Spectral fits to the 0.5-10keV band

We proceed to discuss spectral fits to the full 0.5-10keV band. This requires us to model the soft X-ray spectral features, a non-trivial task since the basic physics determining the soft X-ray spectrum (i.e., absorption vs. emission) is still the subject of much debate. However, since this work restricts itself to the medium resolution EPIC data, it is sufficient to model the soft X-ray complexity with phenomenological absorption/emission components (as oppposed to constructing physical models of the absorber/emitter).

In this work, we employ two phenomenological models of the soft X-ray complexity. In the first model (which we shall refer to as the ``pure warm absorber''), we describe the soft X-ray structure as three simple absorption edges, with threshold energies and maximum optical depths that are left as free parameters in the spectral fits. Since these edges are not of physical interest in this work, we do not report their best fitting values in Tables 1-2. Typical threshold energies (and maximum optical depths) are $E\approx 0.73$keV ($\tau\approx 0.70$), $E\approx 0.85$keV ($\tau\approx 0.45$), and $E\approx 1.0$keV ($\tau\approx 0.15$). These correspond closely with the expected absorption edges of OVII, OVIII, and NeIX/MgX. Blends of oxygen resonance absorption lines, as well as the $L_3$-edge of neutral iron contained within dust grains embedded within the warm absorber, may also contribute to the first of these three edges.

The second soft X-ray model (the ``absorber$+$emitter'' model) consists of these three absorption edges plus relativistically-smeared soft X-ray emission lines of NVII and OVIII, with rest-frame energies of 0.50keV and 0.65keV, respectively. These recombination lines are broadened according to the Kerr black hole accretion disk model of Laor (1991). Both the inclination and emissivity index describing the profiles of these soft X-ray recombination lines are allowed to vary as free parameters of the fit; in order to maintain generality, we do not fix them to be the inclination and emissivity index of the smearing function applied to the reflection features. Although we do not report them, typical best-fitting equivalent widths for the NVII and OVIII lines are 0eV (i.e., the line is not required) and 80eV, respectively. Typical inclinations and emissivity indices characterizing the soft emission lines in our fits are $i=38^\circ$ and $\beta=4.3$, respectively. We do not include any carbon recombination lines (which have also been discussed by Branduardi-Raymont et al. 2001, Sako et al. 2002 and Mason et al. 2003) since they all lie at energies below our low-energy cutoff.

From Table 1 it can be seen that the qualitative conclusions of the hard-band fits are robust to the inclusion of data down to 0.5keV. Recalling that the partial-covering model fails when applied to the joint pn-PCA data, one can see that the only adequate fit is given by the relativistic ionized accretion disk model with a large emissivity index ($\beta\sim 5.5$) and a small inner radius ($r_{\rm in}\sim 1.5r_{\mathrm g}$). The principal difference between the 2-10keV and 0.5-10keV fits is the slightly lower inclination of the latter ($i=44\pm 6^\circ$ versus $i=56\pm 4^\circ$).