We have discovered a significant problem in the gain calibration of recent SIS data, particularly for SIS-1 data taken during AO-8 (calendar 2000). It appears that the Charge Transfer Inefficiency (CTI) is evolving non-linearly with time, whereas we currently use linear extrapolations from the last determination of CTI values on 1997 March 11.

We posted this preliminary report on January 5th, 2001, to warn of this problem. The SIS team has now released a temporary patch, but continues to refine the energy scale calibration.

Symptoms

The profiles of Fe K alpha line of NGC 4151 during the 2000 May observation, obtained using the current set of calibrations. The problem was first noticed in the long observation of NGC 4151 performed in 2000 May. The Fe K-alpha line profiles measured with SIS-0 and SIS-1, using the current calibrations, disagree with each other at a highly significant level.

Re-examination of Cas A calibration observations performed in 1999 summer suggests that the SIS-1 gain was lower than the SIS-0 gain by 1-1.5%. During 2000, the gain difference may have increased to about 3%. It appears that SIS-0 gain has increased relative to current calibration, while that of SIS-1 has decreased.

Such gain miscalibrations can lead to incorrect and mutually inconsistent line profiles in bright sources; they can also affect the determination of continuum shapes in moderately bright sources. The latter is particularly important where the effective area is a steep function of energy, and hence can lead to incorrect values of Nh. [However, other aspects of radiation damage probably also contribute to the low energy calibration problem; see the empirical characterisation of the combined effect.]

Non-linear evolution of CTI

A plot of long-term history of the Ni line energy, measured without CTI correction (but with DFE and echo corrections). The figure above shows the evolution of CTI, as indicated by the background Ni line (from all available 1-CCD Faint mode data without a bright source). Because the Ni line is produced both in the exposure region and the frame-store region of the CCDs, it is not directly usable to calibrate the CTI for the celestial X-ray photons. However, it is a good relative indicator of the CTI.

As seen in the figure, Ni line energies of both SIS-0 and SIS-1 decreased almost linearly up to 1997. Since then, however, the rate of decrease increased in SIS-1, which seems to have reached a saturation point in 1999. In the case of SIS-0, the rate of decrease was almost constant up to 1999, but may have slowed down somewhat since then. These results clearly show that current assumption of linear increase of CTI with time is incorrect.

Some deviation from the linear trend was already seen in the 1999 calibration observations of Cas A; however, the interpretation was made difficult by the slight pointing offset during these observations (a quarter to a third of Cas A was not imaged on the SIS-1 default chip). With hindsight, however, these data do confirm the non-linear evolution of CTI for both SIS-0 and SIS-1.

Temporary patch: a new CTI file

The profiles of Fe K alpha line of NGC 4151 during the 2000 May observation (black: SIS-0; red: SIS-1, both processed with sisph2pi_130201.fits; green: GIS-2). The SIS team has produced an updated CTI calibration file, sisph2pi_290301.fits.

Note this file has been updated on March 29 to correct a separate problem. The earlier version (sisph2pi_130201.fits) has the identical CTI calibration table as this newest version.

This file can be applied to any SIS event files prior to spectral extraction, using the FTOOL sispi. The use of this file results in a marked improvement in the Fe K alpha line profile of NGC 4151 as seen above. We have tested this file with a variety of data, and confirmed that it does not degrade SIS spectra from earlier epochs or at lower energies.

Usage of the new CTI file:

Apply both to SIS-0 and SIS-1 event files, and both to Bright and Bright2 mode files. It can be applied to screened event files (there is no need to re-screen the event files). Simply type sispi, give the data file name, and enter the location of the new calibration file siph2pi_130201.fits in response to Input SIS gain/CTI history file name. Re-extract spectra; the existing region files can be used, and the existing response matrices still apply. WARNING: The SIS team is aware that an energy scale problem still remains using this CTI file, although at a much reduced level (the energy of the narrow core of the Fe line in the NGC 4151 data above are about 20 eV off in SIS data re-calibrated with this new calibration file, with respect to GIS-2). The remaining problem is noticeable only in very high signal-to-noise data. This should be considered a temporary patch.

Future works It now appears likely that the current CTI formula, in which charge loss is assumed to be proportional to the total charge in a pixel, is incorrect. The SIS team is actively investigating an appropriate functional form for the energy dependence, and hope to release a further improved CTI calibration file (and updated software that can utilize it) within the next few months.