2005ApJ...629..172N

We use XMM-Newton blank-sky and closed-cover background data to explore background subtraction methods for large extended sources filling the EPIC field of view, such as nearby galaxy clusters, for which local background estimation is difficult. In particular, we investigate the uncertainties of the background modeling in the 0.8-7.0 keV band that affect cluster analyses. To model the background, we have constructed composite data sets from the blank-sky observations and compared them to the individual blank-sky observations to evaluate the modeling error. Our results apply to data obtained with thin and medium optical filters and in full frame and extended full frame modes. As expected, the modeling uncertainty is determined by how the EPIC background flares are filtered. We find that to keep this uncertainty tolerable, one has to use a much more restrictive filter than that commonly applied. In particular, because flares have highly variable spectra, not all of them are identified by filtering the E>10 keV light curve. We tried using the outer part of the EPIC field of view for monitoring the background in a softer band (1-5 keV). We find that one needs to discard the time periods when either the hard-band or the soft-band rate exceeds the nominal value by more than 20% in order to limit the 90% CL background uncertainty to between ±5% at E=4-7 keV and ±20% at E=0.8-1 keV, for both MOS and PN. This compares to a 10%-30% respective PN uncertainty when only the hard-band light curve is used for filtering, and to a 15%-45% PN uncertainty when applying the commonly used 2-3 σ filtering method. Adding such a soft-band filter on average results in only a 5%-10% reduction of the useful exposure time. We illustrate our method on the nearby cluster A1795. The above background uncertainties convert into systematic temperature uncertainties between about ±1% at r=3'-4' and ±20%-25% (about ±1 keV for A1795) at r=10'-15'. For comparison, the commonly applied 2-3 σ clipping of the hard-band light curve misses a significant number of flares, rendering the temperatures beyond r=10' unconstrained. Thus, the background uncertainties do not prohibit the EPIC temperature profile analysis of low-brightness regions, such as outer regions of galaxy clusters, provided a conservative flare filtering such as the double-filtering method with ±20% limits is used.