Generales

To begin, I analysed each observation individually, removing the Fe line region from the fit (5–8 keV) in order to accurately estimate the continuum emission upon which the reflection is superimposed5. At this point I wanted to test two properties of the continuum: (1) Whether an absorbed power-law is an acceptable description of the continuum, and (2) that the description of the continuum is consistent with the later analysis when the reflection was modelled. It is imperative that the con- tinuum is correctly modelled since the shape of the reflection spectrum is strongly dependent upon it, hence I used this initial analysis as a sanity-check for the later sections. To this end, the motivation for fitting each observation individually in this section was that it allows me to extend the broadband coverage. The main focus, the EPIC-pn and XIS cameras, will only be covering the 4–10 keV region to constrain

5_{The choice of 5–8 keV, which is different to the 4–7 keV region used in Chapter 2, was chosen}

to best suit the broad profile shown later in Observation 4. If 4–7 keV was used, then a strong residual remained above 7 keV. In addition, since much of this study uses a short 4–10 keV bandpass, the 4– 5 keV portion is necessary to well constrain the soft part of the continuum. I note that this could impact the results, since the Fe line profile may well extend to 4 keV; however, later I investigate a larger bandpass down to ∼ 1 keV (§3.3.4), and show that the Fe line profile is too narrow to affect the continuum below 5 keV, meaning that the choice of 5–8 keV does not affect these initial continuum fits, or the profiles displayed in Fig. 3.3.

80 Chapter 3. The truncated and evolving inner accretion disc of GX 339−4

the continuum for the majority of this investigation. I added simultaneous PCA- HEXTE observations to the EPIC-pn, and made use of the PIN on-board Suzaku to extend the coverage up to 100 keV and 50 keV respectively. A constant normal- isation factor must be allowed between each detector; however, the only one well calibrated enough to be fixed is between the XIS and PIN (both of the Observations were taken in the XIS nominal position, hence I fixed the constant to be 1.16). The cross-normalisation between the PCA, HEXTE and EPIC-pn are all uncertain, so the PCA constant was fixed to be 1 and the EPIC-pn and HEXTE datasets were allowed a free constant. These were ∼0.85 and ∼0.8 for the EPIC-pn and HEXTE respectively. I note again that I fit each epoch individually in this section, since, given that there were 5 detectors from three separate missions, simultaneous fitting would potentially render the results degenerate if any parameters were tied between the datasets.

I fitted a single absorbed power-law to the whole broadband spectrum, tying the photon index and normalisation parameters between detectors, whilst allowing a constant to float between them as described before. The continuum model was found to be a good fit to all of the observations (Table 3.4), and no cut-off was re- quired at high energies, although this may exist beyond the upper energy limit. A significant smeared edge was required for Observation 4 suggesting a large amount of reflection was present. The majority of residuals lie beyond 10 keV, the likely source of which is a Compton ‘hump’ from the reflection of hard X-rays by the cool accretion disc. Interestingly, this was not present in Observations 1 and 2, sug- gesting that the level of reflection increased in the higher luminosity observations. Later, in Chapter 4, I show that the level of reflection does in fact increase at higher luminosity in the hard state. Given the good description of the continuum by this model I used it as the base continuum in the later sections unless stated otherwise.

For now I only consider the hard state spectra. I display the continuum data to model ratio with the Fe line region added back into the plot (i.e. ‘noticed’ in XSPEC) for the hard state observations in Fig. 3.3. Immediately one can see a distinct evolution of the Fe line region. At higher luminosities the profile extends further in both the red and blue wings, and the peak appears to shift to higher ener- gies. A higher spin and inclination will broaden the red wing and blueshift the peak respectively. However, these two parameters will be constant between observations so I can rule out their influence.

Two remaining variables can increase the profile’s extension to lower energies. Were the inner disc radius to change, specifically to extend closer to the BH, the relativistic effects would increase and an increase in the broadening of the red wing

3.3 Analysis and results 81

would be observed. Additionally, the emissivity can generate this effect. More cen- trally concentrated emission would mean a larger contribution to the profile by the regions experiencing stronger relativistic effects, so therefore increased red wing broadening. However, the emissivity is not expected to vary so intensively between hard state observations, given that previous investigations have yielded q ∼ 3 (the emissivity is defined to scale as r−q; Table 3.1). Thus, inner radius variation should dominate the red wing variation observed. The shift in the peak of the profile is al- most certainly due to an increase in the disc ionisation. A more ionised disc means emission from higher rest energies, hence a shift in the peak. Also, larger ionisa- tion results in increased Compton scattering and emission from multiple ionisation stages, which can both contribute an overall broadening of the profile (Garc´ıa and Kallman 2009; see also Garc´ıa and Kallman 2010 and Garc´ıa et al. 2011), thus extending the blue wing.

Therefore, inspection of Fig. 3.3, just by eye, implies that at higher luminosities in the hard state the accretion disc is extending closer to the BH, and becoming more ionised.