The accuracy of the different background subtraction techniques described in Sect. 7 was investigated using both blank field and exposures containing sources. The blank field exposures are listed in Table 2 and the source exposures in Table 4.
CAL83 and CAL87 are two "supersoft'' sources located in the Large Magellanic Cloud. CAL83 is only detected in the LECS between 0.1-0.6 keV (Parmar et al. 1998) and CAL87 between 0.2-1.0 keV (Parmar et al. 1997a). HZ43 is a nearby hot white dwarf (e.g., Barstow et al. 1995) that is often used as an extreme ultra violet or soft X-ray "standard candle'' in X-ray astronomy. Due to the softness of its spectrum, it is only detected below 0.28 keV in the LECS. These 3 targets were chosen since they are only detected in a narrow energy range of the LECS, allowing the rest of the spectrum to be used to estimate the background subtraction quality.
4U1630-47 is a recurrent X-ray transient located close to the galactic plane. The BeppoSAX observation was designed to detect quiescent emission between outbursts, but none was found (Oosterbroek et al. 1998). X1755-338 is a bright X-ray dipping source which was observed to have turned-off prior to the BeppoSAX observation (Roberts et al. 1996). These two fields were chosen as tests of the background subtraction techniques in complex regions of the sky at low galactic latitudes.
NGC7172 is a Seyfert 2 galaxy, whose nuclear continuum is seen through an atom cm-2 neutral absorber. A soft excess observed by ASCA was associated with diffuse emission from the group HGC90, in which NGC7172 is located (Guainazzi et al. 1998). In addition, two relatively bright (0.25 count s-1) sources are included as a control sample, to show the relative independence of the results on the choice of the background subtraction method at higher flux levels. These are the Seyfert 2 Galaxy NGC 1068 (Guainazzi et al. 1999) and the coronal X-ray source VYAri (Favata et al. 1997).
In Fig. 10, observed LECS count rates in five energy bands (0.1-0.3, 0.3-1, 1-2, 2-5, and 5-10 keV) are shown for the sample sources. No correction for the instrumental response has been applied. All results in this section are displayed in this way to allow ease of comparison since count rate is a quantity which can be easily estimated when comparing the behavior of a weak X-ray source with the instrumental performances and limitations.
|Figure 10: Count rates (CR) for the source sample in the energy ranges 0.1-0.3, 0.3-1, 1-2, 2-5, and 5-10 keV. The counts include contributions from the source (if present) and any extended emission components within the 8 extraction radii. The ordinate uncertainties are smaller than the symbol size|
|Figure 11: Count rate residuals (CR) as a function of energy for 8 blank fields from Table 2, when the 3 background subtracted techniques are applied. Empty squares indicate the standard background, filled circles the semi-annuli, and crosses the ROSAT-scaled techniques|
|Figure 12: Exposure time weighted sum of the 8 blank field exposure residuals when the 3 background subtraction techniques are applied. Empty squares indicate the standard background, filled circles the semi-annuli, and crosses the ROSAT-scaled techniques|
Figure 11 shows the residuals for 8 blank fields when each of the 3 background subtraction techniques are applied. (The shortest exposure blank field, the SDC Target, in Table 2 was excluded.) In the case of a good background subtraction, the residuals should exhibit a Poissonian distribution centered on zero, and the standard deviation of the residuals provides an estimate of the systematic uncertainties associated with each technique. These results are summarized in Fig. 12, where the exposure time weighted sum of the background field residuals, shown individually in Fig. 11, are plotted. The average mean count rates, ,in the 0.1-10 keV energy range are (-1.5 0.9), (3.6 1.6) and s-1, for the standard, semi-annuli and ROSAT-scaled technique, respectively. Table 5 lists the values of ( is the standard deviation) for the 3 techniques in 3 energy ranges. These values may be interpreted as 3 estimates of the systematic uncertainties of each the background subtraction techniques at high galactic latitudes. In the 0.1-2 keV energy range, the values in Table 5 correspond to fluxes in the range 0.7-1.8 and erg cm-2 s-1 for a power-law spectrum with and photoelectric absorption, , of and atom cm-2, respectively. In the 2-10 keV band, for an of atom cm-2, the limiting fluxes are 2-3 and erg cm-2 s-1 for sources with and 1.5, respectively.
In the case of the blank fields, the standard background provides the most consistent (i.e., lowest values of ) subtraction. This is unsurprising given that the standard background itself includes the comparison fields (see Table 2). In addition, the blank field exposures are at high (>) galactic latitude and avoid features such as the North Polar Spur, and so are expected to be broadly similar. The residuals obtained with the semi-annuli method are the largest. An investigation reveals that this may be in part due to uncertainties in the NXB spectra of the two semi-annuli which contain <8800 counts. It is expected that this will improve as subsequent dark Earth exposures are added to the already existing data set. The ROSAT-scaled background works consistently well within the limits specified in Table 5.
|Figure 13: Count rate difference compared to the mean (CR) as a function of energy when the 3 background subtraction techniques are applied to each of the sample sources. Empty squares indicate the standard background technique, filled circles the semi-annuli, and crosses the ROSAT-scaled technique. Note that the ordinate extrema are a factor 2 larger than in Fig. 11|
Figure 13 illustrates the application of the 3 background subtracted techniques to the sample sources. For each source the 3 techniques were applied and background subtracted count rates in the 5 energy bands calculated. The difference with respect to the mean of the 3 techniques is plotted in Fig. 13. In 6 out of 8 cases no significant deviations at levels higher than a few count s-1 are present. This is comparable to the results on the blank sky fields and indicates that all three methods work well here. The differences between the three methods may be used to estimate the systematic uncertainties associated with background subtraction.
In the case of the two fields close to the galactic plane (4U1630-47 and X1755-338), differences at a level 10-3 count s-1 are evident. This is a factor 2.5 larger than with the blank field exposures and is probably due to incorrect estimation of the contributions of the hard diffuse emission associated with the Galactic ridge, or unresolved point-sources. In addition, in the case of 4U1630-47 the 50 mCrab (or 10 LECS count s-1) X-ray binary 4U1624-49 is located 2 away. X-rays from 4U1624-49 that undergo a single mirror reflection may provide a small contribution to the 4U1630-47 field.
In both these cases, the standard background systematically underestimates the background (as expected), and the scaled ROSAT method overestimates the background somewhat, when compared to the mean values. The reason for this is unclear, but it may be partly due to incorrect ROSAT source subtraction in the complex fields. In the case of the X1755-338 field, the residuals obtained using the semi-annuli method are smoothly distributed between those obtained using the other two methods. This is not the case in the 4U1630-47 field where the residuals deviate from the mean values above 2 keV. These results suggest that the semi-annuli gives the most reliable results for complex fields, with the scaled ROSAT method being the second most reliable. These two cases illustrate the difficulty in obtaining a good background subtraction for sources located at low galactic latitudes, or in complex regions of the X-ray sky.
Finally, the effects of the long term decrease in NXB intensity evident in Fig. 5 were evaluated using all three proposed background subtraction techniques. In all cases the differences in background residuals were substantially smaller than the values given in Table 5. This means that currently no time dependent corrections need to be applied to the LECS background.
Copyright The European Southern Observatory (ESO)