ACIS Pileup

The AXAF CCD Imaging Spectrometer (ACIS) will offer the user a sensitive X-ray camera with a dizzying array of capabilities and observing modes. But unlike optical CCD cameras, ACIS is fundamentally a photon-counting device. This radically alters the application of its CCD detectors to imaging spectroscopy problems. Indeed, X-ray imaging spectroscopy of any bright, compact source with ACIS alone (i.e., without transmission gratings in the beam) will pose a challenge to the prospective AXAF observer. The reason is at once simple and complex: photon pileup.

  
Figure 15: Event pulse height spectra illustrating HRMA+ACIS pileup measurements conducted at XRCF. The energy of the source is 1.49 keV (Al K ). From the top to the bottom panel, source flux increases by about an order of magnitude, from ~ 0.3 to 3.0 counts per frame; these data were obtained with a subarray frame time of about 0.1 sec. The spectral line registered by single Al K photons appears at ~380 ADU in each panel, while the peaks that appear at integral multiples of this amplitude are pileup peaks (consisting of 2 or more Al K photons detected quasi-simultaneously). The intensity of the pileup peaks increases relative to that of the source peak with increasing source flux. (Source: Proposers' Guide)

Pileup (defined here as two or more photons landing in the same or neighboring pixels during a single CCD exposure) will affect many types of ACIS experiments. Any compact or point source for which spectroscopy can be readily conducted with ACIS (via its intrinsic energy resolution) will also be bright enough that pileup will be a problem. That is, since count rates of order 0.2 photons per pixel per frame are sufficient to cause ``significant'' ( ~10 %) pileup (based on simple Poisson statistics), and the nominal full-frame ACIS readout time is about 3 sec, one can expect significant photon pileup -- wherein about 1 in 10 events actually consist of two or more photons -- during observations of any point-like source that can generate more than a few hundred counts in a 10 ksec integration.

  
Figure 16: Simulation of the effect of pileup on the shape of the spectrum of an absorbed blackbody. The solid line in each panel is the ``zero pileup'' PHA spectrum, and the dotted line is the PHA spectrum one might actually detect given the listed ACIS frametime. The dashed line shows the ratio of ``zero pileup'' to ``observed'' spectra. Alternatively, this sequence of spectra can be viewed as that expected for a set of sources of identical spectral shape but increasing source flux, as observed with ACIS' nominal frametime of ~3 sec.

To illustrate the effect of pileup on spectroscopy with ACIS, we can examine results from April 1997 X-Ray Calibration Facility (XRCF) tests of the ACIS flight unit in combination with the High Resolution Mirror Assembly (HRMA). Fig. 15 shows ACIS pulse height spectra resulting from illumination by a monochromatic point source. As the source flux (hence ACIS count rate) increases, the 2-photon ``pileup peak'' (located at twice the actual energy [PHA] of the source peak) grows in relative intensity, reflecting the increased odds that 2 photons are detected in the same or nearby pixels during the same CCD exposure and hence are counted as a single event. Furthermore, as source flux increases, 3-photon and even 4-photon events are readily detected. The power in the pileup peaks relative to that in the main spectral line gives a direct measure of the pileup fraction (the fraction of detected events that actually consist of 2 or more photons). Such multiple-photon events lead to count rate non-linearities. XRCF tests involving HRMA+ACIS will be extremely valuable in quantifying pileup in this as well as in other ways. For example, the character of event charge splitting among CCD pixels-and hence the distribution of event grades-is quite sensitive to pileup fraction, and the statistics of charge splitting may be a useful tracer of the presence and degree of photon pileup.

  
Figure 17: Simulated ACIS images illustrating the effect of offset pointing on image quality. Upper left: on axis; upper right: pointing offset of 2 arcmin; lower left: 4 arcmin offset; lower right: 7 arcmin offset. Note that these images are not displayed with the same angular scale, but ``zoom out'' with pointing offset to emphasize the overall change in shape of the PSF. Postscript version of the above image.

For real (polychromatic) astrophysical sources whose intrinsic spectra are unknown, it will be considerably more difficult to detect and measure pileup than it is for the the monochromatic (XRCF) case, since a continuum of incident photons will produce a continuum of piled photons. The ASC is evaluating the likely impact of pileup on ACIS observations with the help of simulations. Fig. 16 shows a comparison of simulated ``true'' and ``piled'' source spectra as produced by MARX (Model of AXAF Response to X-rays). In this example, source flux is held constant while ACIS exposure time is increased (see below); this sequence is equivalent to fixing exposure time and increasing source flux while decreasing total integration time. The sequence shows how the spectrum of a bright source (where ``bright'' here means erg cm sec integrated over the ACIS band, 0.1 to 10 keV) becomes increasingly distorted with increasing pileup fraction. The net effect of pileup, not surprisingly, is to shift detected events toward higher energies; hence the peak flux in the observed (``piled'') spectrum is suppressed relative to that in the ``true'' detected photon spectrum, and the observed spectrum develops a false high-energy tail.

Clearly, then, AXAF observers will want to avoid pileup when performing imaging spectroscopy with ACIS. There are several methods by which one can attempt to mitigate pileup. (For more information concerning these and other mitigation techniques, consult the Proposers' Guide.)

• Shorten exposure time: By cutting back on CCD exposure time per frame, the user lessens chances of multiple photons landing in the same or adjacent pixels (Fig. 16). For flight ACIS, shorter-than-nominal (<3 sec) exposure times can be achieved either by (1) restricting the portion of the CCD that is read out (``subarray readout''), or by (2) using shorter-than-nominal exposures in full-CCD mode. Option (1) allows the user to cut back on the nominal exposure time (and thus pileup fraction) by up to an order of magnitude. Use of subarrays also cuts down on the ACIS field of view, however, and hence will not be attractive for certain clustered object studies. Option (2) retains the full field of view of ACIS, but at a substantial penalty in time lost to overhead since, regardless of integration time, it takes about 3 sec to read the full CCD frame.

    
  Figure 18: Spectra produced by the same simulated event lists used to generate the sequence of off-axis images in the previous Figure. This sequence illustrates how offset pointing mitigates pileup.
  

• Offset point: A bright source of interest can be placed off-axis to deliberately degrade its image, and hence decrease the odds of pileup by spreading source photons over more pixels. Figs. 17 and 18 illustrate image degradation and spectral improvement, respectively, resulting from offset pointing. In most conceivable cases, this method will be preferable to defocus via SIM offset, since offset pointing does not degrade image quality at field center and poses no problems for spacecraft health and safety. Offset pointing may be particularly useful for observations of a bright source in or near a cluster.

The ASC and the ACIS team are continuing to explore options for mitigating, detecting, and possibly even correcting for pileup in ACIS observations. Whether or not these efforts yield new solutions, it is clear that for many ACIS observations, the need to avoid or at least minimize pileup will be a central issue driving many aspects of observation planning.

Joel H. Kastner