4. Spectroscopy

Next: 4. Discussion Up: 3. Results Previous: 3. The pulse profile Contents

4. Spectroscopy

Figure 3.4 shows the SIS1 fast mode spectrum of the iron line region. The continuum has been modeled with a power law and a broad Gaussian. This broad Gaussian may be due to Comptonization of the 6.4 keV iron line. In any case, a single power law continuum does not fit the data well. The iron emission complex is clearly resolved into three components -- a 6.4 keV fluorescent line, a line at 6.7 keV due to emission by Fe XXV, and a line at 6.97 keV due to emission by Fe XXVI. The best-fit parameters are shown in Table 2. The average GIS spectrum is shown in Figure 3.4. A single power-law continuum did not give a satisfactory fit. The best-fit continuum model had the form

I(E) = e^{- EN_HI}(e^{- EN_Hm}I_mE^- + e^{- EN_Hh}I_hE^- + I_sE^{- (
+ 2)})

(87)

where the interstellar column density, N_HI, was fixed at 1.3 x 10²² cm^-2. The first power law component represents direct emission from the neutron star, the second may be due to reflection from cold matter near the neutron star, and is constrained to have the same photon index. Figure 3.4 shows how this reflected component may arise The third component models interstellar dust scattering and its energy dependence is constrained to be softened by a factor of E^-2 relative to the direct component. This assumes that the Rayleigh-Gans approximation for the differential scattering cross-section holds. Below 2 keV the exact Mie solution for scattering of X-rays by interstellar dust grains is more appropriate. However, due to the fact that the observed intensity falls off quickly below 2 keV it is assumed that the simpler Rayleigh-Gans approximation holds even though it overestimates the scattering intensity (Smith and Dweck (1998)).

**Figure 6:** SIS1 fast mode spectrum of iron line region showing resolved emission lines at 6.4, 6.7, and 6.9 keV.
$\begin{figure}\par\plotfiddle{3lines.epsi}{263.830pt}{270}{55.9}{55.9}{-207.654pt}{308.547pt} \par\end{figure}$

**Table 2:** Observation-average best fit parameters from SIS1 data. Errors are for 90% confidence in one interesting parameter. The energy range has been restricted to 5-9 keV for spectral fitting.
Parameter		Value
$\alpha$	Photon Index	0.68 $\pm$ 0.02
I_P.L.⁶		0.0406 $\pm$ 0.0004
E₁⁷	Fe I	6.377 $\pm$ 0.009
I₁⁸		21 $\pm$ 1
E₂⁹	Fe XV	6.68 $\pm$ 0.02
I₂¹⁰		13 $\pm$ 1
E₃¹¹	Fe XVI	6.97 $\pm$ 0.01
I₃¹²		13 $\pm$ 1
E_broad¹³	Comptonized iron line?	6.33 $\pm$ 0.03
$\sigma_{\rm broad}^{}$	Gaussian width in keV	0.44 $\pm$ 0.03
I_broad¹⁴		47 $\pm$ 2

**Figure 7:** Observation-average GIS spectrum. The iron line region is fitted with three narrow lines at 6.4, 6.67, and 6.97 keV.
$\begin{figure}\par\plotfiddle{obsav.epsi}{263.830pt}{270}{55.9}{55.9}{-207.654pt}{308.547pt} \par\end{figure}$

**Figure 8:** Possible origin of a reflected component in the X-ray spectrum (not to scale). The observer sees direct emission from one polar cap and scattered X-rays from material that is backlit by the other polar cap.
$\begin{figure}\par\plotfiddle{Reflection3.epsf}{190.973pt}{270}{52.8}{52.8}{-183.060pt}{190.973pt} \par\end{figure}$

This model gave a significantly better fit to the observation-averaged spectrum than the model of Ebisawa et al. (1996) in which $\alpha$ was fixed at 1.0.

It is convenient to refer to the three continuum components as the medium, hard, and soft components, respectively, and identify them with the subscripts m, h, and s. To the medium component were added narrow lines at 1.85, 2.01, and 2.64 keV, corresponding to emission from Si XII, Si XIV, and S XVI, respectively. To the hard component was added a 6.4 keV narrow line with centroid energy 6.4 keV. This is due to fluorescence of iron in low ionization stages. This fit yielded an unacceptably high $\chi^{2}_{\nu}$ . The poor fit is due to the presence of residuals around 6 keV. Allowing the 6.4 keV line to have a finite width improved the fit but it was still unacceptable. When a broad Gaussian with centroid energy $\sim$ 6.3 keV was added to the hard component an acceptable fit was obtained as shown in Figure 3.4. This component may be due to Comptonization of the 6.4 keV iron line. The best fit spectral parameters are shown in Table 3.

**Figure 9:** Observation-average GIS spectrum. The model is the same as that in Figure 3.4 but with a broad Gaussian at 6.36 keV
$\begin{figure}\par\plotfiddle{withbroad.epsi}{263.830pt}{270}{55.9}{55.9}{-207.654pt}{308.547pt} \par\end{figure}$

**Figure 10:** Variation of spectral parameters with pulse phase.
$\begin{figure}\par\plotfiddle{pulsephase.epsi}{375.853pt}{270}{74.9}{74.9}{-340.289pt}{412.540pt} \par\end{figure}$

**Table 3:** Observation-average best fit parameters from GIS data. Errors are for 90% confidence in one interesting parameter.
Parameter		Value
$\alpha$	Photon Index	1.29 $\pm$ 0.13
I_medium¹⁵		0.0354 $\pm$ 0.0067
I_6.67¹⁶	Fe XV	3.79 $\pm$ 0.78
I_6.97¹⁷	Fe XVI	3.6 $\pm$ 0.9
I_1.85¹⁸	Si XII	14.8 $\pm$ 4.0
I_2.01¹⁹	Si XIV	8.9 $\pm$ 2.6
I_2.64²⁰	S XVI	4.13 $\pm$ 0.97
N_Hmedium	( 10²² cm^-2)	3.25 $\pm$ 0.24
I_hard²¹		0.0219 $\pm$ 0.0035
I_6.4²²	Fe I	12 $\pm$ 2
E_broad	Gaussian centroid energy in keV	6.34 $\pm$ 0.11
$\sigma_{\rm broad}^{}$ ²³	Gaussian width in keV	0.78 $\pm$ 0.12
I_broad²⁴		28.2 $\pm$ 5.7
N_Hhard	(10²² cm^-2)	25.1 $\pm$ 5.6
I_soft²⁵		0.027 $\pm$ 0.001

In order to examine the iron line equivalent widths the GIS data were fitted with the energy range restricted to 5-8 keV. The continuum model was a power law with iron K-edge absorption. The iron line complex was modeled by three narrow lines with centroid energies fixed at 6.4, 6.67, and 6.97 keV. This model fitted a spectrum extracted from all the GIS high bitrate data well. Pulse-phase averaged spectra were extracted from 24 time intervals. The edge energy and the power law index were fixed at the observation-average value. This yielded acceptable fits to the 24 spectra. The 6.4 keV iron line equivalent width is plotted versus orbital phase in Figure 3.4.

**Figure 11:** Variation of fluorescent iron line equivalent width with orbital phase. Errors are for 90% in one interesting parameter.
$\begin{figure}\par\plotfiddle{eqw390.epsi}{271.782pt}{270}{56.2}{56.2}{-209.452pt}{317.267pt} \par\end{figure}$

Because of the limitations of counting statistics, pulse-phase resolved spectroscopy of these 24 intervals was not feasible. Thus, the observation was divided into five intervals. These are denoted ``eclipse egress'', ``pre-flare non-eclipse'', ``flare'', ``post-flare non-eclipse'', and ``pre-eclipse'' phases. The corresponding orbital phase intervals were $\phi$ = 0.12 - 0.26, $\phi$ = 0.18 - 0.38, $\phi$ = 0.38 - 0.42, $\phi$ = 0.44 - 0.64, and $\phi$ = 0.66 - 0.8, respectively. These spectra were fitted to the continuum model of equation 7.2. The three iron lines were included in the fit. The 6.36 keV broad feature was not required. However, allowing the 6.4 keV line's width to vary improved the fit. The ratio of the 6.4 keV line to the hard power power law intensity and the ratio of the 6.67 and 6.97 keV recombination lines were fixed. Also, $\alpha$ was fixed at the observation-average value. Clear variations in the continuum spectra are apparent in figures 3.4 and 3.4.

**Figure 12:** Pulse-phase averaged spectra at different orbital phases.
$\begin{figure}\par {\centering\leavevmode \vbox to132.319pt{\rule{0pt}{132.319pt... ...t=154.746pt hoffset=-12.895pt vscale=28.0 hscale=28.0 angle=270} }} \end{figure}$

**Figure 13:** Orbital dependence of spectral parameters.
$\begin{figure}\par\plotfiddle{orp.epsi}{350.749pt}{270}{69.9}{69.9}{-306.032pt}{384.985pt} \par\end{figure}$

Footnotes

... I_<#rm#>P.L.⁶: Power law intensity in <#rm#>cts keV^-1 cm^-2 s^-1 at 1 keV
...E₁ ⁷: Gaussian emission line centroid energy in keV
...I₁ ⁸: Line intensity in 10^-4<#rm#> cts cm^-2 s^-1
...E₂ ⁹: Gaussian emission line centroid energy in keV
...I₂ ¹⁰: Line intensity in 10^-4<#rm#> cts cm^-2 s^-1
...E₃ ¹¹: Gaussian emission line centroid energy in keV
...I₃ ¹²: Line intensity in 10^-4<#rm#> cts cm^-2 s^-1
... E_<#rm#>broad ¹³: Gaussian emission line centroid energy in keV
... I_<#rm#>broad ¹⁴: Line intensity in 10^-4<#rm#> cts cm^-2 s^-1
... I_<#rm#>medium ¹⁵: Power law intensity in <#rm#>cts keV^-1 cm^-2 s^-1 at 1 keV
...I_6.67 ¹⁶: Line intensity in 10^-4<#rm#> cts cm^-2 s^-1
...I_6.97 ¹⁷: Line intensity in 10^-4<#rm#> cts cm^-2 s^-1
...I_1.85 ¹⁸: Line intensity in 10^-4<#rm#> cts cm^-2 s^-1
...I_2.01 ¹⁹: Line intensity in 10^-4<#rm#> cts cm^-2 s^-1
...I_2.64 ²⁰: Line intensity in 10^-4<#rm#> cts cm^-2 s^-1
... I_<#rm#>hard ²¹: Power law intensity in <#rm#>cts keV^-1 cm^-2 s^-1 at 1 keV
...I_6.4 ²²: Line intensity in 10^-4<#rm#> cts cm^-2 s^-1
... $\sigma_{\rm broad}^{}$ ²³: Line intensity in 10^-4<#rm#> cts cm^-2 s^-1
... I_<#rm#>broad ²⁴: Line intensity in 10^-4<#rm#> cts cm^-2 s^-1
... I_<#rm#>soft ²⁵: Power law intensity in <#rm#>cts keV^-1 cm^-2 s^-1 at 1 keV

Next: 4. Discussion Up: 3. Results Previous: 3. The pulse profile Contents

Damian Audley
1998-09-04