Astron. Astrophys. 362, 711-714 (2000)

2. Data analysis

Our analysis is based on archival ROSAT PSPC and HRI data and on an observation of RCW86 by the Einstein HRI. Both HRI instruments have a similar spatial resolution of 4" FWHM, but the ROSAT HRI is more sensitive. The PSPC has a resolution of only 30" FWHM, is more sensitive than the HRI and has better spectral capabilities covering the energy range of 0.1 to 2.4 keV with a spectral resolution at 1 keV. The data used for our analysis are summarised in Table 1. The starting point of the analysis are the basic screened event lists. Further processing, like photon extraction and barycentric correction, was done with NASA's ftools v4.2 package.

Table 1. A summary of ROSAT and Einstein observations which cover the point source. Dates refer to the start date of the observation.

2.1. Position of the source

In order to find an accurate position for the unresolved source we used the HRI data and fitted the point spread function to the source using the very sensitive maximum likelihood fits (Cash 1979and e.g. Hasinger et al. 1994). We used a field of 80" 80" roughly centred on the source. The normalisation was coupled to the background level using , being the observed number of photons and , the fitted background and source normalisations. The results of these maximum likelihood fits are listed in Table 2, which also lists the source parameters for the PSPC data as found with a wavelet analysis (Damiani et al. 1997). One of the HRI detections has a significance level of , not good enough to claim a detection by itself, but combined with the other detections the source parameters can be regarded as meaningful.

Table 2. Best fit positions and count rates of the point source. The first three rows list the result of maximum likelihood fits to the ROSAT and Einstein HRI data. The statistic has a distribution with three degrees of freedom. Position errors do not include systematic errors and correspond (90% confidence regions). The PSPC positions and count rates were estimated with a wavelet analysis method (Damiani et al. 1997) using the energy channels 20-200 ( keV).

Unfortunately, positions based on ROSAT sometimes suffer from errors in the attitude calculations which are typically 6" (Hasinger et al. 1994and private communication). The observed scatter in the positions based on the Einstein and ROSAT HRI images, as compared to the statistical position errors, suggests that also our results are affected by systematic errors. Einstein HRI observations are less affected by systematic position errors, the typical systematic position error being 2" (Van Speybroeck et al. 1979). Adding statistical and systematic errors in quadrature the weighted average of the positions based on HRI images is 14h 41m 51.42s and ^o36´ 12.9" (J2000) with a position error of approximately 3". For a two dimensional gaussian this translates into a 95% confidence radius of 5". Note that a very bright, unresolved, radio source inside the remnant with coordinates and ^o34´ 47" (J2000) is clearly not associated with the unresolved X-ray source (Dickel et al. 2000).

2.2. Spectral analysis

For the spectral analysis of the PSPC data of the unresolved source we extracted photons using a circular area with a radius of 32" for the SW (on-axis) pointing and 44" for the other pointings. We estimate that with such radii we cover roughly 90% of the photons coming from the unresolved source (c.f. Hasinger et al. 1992). Background spectra, extracted from an annulus around the source, were appropriately scaled and subtracted from the source spectra. The combined spectrum consists of 177 net source counts. The spectrum was rebinned in order to have at least 15 counts per bin.

For our spectral analysis we used he spectral fitting program SPEX ¹ (Kaastra et al. 1996). Since we want to know whether the source qualifies as the potential stellar remnant associated with RCW 86, we fitted the spectrum with several emission models both with the interstellar absorption value fixed at  cm^-2, the typical absorption value for the X-ray emission of the supernova remnant (Vink et al. 1997), and with as an additional free parameter. The results are listed in Table 3. The best fit values of for all models seem to be in favor of a low absorption column towards the source, but also models with fixed give acceptable reduced values. The fact that models with three parameters result in very low reduced values (i.e. far from the expectation value), suggests that the statistics of the data is not really good enough to fit models with three or more parameters. All models provide reasonable fits to the data with only the thin plasma model with solar abundances and fixed having a reduced substantially larger than 1. The spectrum appears to be rather soft as indicated by the steep power law index, , and the low black body temperature.

Table 3. Results of the spectral fits to the PSPC data. The luminosities are normalised to a distance of 1 kpc. For each model the spectrum was fitted with  cm^-2 and with as a free parameter. Error ranges correspond to , or 90% confidence limits.

2.3. Timing analysis

We searched the four PSPC observations for possible pulsations using the Rayleigh method (Buccheri et al. 1983). This method is one of the most sensitive methods and it does not involve any binning of the data. A sensitive method is needed as the longest PSPC observations yielded only 101 events. We searched in each set for pulsations in the period range 0.02 to 300 s, sampling the frequency range with step of with the total length of the observation. We compared the periodograms to look for peaks showing up in two or more periodograms at or near the same period. Such correlations were, however, not found. The peak values of the Rayleigh statistic, , imply an upper limit to the pulsed fraction of 20%.

As for the variability on the timescales of month, at first sight there is little evidence for variability as all measured PSPC count rates are consistent with a count rate of  cnts s^-1. However, if we convert the Einstein and ROSAT HRI count rates to PSPC count rates using the best fit power law model in Table 3 (the conversion factors are 4.9 and 2.7, respectively) we get the following PSPC count rates (in the same order as in Table 2):  cnts s^-1,  cnts s^-1, and  cnts s^-1. The dependence of the ROSAT HRI/PSPC conversion factor on the chosen model is small (9%), but the conversion factor for the Einstein HRI count rates is more model dependent, varying from 4.6 to 8.0. Even taking into account the model uncertainties it is clear that the observations are not consistent with a constant source count rate, although it is a strange coincidence that low count rates were only observed by the HRI instruments. Source contamination with the PSPC instrument seems unlikely, as no other unresolved sources are seen with the HRI instruments near the point source. Therefore, the X-ray source is very likely variable on a time scale of months to years.

Fig. 1. RCW 86 as observed by the PSPC (SW pointing). The circle indicates the position of the point source.

Fig. 2. A 50" 50" field image taken from the Digital Sky Survey 2. Overplotted is a circle with a radius of 5" centered on the point source and roughly corresponding to a 95% confidence region.

Fig. 3. The PSPC spectrum of the unresolved source in RCW 86. It consists of a combination of 4 individual observations. The solid line shows the best fit black body model with fixed to a value of  cm-2.

© European Southern Observatory (ESO) 2000

Online publication: October 24, 2000
helpdesk@link.springer.de