Table 6 summarizes the published results obtained by X-ray spectral analysis of RCW86, of which the ASCA observations are the first attempt of spatially resolved analysis. We note that only EXOSAT and GINGA results provide a statistical acceptable description of the RCW86 X-ray emission at the probability threshold of 10%. Considering all the other results, the derived values of temperature and abundances, which are two key parameters for the understanding of the SNR and its environment, are often not consistent with each other, and this suggests that this object is still poorly modeled. As a matter of fact, there is still not much agreement on the ultimate nature of the X-ray emission of RCW86, and this is partially due to the fact that this object shows very different properties in the SW and in the N and could not be properly described by fitting its spatially integrated spectrum. This is why in this section we will focus independently on the N and on the SW regions.
4.1. Interpretation of the X-ray emission in the N region
We have seen that the X-ray emission of the N filament is well described by a single temperature thermal emission model, with abundances lower than the cosmic values. This result suggests that we have detected the interaction between the blast wave and the ISM. Long & Blair (1990) have imaged and spectroscopically studied the red part of the spectrum of the group of filaments which lie near the center of our extraction region. In particular, they noticed that radiative and non-radiative characteristics are present in the same filament. Smith (1997) suggests that this could be due to a recent encounter with a dense clump. The very soft component we have detected in the PSPC spectra mentioned in Sect. 3.2.2 might be the footprint of this early interaction stage with a ISM clump, as we have discussed therein.
The best-fit temperature ( keV) found in the N filament could be taken as our best guess of the plasma temperature behind the main shock of RCW86. In this case, we can derive the main shock velocity ( km s-1) and compare it with independent estimates, such the one obtained by Long & Blair (1990). In fact, they measured the broadening of the two components in the H line profile of filaments at N and SW of the remnant and they found km s-1, which seems inconsistent with our estimate obtained at the N position. Such a shock would have keV, and would be easily detected by BeppoSAX. However, Long & Blair (1990) noted that the shock velocity derived by the ratio of the broad and narrow component is 1600-1900 km s-1, which is consistent with our estimate, but they consider more reliable the value obtained using the width of the broad component, because the calibration of the ratio between the two components sensitively depends on assumptions about the post-shock electron temperature. The two estimates could be reconciled if the shock is encountering a density gradient, e.g. a cavity wall as suggested by Vink et al. (1997), and therefore decelerating. In this case, the X-ray spectra may still be dominated by the low density wind-driven bubble ISM, while the optical filament may arise from the shock in the more dense cavity wall. A similar discrepancy between optical and X-ray derived shock velocities is also reported in the North East of the Cygnus Loop, where the broad H component gives km s-1 (Blair et al. 1999), while ASCA X-ray spectroscopy gives km s-1 (Miyata et al. 1994).
The interpretation of the X-ray thermal component in terms of shock propagating in the ISM is also suggested by the observed abundances values. The abundances are in agreement with the value reported in the "NE Hard" region by Vink et al. (1997) (also reported in Table 6), except for O and Mg, for which ASCA reports a value below 0.1; however, the BeppoSAX derived abundances have smaller uncertainties for most of the metals. Vancura et al. (1994) have developed a model of dusty non-radiative shock waves in which they show that metals are locked up in ISM grain and slowly released in the plasma as the grain are destroyed behind the shock. At the shock front, the fraction of metal not depleted onto grains is 0.51 for CNO, 1.0 for Ne, 0.05 for Mg, 0.05 for Si, 0.5 for S, 1.0 for Ar and 0.02 for Fe. Because of grain destruction, the fraction of undepleted material slowly converges toward 1 for all the metals, and this is why we expect to derive metal abundances lower than the cosmic values of Anders & Grevesse (1989). As a rule of thumb, for the derived abundances to be consistent with a metal depletion model, they should not be lower than the fractions of undepleted metals at the shock front, reported above.
The measured metal abundances in the N filament, reported in Table 3 are all above the expected metal abundances at the shock front (except Ne), and therefore consistent with the dusty shock model. Ne is inconsistent because we expect no depletion in grains, but we observe Ne. The measurement of Ne abundances may be however not reliabe with current X-ray instruments, because the Ne lines around 1 keV are embedded in the Fe-L blend. We can calculate the fraction the remaining grain mass () from the derived abundances. is 1 at the shock front and 0 when all the grain have been destroyed and the metal abundances are at their cosmic values. We have
where is the derived metal abundances and is the fraction of metal not depleted onto grains at the shock front, i.e. the expected metal abundance at the shock front. If we consider the abundance of Fe, which is the one with the lowest uncertainty ([Fe]=), we derive by Eq. (1) that the fraction of remaining grain mass is between 0.47 and 0.84. Analogously, the abundance of Si implies a remaining grain mass between 0.21 and 0.64. Combining the two results and following Vancura et al. (1994), we argue that the column swept-up by the shock is cm-2, which is equivalent to a distance behind the shock of pc, where is the post-shock density in cm-3.
4.2. Sedov analysis of the N emission
Sedov analysis is a powerful tool to derive the remnant characteristic parameters from X-ray spectral analysis. Kassim et al. (1994) reviewed the method, pointining that it could be a valid distance and age estimator, and verified it on a set of SNRs with independent distance estimates. They also stress that the main source of uncertainties in the analysis is the value of the initial explosion energy, which must be assumed and which is poorly known for Type II supernovae. Bocchino et al. (1999) extended the applicability of the method to spatially resolved X-ray spectroscopy of SNRs, and pointed out that the method is valid even in presence of ISM inhomogeneities, provided that the component responsible for the X-ray emission of shocked "inter-cloud" medium is properly identified and used.
The method is based on the mutual dependence of the age ( in units of yr), the pre-shock density ( in cm-3), the shell radius ( in pc), the explosion energy ( in unit of erg) and the shock speed ( in units of cm s-1), as given, for instance, by McKee & Hollenbach (1980):
where is in unit of K, is the normalization factor of the spectrum, is the solid angle of the extended source and l is the line of sight extension of the source (Bocchino et al. 1999). The shell radius and the age of the remnant depends weakly on the unknown l, and one can safely use a value of few parsec for shell SNR. A very reasonably estimate of l can be derived if we assume that the emission mainly comes in a thin shell behind the shock front, as the Sedov model predicts:
This relation assumes that the line of sight is tangential to the inner border of the shell, and that the shell is thin. The derived value of l may appear overestimated, but we recall that we are seeing optical and X-ray filaments because of projection of sheet-like structures (Hester 1987), which gives rise to large l.
where T and F are given by the X-ray fits and (in steradians) is measured, e.g. in the HRI. In the same way, we can derive the real shell radius, and therefore the SNR distance, because ( is the apparent shell radius, Kassim et al. 1994). Eq. (2) for the shell radius becomes
In Table 7 we report the quantities computed with Eqs. (5) and (6) using the best-fit values and uncertainties of the 1T fit of the Northern Rim listed in Table 3. The swept-up mass () is computed according to Kassim et al. (1993) to check the consistency of the Sedov approach: since we expect that SN ejected masses are in the range 1-5 for Type Ia SNe and 5-15 for Type II SNe (Woosley & Weaver 1986), values higher than that indicates Sedov evolution. Given the uncertainties in the value, we report the results computed with three different assumptions which covers a wide range of possibilities, namely , 1 and 10. If RCW86 has been generated by a Type Ia event, then should be in the range 0.9-1.5, whereas a larger range is expected in case of Type II events (Kassim et al. 1993). We note that previous estimates of the RCW86 explosion energy point towards (see Petruk 1999 for a review) but they are derived assuming an age of 1800 yr, and using spatially integrated X-ray data and/or CIE emission models. If the swept-up mass is in any case lower or comparable with the mass of ejecta, and this would imply that the remnant is still heavily interacting with the ejecta. This is hardly the case, because the derived metal abundances in the N pointings are not above the cosmic abundances. For this reason, we exclude such a low value. If we consider the generally accepted value , is only in case of a Type Ia SN, while both Type Ia and Type II events are allowed by the results obtained by the higher value .
Table 7. Results of the detailed Sedov analysis applied to the 1T fitting results of the N region. We have used the apparent shell radius reported by Green (1996)
While we can reasonably exclude , we cannot discriminate between and at this stage, apart from the fact that the canonical value is generally preferred in the literature for the supernovae. If , the derived age is fully consistent with the RCW86-SN185 association and the distance is in agreement with Ruiz (1981) and Strom (1994), but not with the distance given by Rosado et al. (1996) (), which argued against the RCW86-SN185 association, and by Milne (1970) on the basis of the relation. To reconcile with the proposed association between RCW86 and the OB association proposed by Rosado et al. (1996), we must assume and a Type II event.
4.3. Reverse shock in the SW region?
The detection of metal abundances above the cosmic values in the hot thermal component of the SW part of RCW86 is in agreement with the ASCA results, and it suggests the possibility that, at this location, the interaction is occurring with the SN ejecta. We note that this is at variance with the results obtained in the N, where there is strong evidence for Sedov-like X-ray emission. This is only apparently surprising, because RCW86 could be expanding in an ISM with large-scale density gradient. In particular, Petruk (1999) has developed a 2-D modeling of the evolution and X-ray emission of RCW86, based on the EXOSAT spatially integrated spectral results reported by Claas et al. (1989), and reported a pre-shock density contrast between the SW and Northeast parts of RCW86 in the range 3.5-4.5. In these conditions, the remnant is expected to have different evolutionary timescales at different shell location, and spatially resolved spectral analysis becomes really necessary to understand the origin of the observed emission.
We can derive the relative filling factor of the two components and their density, using Eqs. (4) and (6) of Bocchino et al. (1999), assuming that the two components are roughly in pressure equilibrium, and using the extension of the source along the line of sight given in the previous section ( pc). We obtain , , and cm-3. We have used a solid angle of the source of steradians, corresponding to the intersection of our circular extraction region, with a polygon tightly enclosing the shape of the "knee" as seen by the HRI. Moreover, using the distance derived in the previous section, we can derive the X-ray emitting mass of the two components, since ; we find and for (15 and 68 , if we use the results of Table 7).
If the SN ejecta expands in a uniform medium (as it would be more probable in case of Type Ia SNe) or in stellar wind bubble, we expect that they give rise to an X-ray component which is cooler than the one due to the shock expansion in the ambient medium. According to Chevalier (1982), the temperature of the reverse shock expanding in the ejecta is at most 1.6 times lower than the temperature of the primary blast-wave (corresponding to a power-law density profile for the ejecta), while we observe . More recently, Truelove & McKee (1999) presented a detailed semi-analytical model of non-radiative SNR in which they show that the temperature of the reverse shock could be as much as times the temperature of the shocked ejecta, but only at very early evolutionary stages ( of the age in which the remnant becomes adiabatic). We note that if the cooler component is really to be associated with the ejecta, we would find that the metal abundances in the SN ejecta are below the cosmic values, because . Even though the exact value of the abundances in the ejecta are difficult to predict, such a low value seems improbable. Vink et al. (1996) and Favata et al. (1997b), for instance, found that the low-T component in the X-ray spectrum of Cas A, which they have associated to the ejecta, has metal abundances well above the solar value.
If the SN ejecta expand in a medium in which substantial mass is present from circumstellar layers ejected before the SN explosion, the situation could be different. Interaction between ejecta and circumstellar medium (CSM) was discussed by Chevalier & Liang (1989) and shown to be consisted with observations of Cas A by Borkowski et al. (1996). The latters noted that in this case the high-T component may correspond to the SN ejecta, while the low-T component may correspond to the CSM. This scenario better fits to our data, because we find that the metallicity of the high-T component is above the cosmic value. The CSM abundances are expected to be approximately of solar-type, and Borkowski et al. (1996) found 0.40 for Fe and for the other metals.
If it is true that the X-ray emission of the SW shell of RCW86 is not associated to the ISM but to ejecta and CSM, the amount of shocked mass unrelated to the ISM is ( if ). This suggests that the progenitor SN is of Type Ia if and of Type II if , in agreement with the conclusions reached in the previous section on the basis of the emission from the N shell.
Further insights on the presence of shocked ejecta can be provided by deep optical spectrophotometry. Leibowitz & Danziger (1983) examined the spectrum at different location of a very bright complex of filaments located around 14h 40m 20s and -62d 39m, concluding that the abundances of iron should be lower than in Kepler, RCW103 and IC443. Their slit position falls in a region where the X-ray emission is soft and dominated by the component, therefore their results are in rough agreement with our measured . To properly address the topic of shocked ejecta, we need optical spectroscopy in regions dominated by the high-T component, where the optical surface brightness is much lower than the filament observed by Leibowitz & Danziger (1983).
Finding emission from the ejecta in RCW86 is crucial to the proper modeling of this object. In fact, if the connection of the remnant to the OB association at 2.8 kpc proposed by Rosado et al. (1996) is valid (at variance with the connection with SN185), then the solution with , yr and , presented in Table 7, must be favored. But, apart of the problem of a value much greater than the canonical value, this solution implies a very large swept-up ISM mass, and therefore it is not in agreement with the detection of ejecta in the SNR shell 2. For this reason, on the basis of the X-ray data, we favor the solution which implies a relatively young SNR, , a Type Ia progenitor and the association with the historical supernova.
4.4. Spatial analysis of the two-temperature components in the SW region
The hardness ratio maps presented in Fig. 4 provide a qualitative measurement of the spectral differences of the X-ray emission. Since the statistics and the BeppoSAX PSF do not allow us to investigate the observed regions with a full fit approach at angular scales smaller than several arcmin, we have tried to assess spatial spectral variations, which yield the hardness ratios in Fig. 4, with an alternative approach. This approach assumes that the observed spectral variations are entirely caused by variations of the ratio of emission measure of the two components detected in the SW and neglects that they could be due to temperature variations, or absorption effects. This is indeed the case of the Northern rim of the Vela shell, as pointed out by Bocchino et al. (1999), who have also developed a formalism which allows the filling factor of the cooler (or the hotter) plasma to be derived from the ratio and the density from the ratio . Also Ozaki & Koyama (1997) performed spatially-resolved spectral analysis of the IC443 SNR and have found that all the SNR subregions investigated have similar temperatures, the variations being mostly due to ionization time variations.
In order to model the hardness ratio in terms of the ratio, we have generated 40 simulated spectra with a two temperature model and parameter values fixed to the best-fit results listed in Table 4, column Var.Ab., except the value of the ratio, which spanned a range between 0.1 and 400 (the best-fit value is ). We have convolved the spectrum with the instrument response matrices and the effective area files which we have used for the spectral analysis of the LECS and MECS data. Then, for each spectrum we have computed the value of the ratio , and we plotted it in Fig. 8 versus the input values of the emission measures ratio. The curve allows us to estimate the ratio from the hardness ratio maps. In Fig. 8, the top x-axis reports the value of the filling factors corresponding to the emission measure ratios of the bottom x-axis, computed using Eq. (4) of Bocchino et al. (1999) and assuming keV and keV.
Fig. 8 shows that hardness ratios (HRs) much greater than 0.85 and much lower than -0.15 cannot be produced in the framework of only varying the ratio. Fig. 9 is the histogram of the HR values reported in Fig. 4 and shows that only % of the pixels has values outside this range. This is an indication that a two thermal component model with varying is a rather good description for most of the pixels. In the remaining % of the pixels, there are probably additional effects, e.g. temperature variations. In fact, we note that the pixels having are mostly located at the inner edge of the dark region in Fig. 4, which is nearer than the "knee" to the center of the remnant, and therefore we expect an increase of the temperature there. In the "SW hard" region identified by Vink et al. (1997), the HR is in the range 0.5-0.9, corresponding to a filling factor , while in "SW soft" region we can identify two blobs where HR is between -0.2 and 0 (labeled SW1 and SW2 in Fig. 4) and and an "intermediate region" between them which connects the "SW Hard" region with the outside of the shell, in which HR is between 0 and 0.5 (). Therefore, most of the X-ray bright emission region in the "knee" is characterized by a low filling factor of the high temperature component, which becomes dominant only in the inner edge of this feature and probably in the central region of the remnant.
We note that the presence of the soft X-ray component at the edge of the rim favors the possibility that this component could be associated with the CSM, rather than to the ejecta.
© European Southern Observatory (ESO) 2000
Online publication: August 17, 2000