Update 2005 Feb 28: Improved understanding of the CTI/background correlation for FI CCDs has confirmed that after removing the effects of the background variation, the corrected CTI data is completely consistent with no step-function change in Oct 2003. The 3-sigma upper limit on I-array CTI change is 9 x 10-7. Similar analysis of BI data reduces but does not eliminate the apparent CTI change. The measured CTI increase after removing both the long term linear trend and background variations is (5.5 ± 3.5) x 10-7. The plots and text below have not been updated to reflect this.
This is a quick analysis to determine whether ACIS CTI suffered a step function increase as a result of the series of large coronal mass ejections in October-November 2003. The largest, on Nov. 11, was associated with the strongest X-ray flare ever measured.
The data shown here is available on my CTI monitoring web page. For reference, a CTI change of 1 x 10-6 corresponds to a maximum pulseheight change (for events furthest from the readout) of about 0.1% at 5.9 keV and about twice that at 1.5 keV.
The top panel shows the measured I-array parallel CTI since January 2000. The vertical dotted line marks the approximate time of the series of large CMEs. The bottom panel zooms in on the past year.
There is clearly a step-function change in measured CTI at the time of the CMEs. The size of the jump is (4.0 ± 0.2) x 10-6 (from measuring the mean CTI in the 55 days before and after the CME). Since that time, the measured CTI has gradually decreased to nearly the previous level.
This change in measured CTI will manifest itself in a change in pulseheight. Across the CTI jump, the maximum pulseheight change is ~0.4% at 5.9 keV and ~0.8% at 1.5 keV at row numbers far from the readout. At lower row numbers the change will be smaller. This is being incorporated into the calibration products that are used by the tgain tools and acis_process_events (CIAO3.1)
The measured CTI depends on a number of factors. An increase in CTI could be due to:
The detector temperature was nominal throughout this period and should not be the cause. Changes in the trap time constants would manifest as changes in the fraction of lost charge re-emitted in the trailing pixel - this is not seen in the data.
That leaves sacrificial charge/cosmic ray background and additional radiation damage as the possible culprits.
The above plots show the high energy reject rate from S3. This rate is correlated with > 1 MeV cosmic rays. The change in background rate seems to mirror the change in CTI, which indicates that at least some of the apparent change in CTI is due to sacrificial charge and not radiation damage. The large jump and long duration recovery make this event unusual, at least in the four years of Chandra's lifetime.
The above plots show the CTI after applying a standard correction for the background variations. The jump in CTI is smaller at (2.7 ± 0.6) x 10-6. At the present time, the corrected CTI has decreased to close to the previous level. The horizontal dashed line is a linear fit to the two years of data preceeded the CTI jump. The most recent data points are completely consistent with that gradually increasing trend. After accounting for the gradual increase, the 3-sigma upper limit on CTI change is 6.0 x 10-7.
Was there any increase in CTI due to the October CMEs? There are two plausible mechanisms either of which (or both) might be at work here. The jump in CTI could have been due to radiation damage from the CME that was annealed over the following months, or the jump and decrease are due to cosmic ray sacrificial charge that is not being completely accounted for by the current algorithm. Since the lightcurve of background events is unusual during this period, it is possible that the makeup of the background was also unusual, so that the standard background correction was insufficient.
The above plots show the parallel CTI on S3. Previous studies have shown that the type of CTI in the BI devices is not affected by cosmic ray background sacrificial charge. There is an apparent CTI jump of ~1 x 10-6 at the time of the CMEs.
Averaging the CTI for the six months preceeding and following the CMEs, the change in CTI is (8.6 ± 0.7) x 10-7. A small increase is expected from the gradual trend seen in historical data. After removing the gradual increase, modeled by a linear fit to the two years of data preceeding the CMEs, the increase in CTI is (7.0 ± 0.9) x 10-7.
There is no evidence for a change in serial CTI on either type of CCD. The 3-sigma upper limit on I-array serial CTI is 7.2 x 10-6 and on S3 serial CTI change is 5.4 x 10-6.
There is no evidence for a change in CTI in the framestore region of either type of CCD. The pulseheight of events at the bottom of the imaging region has not changed due to the CMEs.
There is no evidence for a change in serial CTI or parallel CTI in the framestore. By itself this indicates that any damage must have been done by soft particles, however ACIS was stowed at the HRC-S position for the worst of the storm. The CME was one of the hardest Chandra has seen, so the damage may have been done by softer secondaries produced by the hard CME particles. It is difficult to reconcile the apparent CTI change on the BI CCDs with the lack of change on the FI CCDs. XMM has reported no change in CTI for MOS or PN, but has a different orbit, so could have experienced a different level of geomagnetic shielding.