Astron. Astrophys. 360, 671-682 (2000)

3. Results

3.1. Spatial analysis

In Fig. 2 and Fig. 3 we show the ROSAT HRI, LECS and MECS images of the SW and N filaments of the RCW86 shell. The LECS image is restricted to the 0.1-2.0 keV band, while the MECS image is restricted to the 2.0-10.5 keV band. The HRI image is in counts per 8 arcsec pixel and smoothed with a gaussian of 1.5 pixel . The BeppoSAX images have been corrected for vignetting effects, have a pixel size of 32 arcsec, and have also been smoothed with a Gaussian of 1.5 pixel . The Point Spread Function (PSF) of the SAX instruments (3.0´ at 1.5 keV for the LECS and 2.5´ at 6.4 keV for the MECS, 80% of encircled energy radius) does not allow the structure of the X-ray filaments to be traced, which is instead clearly visible in the HRI image. Nevertheless, significant differences of the emission is evident down to the angular scale.

Fig. 2. ROSAT HRI 0.1-2.4 keV (left panel ), BeppoSAX LECS 0.1-2.0 keV (middle panel ) and MECS 2.0-10.5 keV (right panel ) images of the SW part of the RCW86 shell. Units are counts per 8" pixel for the HRI image and counts per second per 32" pixel for the BeppoSAX images. Linear contours (10 linearly spaced between 0 and maximum) and grey scale are also shown. The radius circles represent the extraction region for the LECS+MECS spectra

Fig. 3. Same as Fig. 2 but for the N part of the RCW86 shell. The asymmetric shape of the field of view, which includes the feature at and , is due to the exclusion of the hot pixels caused by the MECS calibration sources

Spectral differences across the filaments are better shown by the hardness ratio maps in Fig. 4. The maps, binned using a pixel size of , have been obtained with the LECS 0.1-2.0 keV and MECS 2.0-10.5 keV band data; the value at each pixel has been computed using the formula (HR), where C is the number of counts. The resulting uncertainty in the hardness ratios are between and for HR, and between and for HR. In the SW (Fig. 4, left panel) the soft emission is more external with respect of the hard emission throughout the "knee", except in the East, at RA=14^h 42^m 30^s and DEC=-62^d 45^m, where the soft emission is nearly absent and the hard emission is present at the limb of the shell. This position corresponds to the S1b position of Smith (1997), where no [S II] emission is observed, but which does display Balmer-dominated H filaments. It is interesting to note that while the relation hard emission - Balmer dominated optical filaments seems to hold, the inverse correspondence soft emission - radiative optical filaments is not always fulfilled. In fact, the soft region at the north end of the knee (14^h 41^m, -62^d 30^m) corresponds to a region where the [S II] emission is very low and the filaments are mostly Balmer dominated (W1b, W2b and W3b in Smith 1997).

Fig. 4. Hardness ratio maps of the South Western part (left panel ) and of the Northern part of the RCW86 (right panel ). The maps have been obtained from the LECS 0.1-2.0 keV and MECS 2.0-10.5 keV images (Fig. 2 and Fig. 3) binned to a pixel of 2´; the hardness ratio for each pixel has been computed by the HR formula

In the N (Fig. 4, right panel) the situation appears to be different, since the soft emission has not a sharp localization and seems more diffuse across the field of view. The soft region around and is the continuation of the northern and soft end of the knee.

3.2. Spectral analysis

3.2.1. The LECS and MECS spectra

LECS and MECS spectra were accumulated in the 8´ circular regions reported in Fig. 2 and Fig. 3. For the LECS, appropriate effective area and response matrix files have been computed with the task LEMAT version 3.5.3, while for the MECS we have used the standard response matrix. In both cases, the effective area files have been computed using the source profile extracted from the HRI image. Background spectra was collected in a BeppoSAX observation of Proxima Centauri, which is off RCW86, because there are no suitable background regions in the RCW86 observations. For the LECS we have followed the "background semi-annuli" method of Parmar et al. (1999b) to collect and normalize the background spectrum in the observation of Proxima Centauri, and then we have used it in the RCW86 observation, while for the MECS we have used a similar method, collecting the background in a large annulus around Proxima Centauri, normalizing it to the same area used for the extraction of the source spectrum, and then using it in the RCW86 observation. The "standard background" method also quoted by Parmar et al. (1999b) has proven to be unusable for RCW86, because at the position of the remnant we expect a non-negligible contribution of the Galactic Ridge X-ray emission and this method would underestimate the background, especially at high energies. The "Scaled ROSAT PSPC all-sky survey" method of Parmar et al. (1999b) has also been discarded since it would provide a background spectrum with too few counts and contaminated by the remnant itself.

The spectra have been rebinned to 1/3 of the effective spectral resolution (FWHM), and then rebinned again to ensure that a minimum of 20 counts are in each channel. The LECS and MECS spectra have been jointly fitted with one temperature and two temperature Non-Equilibrium of Ionization (NEI) emission models included in the SPEX package (Kaastra et al. 1996) plus the absorption component of Morrison & McCammon (1983). The reference abundances are those of Anders & Grevesse (1989), which will be referred in the following as "cosmic". Before fitting, we have multiplied the LECS spectra by the constant 0.87, to take into account the relative normalization between LECS and MECS (see e.g., Favata & Schmitt 1999); we have also verified that varying the multiplication factor yields higher values.

One of the dangers of fitting NEI models is to get easily trapped in local minima because of the correlation between the temperature and the ionization time (). Including metal abundances as free parameters makes this problem worse. For this reason, we have followed a rigid scheme in fitting our data, starting with a single-component NEI model having fixed abundances, calculating the and the corresponding probability, and setting more free parameters only if the fit does not return an acceptable probability. When fitting with variable individual abundances we have chosen to vary only CNO, Ne, Mg, Si, S, Ar and Fe, because these metals mostly affect the BeppoSAX spectra. In particular, the measurement of abundances of some elements is driven by the signature due to line emission in some narrow bands, namely around 1 keV and 6.5 keV for Iron, in the 1.2-1.5 keV band for Mg, 1.8-2.0 keV band for Si, 2.4-2.6 band for S, and the 2.8-3.0 for Ar. Notwithstanding the unaivodable correlation between some of the remaining spectral parameters, the metal abundances of hot thin plasma can be reasonably derived with LECS-MECS spectra, if spectra with a sufficiently high signal-to-noise ratio are available (Favata et al. 1997a; Favata et al. 1998). The results of the one-temperature NEI model fittings obtained in this way are shown in the Table 3.

Table 3. LECS+MECS fitting to single temperature models. The uncertainties are computed using , corresponding to 90% confidence level (Lampton et al. 1976)

Neither a 1T model with abundances fixed to cosmic nor a model with variable metallicity (Z) can describe the data at the 10% probability level. This is also true for the variable metal abundances 1T model and the SW spectrum, while the N spectrum is satisfactory reproduced by this model. In Fig. 5, we show the spectrum at the N position, along with the best-fit model of Table 3 and residuals. Since a proper modeling of the X-ray emission from SW cannot be found in the framework of 1T model, we also fitted the SW spectrum with a two-temperature (2T) NEI emission model, obtaining a bad using abundances fixed to cosmic, but an acceptable using variable metallicity for the two components ( and ). The latter result is reported in Table 4, and indicate very different metal abundances for the two components. In Fig. 6, we show that contours associated to the confidence levels of 68%, 90% and 99% of the two interesting parameters and Z. We have chosen to represent these two parameters to explore the possible interplay between them, since high abundances may compensate for low values. While the  keV component has Z significantly above 1.0, the cooler  keV component has , compatible with the best-fit iron abundances of the  keV best-fit model of the N spectrum. Fig. 7 shows the spectrum at the SW position with the best-fit model. Since the measurement of the high-T component abundances is important to the detection of reverse shocks in RCW86, we have verified that our fit has found global minimum rather than a local minimum in the space. The best-fit value of the ionization time ( in s cm^-3) is rather unusual for a SNR, even though it is consistent with the ASCA results obtained by Vink et al. (1997), and indicate a recent interaction between the shock and its environment.

Fig. 5. LECS (diamonds) and MECS (triangles) spectrum at the N position of RCW86 with the 1T variable abundances best fit model (continuous line) and percentual residuals

Fig. 6. contour plots in the plane for the high-T component (left ) and the low-T component (right ) of the 2T NEI fit with variable metallicities. The 68%, 90% and 99% confidence level contours for two interesting parameters are displayed. The dashed line corresponds to

Fig. 7. LECS (diamonds) and MECS (triangles) spectrum at the SW position of RCW86 with the 2T variable Z best fit model (cont. line)

Table 4. LECS+MECS fitting to two temperature models in the South Western region. The uncertainties are computed using , corresponding to the 90% confidence level (Lampton et al. 1976)

3.2.2. ROSAT PSPC spectra

The PSPC spectra have been collected in regions identical in size and position to the regions used for the LECS and MECS spectra. The background was collected in the same PSPC image, at an off-axis position free from diffuse emission. The spectra have been rebinned to ensure at least 20 counts in each bin. The spectra have been first fitted with a 1T NEI model with metal abundances both fixed to the cosmic values and jointly varying (free Z), resulting in both cases in a too high value. The fact that 1T NEI spectra fail to reproduce the X-ray SNR emission in the PSPC band was also observed for the Vela SNR and there explained with the multi-component nature of the plasma in the interaction regions (Bocchino et al. 1997). In our case, the BeppoSAX fitting results provide us with useful hints about the components present in the plasma, and therefore we have checked whether the model best-fitting the BeppoSAX spectra is able to describe also the PSPC data. Even if the normalization of the thermal components detected with SAX are left free to vary, these fits yielded again an high , suggesting the presence of an additional PSPC soft component not detected in the BeppoSAX spectra. Therefore, we added a thermal Collisional Ionization Equilibrium (CIE) component to the BeppoSAX-derived components, because a second thermal component was also reported by Vink et al. (1997), and we fitted the PSPC spectra. The results, summarized with the other PSPC fitting results in Table 5, shows that the improve noticeably with a very soft component at for the SW spectrum and  keV for the N spectrum. If we fit the LECS+MECS spectra with this supersoft component and the previously derived components with varying normalizations, the derived normalization factors of the soft component are consistent with 0, indicating that only an upper-limit to the PSPC soft component can be obtained on the basis of the BeppoSAX spectra alone.

Table 5. PSPC fitting results on both the SW and N regions. Values with uncertainties (computed with the criterium , Lampton et al. 1976) correspond to free parameters.
Notes:
a) These results have been obtained by fitting a 1T CIE model plus two component with their parameters fixed to the best-fit values listed in the Var. Z column of Table 4.
b) These results have been obtained by fitting a 1T CIE model plus one thermal component with its parameters fixed to the best-fit values derived by the fitting of the SAX Northern rim spectrum, and listed in the Var. Ab. column of Table 3.

As Parmar et al. (1999a) summarize, it is not uncommon that the PSPC detects a soft thermal component with  keV which is not detected in the LECS instrument. This was the case not only for X-ray cosmic background spectrum, but also for the spectrum of some active stars which have been observed with the PSPC and other X-ray and UV detectors (see e.g., Griffiths & Jordan 1998; Brickhouse & Dupree 1998). Parmar et al. (1999a) argue that this might point towards a systematic effect in the low-energy calibration of the PSPC. However, we note that the presence of a supersoft component is in agreement with the idea that RCW86 is expanding in an inhomogenuous medium, as suggested by radiative and nonradiative optical filaments coexisting next to each other (Leibowitz & Danziger 1983; Smith 1997). In this framework, the supersoft component may be the tracer of the interaction of gas with density in between the low density associated with the main shock in the ISM and ejecta, and the high densities associated with the bright optical filaments, as, for instance, is the case for the Vela SNR (Bocchino et al. 2000). Moreover, we also note that the limited spatial resolution of the LECS at low energies (6.1´ at 0.28 keV, 80% encircled power radius) is significantly lower than the resolution of the PSPC (33" at 0.3 keV, 80% encircled power radius). This implies that soft energy photons of the supersoft component are spread over a wide area in the LECS image, and even out of our 8´ LECS extraction region, while this effect is much smaller in the PSPC. This eventually implies a difficulty to detect super-soft components greater with the LECS than with the PSPC.

© European Southern Observatory (ESO) 2000

Online publication: August 17, 2000
helpdesk.link@springer.de