Calibration Access and Data Handbook

Procedure

The current MOS procedure envisages a less complicated correction mechanism than for the PN. However as the physical principles are the same it is plausible that MOS routine may need revising at some future date to more nearly approximate the PN.

The CTI correction applies a scalar correction to the event list set, on an event by event basis, in order to correct for loss of charge which has occurred in reading out the CCD array.

Note that the $EMENERGY$ task just requires the PHA values be corrected. There is no flag or other data carried forward to the RMF-generating routine for example, to account for a ``noise'' factor in the correction process.

We expect to correct for a fixed amount of charge loss, and a fractional amount of charge loss, which depends on the number of row and column transfers completed. The early analysis of ground calibration data shows some possible detailed effects that need to be accounted for. For example:

• optical background - with the OPEN filter position, a significantly more efficient charge transfer resulted. This must be due to the filling of traps by the continually circulating small level charges generated by optical photons.
• a count - rate effect where the CTI may depend on the filling of traps by precursor events. Although this is explicitly handled for the PN camera we envisage in the MOS only to use some local count rate criteria to modify the CTI values. This is not yet implemented.
• There seem to be (rare) cases where a localised trap creates a step-wise increase in charge loss. If the CCF components accurately give a fixed loss component per column, this can mimic loss of extra charge from some point in the serial register (although then a node dependency is required), and also this accounts for possible spatially varying charge loss factors in peripheral columns for example. At present the data do not seem to warrant a description of additional step-wise loss occurring at an individual row.

Note that the intended use of the correction occurs in the pipeline at $EMENERGY$, so that the rawX,rawY attributes are related to the co-ordinates range (1,600), whereas if the same routine were to be used to correct diagnostic data, then a conversion is needed. This would be invoked on looking for the PUT_XY flag setting. At the moment the latter case is hardly conceived as needing some CTI correction therefore ignored and assume a future SEPARATE CAL routine might be needed. Use in $EMENERGY$ also implies we can only accurately correct $BEFORE$ a modification for a background light contamination has occurred, and also that in principle we should correct ALL E$_{i}$ components of an event, and not just the central pixel value.

Additionally we note that in principle CTI may change with mode due to the differences in clock timings employed, so that a mode dependence in the CAL state and the CCF files must be carried (TBC from ground data).

Note in the current context of applying the correction to a whole array there is not a concept of handling a frame-dependent count rate dependence to modify the CTE values. The best method for doing this should be established. For example should $EMENERGY$ actually ingest a frame file, and calculate from that some count rate estimate - not easy if the dependence is really column dependent.

Algorithm:
A detailed in-flight trend-analysis of the MOS Charge Transfer Inefficiency (CTI), using the two main emission lines (Mn and Al) from the internal calibration sources, spanning three years of operations has shown that :

(1) at a given energy the energy loss per transfer due to CTI (Charge Transfer Inefficiency) is a linear function of time, in time intervals between major discontinuities due to solar flares or change of CCD operating temperature.

(2) the energy losses per transfer scale as a power law function of energy, by comparing the time evolution (degradation) of the CTI at Mn (6keV) and Al (1.5 keV) energies.

Assuming that the energy losses at other energies can be extrapolated from these two energies (Al and Mn emission lines from the internal calibration source), the MOS CTI has been successfully modeled and coded as follows in the CAL CtiCorrector.cc with ALGOID = 2 (in the latest set of CCFs) for a pixel at CCD coordinates (RawX,RawY) :

$eOut = eIn + rawY*ctiY + rawX*ctiX - offset(rawX,rawY)$ with :

$ctiX = ( serial[0] + serial[1] * deltaT )* eIn^{(serial[2])}$

$ctiY = ( parall[0] + parall[1] * deltaT )* eIn^{(parall[2])}$

where :

• eIn is the uncorrected PHA value.
• eOut is the CTI-corrected PHA value.
• offset(rawX,rawY) are energy offsets of individual column segments.
• deltaT is the time difference between launch and the observation date, expressed in seconds.
• all other serial[i] and parall[i] coefficients are CCD dependent and stored in the CCF MOS CTI extensions CTI_EXTENDED. from version 8 onwards, using ALGOID = 2. (Previous CCF versions use ALGOID = 0 and 1)

The transfer losses are dominated by the parallel CTI. Its typical power law dependence with energy (PHA) of the order of 0.65 for all 14 CCDs (parallel[2] coefficients) at the CCD operating temperature of -100C, but seems to have dropped to lower values (0.5 or lower), since the detectors were cooled down to -120C in revolution 533.

The elapsed time since launch is computed from the date indicated in the FITS header keyword REF_DATE of the first binary extension.

The plots of MOS parallel and serial CTI trends since launch for all CCDs at Mn and Al energies are available on the internal web at:
http://xmm.vilspa.esa.es/ xmmdoc/MOS/mos_cti.html