Innovations in ACIS Data Analysis
Patrick Broos (Penn State University) , Leisa Townsley (Penn State University)
Fields with thousands of X-ray point sources pose significant data analysis challenges to the Chandra/ACIS observer, ranging from source detection to spectral modeling. We describe three innovations we are exploring as part of the ACIS Team's on-going development of the ACIS Extract (AE) analysis package, which has been publicly available since 2002. (1) Point source detection in crowded fields is difficult. We describe a procedure that first proposes sources by identifying peaks in a simple Richardson-Lucy restoration of ACIS image data, and then carefully computes a confidence level for each candidate source using the AE machinery. (2) Accurate local background estimation is critical for effective source detection and for spectral analysis of weak sources. We describe a model-based approach to background estimation, now available in AE, that addresses the crowded field problem. (3) Perhaps the best approach for modeling the low-count spectra produced by the majority of ACIS sources is simultaneous fitting of the raw (not grouped) source and background spectra using the C-statistic. However, the background model employed by XSPEC (when the C-statistic is selected and a background spectrum is supplied) is known to produce biased results. We describe a simple alternative background model that seems to perform well.