3. Joint spectral fits
3.1. Single power-law fits to 1 keV data
We have used XSPEC for spectral analysis. The BBXRT and ASCA results (Sect. 1) show that the CXRB spectrum between 1-10 keV can be well-fitted with a single power-law with . Thus we first made single power-law fits to the keV data as a simple check for consistency among instruments. The PSPC, GIS and SIS pulse height spectra for keV (PSPC:1.0-2. keV, GIS:1.0-10. keV, SIS:1.0-7. keV) have been fitted with a model of the form , where is the photon intensity in units of and E is the photon energy in keV. In the fit, the two parameters, A (normalization) and (photon index) are varied separately or only A is separate while is joined. Fitting results are summarized in Table 2 (fit id. A1-A4) with 90% errors (). In A1-A4, all free parameters were varied during the error search.
Table 2. Results of the spectral fits (see text for model and parameter definitions)
Confidence contours in the space, considering statistical errors only, for A1 and A3 are shown in Fig. 1. The disagreements among instruments in Fig. 1 show the level of systematic errors. Possible sources of these systematic errors are discussed in Sect. 4. Adding an keV excess component to the model did not change the results significantly.
3.2. Broad-band fits (0.1-10 keV)
We have made joint fits to the overall ROSAT - ASCA spectra over the 0.1 - 10 keV range (0.1-2 keV with PSPC; 0.7-10 keV with GIS and 0.7-7 keV with SIS) considering the following components: (1) an extragalactic power-law component with parameters A and (see above); (2) the hard thermal component, probably associated with the Galactic halo, with plasma temperature [keV] and normalization in the XSPEC convention (per steradian); (3) the soft thermal component, from the Local Bubble with and . Components (1) and (2) are absorbed by the interstellar gas with a hydrogen column density fixed to Galactic values ( [ ], Dicky & Lockman 1990). For components (2) and (3), a Raymond & Smith plasma (distributed as a part of XSPEC), with the solar abundance was assumed. Spectral shape parameters and ratios of normalizations of different spectral components have been joined for all instruments. The overall normalizations have been allowed to vary separately represented by a parameter or , which is the relative normalization to the PSPC value. The pulse height spectra and models are shown in Fig. 2 and fixed/best-fit parameters are listed in Table. 2 (B1, B2). The 90% error in Table. 2 are formal statistical errors, which have been derived with and temperatures fixed at nominal values while all the normalizations are fitted.
The good fits with the thermal components in the PSPC keV data show that the after-pulse (AP) events can be neglected in these observations. Since the ROSAT spectra of resolved X-ray sources, mostly extragalactic, are steeper (, e.g. Hasinger et al. 1993) and a larger flux than inferred from a single power-law fit has already been resolved at keV, we expect a turn up of the extragalactic component below keV, where the hard thermal component also start to emerge. We could also obtain satisfactory fits with a model where the extragalactic component has a break below 1 keV (fixed to at keV) with the main modification of the hard thermal component normalization ( =0.98 and 1.08 for LH and LX respectively). We could not find satisfactory fits with no hard thermal component but with an extragalactic component excess below 1 keV (either by a broken power-law or an addition of a steeper power-law component). This is because of a clear oxygen line feature around 0.5-0.6 keV in the PSPC spectrum (see also Hasinger 1992).
© European Southern Observatory (ESO) 1998
Online publication: May 12, 1998