This screening also eliminates almost all of the afterglow events.
A quick look through the 5×5 event islands recorded in the Very Faint mode reveals that a large number of events are in fact the end-points of big particles tracks. A standard grade analysis which uses 3×3 islands, fails to screen these events. Several examples are shown below.
Such events can be screened by rejecting those in which any border pixel in the 5×5 island is above the split threshold. As I show below, such a screening does not remove any significant fraction of good photons from a faint (i.e., with no pile-up) source.
The implementation of the filtering is straightforward. If any border pixel in the 5×5 event island exceeds a split threshold, the event is flagged in the status, fltgrade or grade column. A subsequent filtering with dmcopy or fcopy can remove the bad events.
The only issue here is to properly deal with the CTI effect for high energy photons. Because of this effect, there is a charge trailing the bright central pixels along the direction of the chip readout. For high energy photons, the trailing charge can exceed the split threshold at the border pixels phas(2:4,5). The obvious solution is to subtract the expected average trailing charge corresponding to observed intensities in pixels phas(2:4,3). A quick solution I am using in the current implementation is to use a conservative value of phas(*,5)/phas(*,3)=2.7% corresponding to large CHIPY's in the ACIS-I chips at -110C at high energies.
TO DO: This CTI correction, of course, depends on the chip, the focal plane temperature, and CHIPY of the photon, which should be taken into accoint in the code at some point. These should have a minor effect on the efficiency of the rejection and loss of real photons.
In the recent CIAO versions, this cleaning algorithm is implemented in acis_process_events, see http://cxc.harvard.edu/ciao/threads/aciscleanvf. Briefly, the user has to use the option check_vf_pha=yes to flag bad events and then filter them by, e.g., fcopy evt.fits"[events][status==bxxxxxxxx0xxxxxxxxxxxxxxxxxxxxxxx]" good_evt.fits.
Also, a stand-alone fortran program is available at
It should be compiled with the CFITSIO library.
The program flags bad events by setting a bit in the STATUS column, or by changing FLTGRADE or GRADE to 255. Bad events can then be filtered out by dmcopy or fcopy. Usage:
clean55 [-bN | -fg | -g] [-ct coeff] [-spth sp_thresh] f1.fits f2.fits ... -bN : change N-th bit in the STATUS column. -b20 is the default -fg : instead of STATUS, change FLTGRADE to 255 -g : instead of STATUS, change GRADE to 255 -ct coeff : set the average ratio phas(*,5)/phas(*,3); default is 0.027 -spth sp_thresh : set split threshold (integer value); default is 14
Fcopy filtering example (assuming that bit 20 equals 1 for bad events):
fcopy evt.fits"[events][status==bxxxxxxxxxxx0xxxxxxxxxxxxxxxxxxxx]" good.fits
The particle background reduction is best demonstrated on the ACIS background observation performed in the stowed position, not exposed to the sky (see http://cxc.harvard.edu/contrib/maxim/stowed).
To demonstrate that the loss of good X-ray photons is negligible, we present tests for fields with celestial sources: Perseus cluster, a deep distant cluster observation (Lynx field), a cooling flow cluster MS2137-2353. Lynx field was observed in May 2000 when ACIS temperature was -120C; all other observations were made in the fall of 1999 at -110C. The high background periods were excluded.
The figure below compares the particle background spectra from chips 0236 before and after the VF cleaning:
The top panel shows the background spectra before cleaning (black) and after cleaning (red). The ratio of these spectra is shown in the bottom panel. The background reduction is a factor of 2.5 near 0.5 keV, 1.4 near 1 keV, 1.1 between 2-5 keV, and reaches 1.4-1.5 above 7 keV.
Total image in the 0.5-2 keV band:
Excluded events in this band:
As suggested by these images, almost no good photons are lost as a result of the VF cleaning. This is shown in more detail in the section below.
The spectrum of the central part of the cluster, excluding the piled-up nucleus is shown below. Background was subtracted using a region which was relatively free from the cluster emission.
Red crosses show the spectrum after cleaning. The bottom panel shows the ratio of the clean and total spectra. Within the error bars, the data are consistent with a 2% loss of good photons, independent of energy. This is a geometric factor representing the probablity of the particle background event falling inside the 5×5 pixel island of the real X-ray photon.
Since the fraction of real photons removed by the VF cleaning is low and does not show any energy dependence, one can continue using the CCD QE and RMF calibration applicable for standard grade selection.
Again, the background reduction is best demonstrated on the ACIS-stowed observation. The figure below compares the particle background spectra from chip S3 before and after the VF cleaning:
The top panel shows the background spectra before cleaning (black) and after cleaning (red). The ratio of these spectra is shown in the bottom panel. The background reduction is a factor of ~ 5 near 0.3 keV, 1.05-1.1 between 0.5 and 4 keV, and reaches 1.25 above 6 keV.
This test case shows the effect of the VF cleaning on the real X-ray photons in the back-illuminated S3 chip. The spectrum of the cluster center (with the background subtracted) is shown below.
This figure shows the background subtracted cluster spectrum before and after the VF cleaning (red and black, respectively), as well as the background spectrum before and after cleaning (blue and green). Again, the change of the source spectrum is negligible (lower panel).
Interestingly, the filtering described above also excludes the ``afterglow'' events. For example, the first 6 events detected by acis_detect_afterglow in one of the above observations are shown below:
All these events are obviously excluded due to the high signal in the border pixels
For bright sources, the VF filtering should not be applied because it can remove good photons due to overlap of the events' 5×5 pixel islands. Analysis should proceed as usual for the FAINT mode.
Last modified: 11/15/10
The Chandra X-Ray
Center (CXC) is operated for NASA by the Smithsonian Astrophysical Observatory.
60 Garden Street, Cambridge, MA 02138 USA. Email: firstname.lastname@example.org
Smithsonian Institution, Copyright © 1998-2004. All rights reserved.