MATCHED-FILTER SOURCE DETECTION APPLIED TO THE ROAST PSPC AND THE DETERMINATION OF THE NUMBER-FLUX RELATION

We investigate the application, for X-ray astronomy, of a source detec tion algorithm similar to the so-called matched filter (MF) technique. From the instrument point-spread function (PSF) we construct filters which are much closer to optimal for point-source detection than are t he square functions used in the traditional sliding box (SE) method an d at the same time ensure local background subtraction. We develop the technique primarily for use with the ROSAT PSPC images. Through inten sive Monte Carlo simulations we calibrate the method for the real PSPC PSF and the values of background intensity typically found for medium and deep exposure (>10,000 s) ROSAT fields. From the simulations, all those functions necessary to correct the measured source intensities and the number-flux relation (log N-log S) are derived. It is shown th at, indeed, we are able to obtain unbiased estimates of source intensi ties and the log N-log S function down to S/N similar to 2, as long as the PSPC PSF approximation we use is valid. We compare our method wit h the SE technique, as it is traditionally used, and find that the MF algorithm always works significantly better than the SE method. In par ticular, the latter overestimates the source fluxes and the log N-log S function near the detection threshold and therefore requires additio nal calibrations to correct for these biases.