, we estimate M_ej ∼ 5–8 M_☉. These estimates are based on the assumption that the remnant is strictly in the ED phase. The deceleration of the remnant during its transition to the ST phase would increase the explosion energy and decrease the ejecta mass. We note that due to the high wind density near the center, radiative cooling may have been effective in the very early age of the remnant (which may be responsible for the optical emission from the ejecta; LH03) and this could have affected the overall evolution. However, at this early time, most of the explosion energy would be in the form of the kinetic energy of the expanding ejecta, and the loss of energy by radiative cooling should be negligible in the total energy conservation and the overall remnant evolution.

The shock velocity of the blast wave (v_b) is given as

$$v_b = 2500\;\mathrm{km\;s}^{-1}\left(\frac{n-3}{n-2}\right) \left(\frac{r_b}{7.7\,\mathrm{pc}}\right) \left(\frac{t}{3000\,\mathrm{yr}}\right)^{-1}.$$

For the observed electron temperature range of 0.6–0.9 keV, this shock velocity indicates T_e/T_p ∼0.1, consistent with those in other remnants with comparable shock velocities (Ghavamian et al. 2007). The optical observations of [O iii] ejecta show an average expansion velocity of ∼1700 km s⁻¹ (Ghavamian et al. 2005). This velocity may be considered an upper limit for the current reverse shock velocity (the velocity in the rest frame of the ambient gas; $\frac{dR_r}{dt}$). Using Equation (A13) of LM03, we derive $\frac{dR_r}{dt} \simeq \frac{x_0}{t_0}$ near t = t₀. The derived velocity $\frac{dR_r}{dt}$ is insensitive to t because of its logarithmic dependency. We estimate $\frac{dR_r}{dt} \sim 2040\;\mathrm{km\;s}^{-1}$ for E₅₁ ∼ 1 and $\frac{dR_r}{dt} \sim 1660\;\mathrm{km\;s}^{-1}$ for E₅₁ ∼ 0.5, which are consistent with the optical observations (Ghavamian et al. 2005).

While we have assumed a spherically symmetric explosion, recent observations and hydrodynamic modelings of SNe revealed a compelling case that high-mass SN explosions are intrinsically aspherical events (Wang & Wheeler 2008; Burrows et al. 2005; Janka et al. 2005). Previous observations of G292.0+1.8 indeed suggested that the explosion of G292.0+1.8 may have been asymmetric. Park et al. (2007) showed that the ejecta temperature is significantly higher generally in the northwestern and western regions of the SNR than in the southeastern regions. They suggested that this large-scale nonuniform distribution of the ejecta temperature may be caused by an asymmetric SN explosion: the explosion could have been more energetic toward NW–W than in SE. Winkler et al. (2009) found that [O iii] knots in north–south (N–S) axis are faster than those along the east–west axis, and suggested a possibility that explosions were more energetic along N–S axis. Also, there is a possibility that the azimuthal density variation around the remnant might be related with the asymmetric explosion (Section 4). However, these scenarios are not fully consistent with each other and further studies are needed. Our future study, which will focus on the X-ray characteristics of the shocked ejecta material, will help us reveal the detailed nature of the explosion of G292.0+1.8.

For the progenitor mass of ∼20–40 M_☉ estimated from previous X-ray observations (Hughes & Singh 1994; Gonzalez & Safi-Harb 2003; Park et al. 2004), the overall observed wind properties are in plausible agreement with the general physical picture that the progenitor was a massive star which lost a significant amount of the mass through the RSG winds. However, given the uncertainties in the observational results and theoretical models, the current results do not provide a unique evolutionary track for the progenitor star (of a specific initial mass) which led to the explosion of G292.0+1.8. Based on our results, we briefly discuss a few possible scenarios for the evolutionary track of the progenitor.

The progenitor star of initial mass of as low as ∼20 M_☉ seems to be consistent with lower end of our wind mass estimate, where the star had a mass loss of ∼10–15 M_☉ as the RSG wind while the rest of the mass went into the ejecta and the neutron star. In this scenario, the kinematics of the remnant can be explained by an explosion of E₅₁ ∼ 0.5. While the stellar evolutionary models for a single star of M ∼ 20 M_☉ seem to predict smaller mass losses than 10 M_☉, a possible companion star may have increased the mass loss, as proposed for the case of Cas A (Young et al. 2006). A significantly increased amount of mass loss is expected if the progenitor is more massive. For example, the models suggest that a star of 30 M_☉ loses ∼20 M_☉ as the RSG wind and than explodes as the RSG star (Woosley et al. 2002). While a black hole might be expected after the explosion of such a massive star, the final fate of the compact remnant is inconclusive as observations show that a neutron star can be still formed from even more massive stars (>40 M_☉; Muno et al. 2006). A progenitor star of >35 M_☉ likely has evolved into the WR phase after the RSG phase. This might be the case of G292.0+1.8 if its WR phase has been short. In summary, given that the wind is most likely the RSG wind, the evolutionary models seem to prefer a progenitor's initial mass between ∼20and35 M_☉. This is plausibly consistent with the lower range of our wind mass estimates (∼15 M_☉), although the higher mass progenitor cannot be completely ruled out.