INTRODUCTION
All data from the ACIS instrument on the Chandra X-ray Observatory show effects from the non-zero charge transfer inefficiency (CTI) in the ACIS detectors.  This web page presents results from research into data processing techniques that can partially remove these effects.  A review of the CTI problem in ACIS is available.

The work discussed here was conducted by Leisa Townsley and Patrick Broos (with help from several others) to support their involvement in the analysis of ACIS data.  This work is NOT an official product of the ACIS Team and has NOT been endorsed by the ACIS Team or the Chandra X-ray Center (CXC).

The CXC employs a gain de-trending algorithm as part of the standard Level 1 processing.  This largely removes the position-dependence of event energies, which is the biggest effect of CTI; the Energy column of the standard Level 1 or Level 2 event list gives the result of this gain correction.  The performance of this gain de-trending is briefly compared to the performance of the technique presented here in an upcoming paper.  The CXC is developing a CTI corrector based on techniques developed by the ACIS group at MIT as well as on the algorithm presented here.  See this webpage for more information.
 

THE TECHNIQUE
Our CTI model and its application to "correcting" ACIS event lists are described in an ApJ Letter.  It's posted here with original figures and at astro-ph with compressed figures.   A more detailed paper is on the way.  In short, we have used calibration data to tune a phenomenological model of the CTI.  We use an iterative forward-modeling technique to recover the best estimate of the original 3x3-pixel event island that is consistent with the CTI-corrupted event that is observed.  The best estimate is returned as the CTI-corrected event.
 

Three papers have been published describing our work.  You can get to these from ADS if you are careful to check the Physics box in "Databases to query" at the top of the page.

Mitigating Charge Transfer Inefficiency in the Chandra X-Ray Observatory Advanced CCD Imaging Spectrometer
\bibitem[Townsley et al.(2000)]{2000ApJ...534L.139T} Townsley, L.~K.,
Broos, P.~S., Garmire, G.~P., \& Nousek, J.~A.\ 2000, \apjl, 534, L139

THE CODE
Software that implements the CTI model and correction algorithm described in the paper is available.  This code is provided in hopes that it may assist other investigators interested in the ACIS CTI problem.  ACIS observers with access to IDL may also download the code and correct their data.  The CTI corrector works on a standard Level 1 event list and returns a modified Level 1 event list.  Users are then expected to complete the other filtering steps (e.g. grade, GTIs, flaring pixels) that their data require.  Only data employing the standard full-frame readout (no sub-arrays or CC-mode) can be corrected.  This is because the trap time constants interact with the readout time, thus these other modes will exhibit different CTI effects.
 

DETAILS

* The corrector has been calibrated for focal plane temperatures of -110C (Aug 1999 - 29 Jan 2000) and -120C (29 Jan 2000 - present), using all appropriate External Calibration Source (ECS) data for -110C and all ECS data available at the time of calibration for -120C (about a year's worth).  This amounts to well over a million events per amplifier.

* Each CCD exhibits slightly different CTI effects, so parameter tuning must be performed on each device (and each amplifier of that device) separately.  This is a labor-intensive problem.  With the help of many people on the PSU/ACIS team, the CTI corrector is now available for all CCDs in the ACIS imaging array (I0, I1, I2, I3, S2, S3) at both -110C and -120C.  It is valid for S2 and S3 with or without the gratings.

* The CTI corrector changes the event amplitude (energy) and charge distribution (grade) to account for the charge loss and charge smearing effects of CTI. Thus CTI-corrected data need matching response matrices and QEU files for accurate spectral analysis.  Using our Monte Carlo CCD simulator in conjunction with the CTI model, we have generated these (PI) RMFs and QEUs for I0, I1, I2, I3, S2, and  S3 at both -110C and -120C.  They are available to ACIS observers as a gzipped tarfile called CTI_products.tar.gz.  Be sure to check out the README file.

* S3 RMFs:  Since the spectral resolution of this BI device does not exhibit a strong spatial dependence, a single RMF is adequate for sources on this device.  The CTI correction is calibrated so that -110C data use a CCD gain that reproduces (as closely as possible) the energies of -120C CTI-corrected data.  This was done so that -110C and -120C data can be combined for spectral analysis.  Due to small temperature dependencies in the CCD spectral redistribution function, users should still apply the -110C RMF to -110C data unless those data will be combined with -120C data for spectral analysis, in which case the -120C RMF should be used.  These S3 matrices assume standard ASCA Grade 0,2,3,4,6 grade filtering (applied AFTER CTI correction).

* FI RMFs:  The CTI corrector improves the row-dependent degradation in spectral resolution caused by CTI but it cannot remove these effects completely.   Thus position-dependent RMFs are still necessary.  Since CTI effects are reduced after correction, though, only a small number of RMFs suffice.  We provide 8 PI RMFs for each device/temperature combination; each is a 128-row (all column) average.  We additionally provide a full-CCD RMF and a 64-row average RMF around the I3 aimpoint.  These matrices should be adequate for most applications, including extended emission.  All assume standard ASCA Grade 0,2,3,4,6 grade filtering (applied AFTER CTI correction).

* Our method for RMF generation yields a large supply of simulated events that can be re-filtered to generate custom RMFs.  The most obvious application is RMFs that use non-standard grades.  Different spatial filtering is also possible.  If you have a compelling reason to need such custom RMFs, please contact Leisa Townsley  for assistance.

* The corrector has been calibrated primarily at the energies present in ACIS ECS data (0.680, 1.486, 2.711, 4.155, 4.511, 4.932, 5.895, 6.490, 9.711 keV).  It has been tested on high-energy instrumental lines (Ni K and Au L) present in celestial data and appears to behave nominally.  It has been minimally tested at low energies (lines in the SNR E0102-72.3) and again appears to behave nominally.  The gain behavior in the 0.2-0.5 keV range for the BI device S3 is untested due to a lack of calibration information.

* This work is not part of our primary responsibilities, so support for installing and using this tool will be limited.  If you believe that the CTI corrector will significantly improve your science results but you lack access to IDL, please contact Leisa Townsley for assistance.
 

We welcome feedback from other ACIS and CTI investigators who choose to try out the corrector code.