The Comptonised emission

In Figure 4.5 I display how the photon indexΓ evolves with the total unabsorbed luminosity. The shape is remarkably similar to the HID (Figure 4.2), and indicates well how the hardness of the X-ray spectrum is not solely due to the influence of the thermal emission from the accretion disc. The blue points in Figure 4.5a mark the hard state, wherebyΓ mainly lies in a narrow range of 1.5–1.7, and appears to remain rather constant despite the hard state spanning over two orders of magnitude in luminosity. However, Figure 4.5b separates the periods of rise (blue) and decay

4.4 Analysis and results 123

Figure 4.4: Fractional RMS versus spectral hardness displaying how the interme- diate and soft spectral states were determined (§4.3.3). The SIMS-A are a cluster of points with low variability (fractional RMS<5 %) and a relatively high spectral hardness (>0.175 %), which I propose as a distinct spectral state (see §4.3.3 and §4.5.2). The hard state was determined by the RMS-intensity relation (Figure 4.3).

(red), and displays quite clearly that above LX∼ 1037.5 Γ tends to become softer

as the source rises up the hard track, consistent with an increase in seed photons to cool the corona (see Done et al. 2007 and references therein). Furthermore, as the source reaches the hard state in decay the photon index is clearly softer than the rise at that luminosity, appearing to retain a softer slope towards quiescence (although see also Stiele et al. 2011). At the lowest luminosities the rise also appears to be softer (see also Sobolewska et al. 2011); however, as a result of the data selection criteria, only two points lie at LX<1036.5limiting any strong conclusions that can

1 2 4 C h ap te r 4 . X -r ay re fl ec tio n th ro u g h o u t th re e o u tb u rs ts o f G X 3 3 9 − 4

Figure 4.5: The total source luminosity plotted against the power-law photon indexΓ. Left: Points coloured by their respective state. Right: The same diagram, however, the soft state, SIMS and SIMS-A observations are now grey, with the HIMS and hard state separated according to the rise or decay phase of the outburst. Observations are plotted randomly to avoid any visual bias. The y-axis is plotted in units of erg s−1.

4.4 Analysis and results 125

During the state transition the photon index undergoes very distinct softening, eventually reaching Γ ∼ 2.5. In fact, the bulk of the evolution appears to take place in the HIMS, rather than the SIMS, which itself is consistent with the val- ues recorded in the soft state. The soft state shows some scatter (2.4–2.9), but is ultimately dominated by large confidence intervals due to the diminished signal in the hard band. The SIMS-A are also typically steeper than the SIMS (see also Table 4.4). The decay phase (Figure 4.5b) displays clear hardening as the source makes its way through the soft-hard transition. In Figure 4.5 I plotΓ against the total lumi- nosity calculated between 0.1 and 1000 keV, which accounts for the non-negligible disc flux below 3 keV, therefore displaying the source decay through the transition not so apparent in the 3–10 keV HID (Figure 4.2). This then serves to exhibit clearly how the source hardens monotonically with luminosity towards the hard state.

Throughout this investigation, I model the Comptonised emission with a cut- off power-law fixed at 300 keV, in order to remain consistent with the assumed illuminating spectrum in the reflection model (Garc´ıa et al., 2013). However, as part of my routine, I initially allowed the high energy cut-off to be a free parameter as it has been shown to be significantly lower in the brighter phases of the hard state and subsequent transition (Motta et al., 2009). I allowed the cut-off to be free within the range 10 to 300 keV; however, should it converge on a value lower than 15 keV, or higher than 250 keV, respectively, I deemed the fit to be in error and fix the parameter to be 300 keV. I took this action because below 15 keV the cut-off tended to be fitting the curvature of the disc component, rather than the power-law. Above 250 keV the effect of the cut-off becomes insignificant due to the upper limit on the bandpass of the HEXTE instrument (100 or 200 keV; §4.3.2).

In Figure 4.6 I plot the time-evolution of the high energy cut-off in the 2002 and 2007 outbursts. No cut-off was resolved in any of the 2004 observations; however, the transition luminosity in this outburst was lower than any of the 2002 and 2007 observations which did. Time-zero in Figure 4.6 marks the first observation where the cut-off was determined freely and displays a gradual decrease in energy as the source continues its rise through the hard state. All observations intermediate to the respective first and last resolved observation in Figure 4.6 were able to fit the parameter freely. Furthermore, the final three observations in each outburst took place in the HIMS, confirming this trend continues into the state transition. The high energy cut-off is thought to represent the temperature of the electrons in the corona, hence meaning the power-law emission symbolises thermal Comptonisation. The softening of the photon index and decreasing high energy cut-off are both consistent with increasing amounts of seed photons cooling the corona.

126 Chapter 4. X-ray reflection throughout three outbursts of GX 339−4

Figure 4.6: Evolution of the high energy cut-off with time. Zero marks the first observation where the cut-off was detected (MJD 52372 and 54122 for the 2002 and 2007 outbursts respectively). No cut-off was resolved in the 2004 outburst observations.

The trend I have found is very consistent with that found by Motta et al. (2009) using the same datasets; however, they resolve the cut-off over a longer period than us all the way into the soft state. In this study I am more focused on the reflection, hence to maintain a reasonable timescale for the fitting routine I did not apply such stringent and detailed criteria for detecting the cut-off, hence I refer the interested reader to the work of Motta et al. (2009) for a more in depth study of the cut-off.