The work presented in the following chapters has made use of the large FOV and high time resolution of the PCA to search for and monitor X-ray binaries in the SMC. The data used come from a key programme in which twice-weekly monitor- ing observations of the SMC bar region are carried out. Galache et al. (2008) give details of this programme and present light curves of the SMC pulsar population at that time. Here I will summarise the key aspects of the programme, the data analy- sis methods and changes made to the reduction and analysis pipelines developed by Laycock (2002) and Galache (2006).
Many different positions have been observed during the programme, though in recent years we have settled on two slightly overlapping positions covering most of the Bar. These positions are depicted in Figure 2.5. The large and small circles are the 2◦ full width zero intensity (FWZI) and 1◦ FWHM fields of view for the two positions. The observations are made in the GoodXenon event mode which utilises the full timing and energy resolutions of the PCA (1µs and 256 channels). Each photon is tagged with a time, energy and the PCU configuration in this mode. The data are extracted in the 3–10 keV range in order to maximise the signal-to-noise ratio (SNR) for a HMXB. Jahoda et al. (2006) present the quantum efficiency of each of the layers in PCU2. At the energy range of interest, the efficiency of layers 2 and 3 are negligible in comparison to layer 1 and so we only extract data from this top layer. In addition to this, PCUs 0 and 1 unexpectedly vented their propane veto layers on May 12th2000 and December 25th 2006 respectively. Though a new
background model was introduced to account for this in PCU0, it was decided not to use data from this PCU after this date. PCU1 is very rarely used for faint sources, though data are also rejected from this PCU on the rare occasion it is active in our observations.
A reduction pipeline was written in C-shell script to provide a fully automated way of producing a cleaned and barycentre corrected light curve from raw RXTE data files. The pipeline uses many of the FTOOLS needed for the standard reduction of RXTE data and was originally written by Silas Laycock (Laycock, 2002). Major revisions were made by Jose Galache during his Ph.D to account for changes in the observatory conditions and reduction methods (Galache, 2006). The key points that are important to the discussion in this thesis are described now. After light curves have been extracted from the raw event files they are binned at 0.01s. They are then background subtracted and barycentre corrected before finally having the count rate corrected for the number of active PCUs. The final light curves are passed through a Lomb-Scargle periodogram to search for pulsations. The error associated with any period found is calculated based on the formula for the standard deviation of the frequency given in Horne & Baliunas (1986):
δω = 3πσN
2N1/2D A (2.1)
whereσN2is the variance of the light curve, N is the number of data points, D is
the length of the data and A is the amplitude of the signal given by:
A = 2
r z0σN2
N (2.2)
where z0is the Lomb-Scargle power. This error analysis is not perfect, but the time
and computer power needed to run Monte-Carlo simulations on over 1500 power spectra has meant that a more sophisticated analysis has not yet been implemented. Further minor revisions to the pipeline were made by Matthew Schurch during his Ph.D (Schurch, 2009). Minor revisions made during my Ph.D include:
• addition of new AO information and adaptation of the ObsID recognition section to account for the most recent data sets and newly devised raster ob- servations to try and locate newly discovered pulsars
within the pipeline meant all data were reduced with the energy channels chosen for epoch 5 data. Epochs prior to epoch 5 were meant to be reduced using different gains as the PCUs are constantly changing.2
• changing the filtering criteria to remove bad data related to PCU breakdowns. If an observation, or part of an observation occurs less than 150 s before or less than 600 s after a breakdown event (flagged with the TIME SINCE BRK keyword) that observation, or part thereof, will be thrown away. The selection expression for PCU2 data, for example, is thus (PCU2 ON == 0k (PCU2 ON == 1 && (TIME SINCE BRK2 <-150k TIME SINCE BRK2 >600))). This expression has been added to the more general selection criteria described in Galache (2006).
Due to the quickly expanding data archive, Galache (2006) developed two fur- ther analysis packages designed to automate the period analysis of the light curves. The first program, PUMA (PUlsar Monitoring Algorithm), produces a Lomb-Scargle power spectrum for each light curve and searches it for periodicities from known pulsars. The typical length of an observation is around 10 ks meaning periods of up to∼ 3000s can be searched for. If it finds a known pulsar, it will measure the pulsed amplitude at that frequency. It will also record bright pulsations from previously unknown systems. The second program, ORCA (ORbital Calculation Application), reads in the pulsed amplitude histories of each pulsar as measured by PUMA and makes a long-term light curve for each one. It then searches the light curves for pe- riodic signatures indicative of orbital modulation. Further details on both programs can be found in Galache (2006) or Galache et al. (2008).