Analyzing the Spatial Distribution of Piled Sources Glenn E. Allen May 23, 2002 Pileup: The effects of the pile up of more than one X-ray event in an ACIS event-detection cell (i.e. a 3 pixel x 3 pixel island) has been shown to be a feature for moderately bright point sources and, perhaps, some very bright features in extended sources. Pileup affects both the measured X-ray spectrum and the spatial distribution of the photons from a source. Spectral models for moderately piled sources have been implemented in Sherpa, ISIS, and XSPEC. However, there is no means of modeling the spatial distribution of a moderately piled source. Piled PSF analyses: There are two potential techniques that could be used to analyze the spatial extent of a moderately piled source: (1) a detailed event-by-event detector simulator or (2) a parameterized model that incorporates the statistical properties of an ensemble of source events. An outline of how each of these analysis techniques might proceed is listed below. 1. Event-by-event simulator One scenario for such a simulator is as follows Spectral model for source (e.g. the results of fitting a piled spectrum) | | V Ray trace code for the HRMA (e.g. SAOSAC, MARX; requires accurate calibration of the mirrors) | | V CCD event simulator (e.g. MIT or PSU CCD simulators; requires accurate calibration of the QE, gain, and response and an | accurate model of CTI, all of which | are time dependent; may require an V accurate model of the various backgrounds) Event detection algorithm (This algorithm must match the one used onboard and may require use of | the bias maps for the observation.) | V Level 1-like file (The event grading algorithm and PHA sum rules would have to match the | algorithm and rules used for the | observation being simulated.) V Filter, select, and analyze (This process would have to be data identical to the process used for the observed data.) 2. Parameterized model An alternative approach is to use a model that incorporates the statistical properties of a piled source. Spectral model for source (e.g. the results of fitting a piled spectrum) | | V Ray trace code for the HRMA (e.g. SAOSAC, MARX; requires accurate calibration of the mirrors) | | V Statistical assignment of (This would require an accurate pulse heights and coordinates parameterization of distribution of charge for a given source spectrum.) | | V Level 1-like file (The event grading algorithm and PHA sum rules would have to match the | algorithm and rules used for the | observation being simulated.) V Filter, select, and analyze (This process would have to be data identical to the process used for the observed data.) Limitations: I expect that neither technique would work for extremely piled sources for two reasons. If a source is piled enough to burn out the core of an image of the source, then it is difficult to obtain an accurate model of the spectrum of the source. If the core is missing, then modeling the spatial distribution of the events would rely on fitting the relatively poorly calibrated winds of the PSF. Work Effort: 1. Event-by-event simulator The best existing ray-trace code is SAOSAC. I believe a substantial effort would be required to test and tweak the PSF for unpiled off-axis sources before one could begin to hope to use it as part of a pileup simulator. Both PSU and MIT have CCD simulators. The PSU simulator has been used to model the spectral effects of pileup. While the PSU simulator does include a model of the spatial distribution of charge for piled (or unpiled) events, to the best of my knowledge, no effort has been made to ensure that the PSF of simulated events (either piled or unpiled) is consistent with the PSF of the Chandra telescope. Use of this simulator (or the MIT simulator) would most likely require an enormous effort by the team developing the simulator and by the calibration team to make sure that the gain, QE, response, and affects of CTI are treated appropriately. I know that these things are not entirely accurate at the moment and expect that many months (or perhaps years) of effort on the part of several people would be required to perform this work. The remainder of the simulation process is relatively trivial. 2. Parameterized model The comments made about the ray-trace code above apply here too. MARX has a model of pileup that approximates the spectral effects of pileup, but these effects are not entirely consistent with the actual spectral effects of pileup and no effort has been made to simulate the effects of pileup on the spatial distribution of events. The simulation of the mirrors in MARX is also an idealization of the actual behavior of the mirrors and is probably not appropriate off axis. A substantial effort (at least several months) would be required to marry SAOSAC (a better mirror simulator) to the code and to determine how to parameterize the distribution of charge for piled and unpiled events. Use of either technique would require a large amount of work. Given the limited resources of the CXC staff, it would be useful to determine what portion of the user community would benefit from a model of the spatial distribution of piled sources and prioritize the project accordingly Glenn