4. The HRI case

Given its generality, the algorithm is portable to other detectors. In the case of the ROSAT HRI, we have used the same cell geometry, but different choices of and . We adopted the PSF given by David et al. (1995) and we have verified that the highest value of usable for count extraction purposes is 0.85. For greater , the calculation of the cell radius involves integration of the tails of the PSF rather than the core, and since the tail is less well known than the core, that would yield high uncertainties in the assessment of the cell's size.

We have performed simulations similar to those described in the PSPC case. We verified that the screening algorithm is more efficient when tailored on a template source lying at 15 off-axis in the HRI FOV. Since the HRI FOV is , this result is therefore similar to the PSPC case, in which the optimal template source lies at 30 off-axis in a radius FOV.

We have also verified that the screening procedure yields very low gain on detected sources (%). The corresponding rejection percentages are also negligible, usually between 1 and 3% of the total exposure times. We stress that this result is not due to low background counting statistics in the detection cell, because the ratio between the number of source and background counts in the cell is in both PSPC and HRI typical observations. Other explanations of the low HRI screening efficiency could be:

1. The HRI background is dominated by the particle component. In particular, the peaks in the HRI background light-curve are due to high particle rate intervals. The peak amplitude, however, is mostly only a factor of 3 higher than the median background value, with few cases in which it is a factor of 10 higher (Fig. 3). Instead, in the PSPC case, the peaks are mostly a factor of 10 higher than the median, with few cases in which they are only a factor of two higher. Our algorithm, therefore, tends to exclude more time intervals in the PSPC case rather than in the HRI case, because the HRI background light-curves are relatively less contaminated by spikes than those of the PSPC.
2. The HRI spikes are broader than the PSPC ones, because they are due to the particle component which is just a function of the sub-satellite position and does not change as rapidly as the SB PSPC spikes.

We have carried out an additional set of simulations to investigate which of our proposed explanations are correct. We have simulated an HRI background light curve as a Gaussian spike superimposed on a constant level, and we run the screening procedure varying the Gaussian amplitude and sigma one at a time. The screening efficiency turned out to be rather insensitive to the width of the spikes, but very sensitive on the ratio between the Gaussian amplitude and the constant background level. In particular, the screening gain is very low when the ratio is comparable to that observed in typical HRI images, and it yields a not negligible gain (>1%) when the ratio is at least twice as observed, as usually occurs in PSPC observations.

  
Figure 3: Same as Fig. 2 (click here) but for the background light curve of an HRI field pointed towards the NGC 2422 open cluster (38075 s)

Figure 3 (click here), left panel, reports the light curve of an HRI case study, a sequence pointed toward the NGC 2422 open cluster with detected sources (RH 201828, Barbera et al. in preparation).

After inspecting a number of HRI real background light-curves, we conclude that the screening of HRI data is not a crucial point in HRI data analysis, but it could likely be very effective when the particle and/or the SB spikes heavily contaminates the data. This may occur either when the spike amplitudes are large (like in PSPC) or when the mean background level is low (this may be the case of the AXAF detectors).