Simulating ACIS Pileup with marx

Confusion is a word we have invented for an order which is not understood.

– Henry Miller


In this section, we will discuss how to use the marxpileup tool in marx to simulate the effects on Chandra data of photon pileup in the ACIS CCDs. As a post-processing module, the marxpileup tool is designed to work with existing marx simulations. Users first create a simulation using marx Running marx simulations and then, if necessary, run pileup on the results of that simulation to study pileup effects. In this manner, users simulating faint or extended sources which are less susceptible to pileup can produce marx simulations more quickly. We will briefly describe the pileup process itself and then proceed to outline the use and output of the pileup tool.

What is ACIS Pileup?

At some level, all Chandra observations performed with the ACIS imager will suffer from the effects of pileup. Pileup occurs when two or more photons land in the same pixel location in a given ACIS readout time. In this situation, ACIS “detects” a single photon with an energy which is roughly the sum of the two component photons. Some simple schematic representations of such “piled” photons are shown below. The pileup process can affect ACIS data in a number of ways including:

Photometric Inaccuracy
The event detection algorithm cannot distinguish between single, large pulse height events and composite, piled events. Consequently, the detected count rate will be reduced with respect to the true count rate in the absence of pileup.
Spectral Distortion
By combining multiple incident, photons into a single “detected” event with a larger pulse height, pileup effectively “hardens” the observed ACIS spectrum.
Point Spread Function Distortion
The severity of the pileup effect is governed in part by event density. Since the core of the PSF has a higher event density than the wings, it will be affected to a greater degree. This effect will tend to broaden the PSF as the ratio of core to wing events decreases.
Grade Migration
As the degree of pileup increases, the distribution of event grades will change. Multiple photon or “piled” events will tend to have “bad” grades which include detached corner pixels, such as ASCA grades 1,5, and 7. This migration will have repercussions for standard data analysis which often begins by discarding such “bad” grades.
Two events (symbolized by red and blue bars in a bar chart) are stacked on top if each other.

A schematic representation of a “piled” event. In this simple illustration, two events (red and blue) are detected in the same location within an ACIS frametime. The pulse heights are shown here in units of keV. Note that the magnitude of the “piled” distribution is essentially the sum of the two component distributions. Hence, instead of detecting two roughly 3 keV photons, we would have detected a single 6 keV photon.

Schematic representation of a  "split"  event

A schematic representation of a small portion (8x8 pixels square) of a CCD in which two events have been detected. The highest amplitude event (blue) is the pixel of interest. Since the standard event-detection routine classifies events in a 3x3 square pixel neighborhood, rates in all other pixels marked in blue must also be considered. In addition, some events two pixels away, like the red event, could pile via splitting onto the event of interest. In this example, both events deposit electrons in more than one pixel - the blue event in two pixels and the red event in three. The peak of the blue event has the highest signal and this would define the “center” of the 3x3 pixel island that is transmitted to the ground. Clearly, the blue event would be assigned too much charge, in this case. Thus, the rates in pixels marked in yellow must be considered when assessing the pileup rate in the blue event (but not all events in this border will actually pile with the blue event; the actual number depends on the details of the branching ratios into each grade).

Chandra observations of bright point sources are the most likely to be significantly affected by pileup. Extended or faint sources will be less affected although sharp features such as unresolved cores or bright knots or filaments could still be vulnerable. Although pileup cannot be avoided entirely, a number of techniques can be employed to mitigate its effects somewhat. For a more detailed discussion of pileup, the reader is referred to acis_pileup.

Overview of the pileup Tool

The marxpileup tool implements the pileup algorithm developed by John Davis (MIT). This same algorithm has been implemented into the ISIS, SHERPA, and XSPEC spectral fitting packages. While this implementation of the pileup algorithm emulates most of the qualitative effects of ACIS photon pileup, users should keep in mind that we are still calibrating the procedure. The ACIS pileup model is statistical and is not an a priori photon-silicon interaction model which generates charge clouds and then PHAs per event “island.” The model is valid on-axis for point sources for low to moderate pileup. While valid for qualitative predictions of the effects of pileup on the PSF, it has not been verified for image reconstruction. Detailed studies of the effects of pileup on the HRMA PSF including comparisons to actual on-orbit data are still underway. The model is very good for spectral modeling of light to moderately piled point sources. Users should interpret all results including the effects of pileup cautiously.