CCD Charge Transfer Efficiency Changes
The STIS CCD is a 1024 X 1024 pixel, thinned, backside illuminated SITe CCD, with sensitivity from below 2000 up to about 11000. In a space environment, CCD detectors accumulate radiation damage, which over time leads to increasing dark current, more hot pixels, and reduced charge transfer efficiency. After twelve years on orbit, the effects of such radiation damage on the STIS CCD have become substantial.
The Charge Transfer Efficiency (CTE) measures the fraction of the charge in a given pixel that is transferred to the next row during the readout. The CTE is in practice always less than unity because some of the charge is caught in "Traps" in the pixel and is left behind during transfer of that pixel to the readout amplifier. The fraction of charge lost depends on the signal levels of both the source and the background. Losses are most substantial for faint sources on low background levels. Since in a CCD, charge must be transferred many times before reaching the readout register, even a small decrease in CTE can have a large effect on the measured count rate near the center of the detector. For n transfers, the fraction of detected charge will be CTEn. In Figure 1, we show how the measured Charge Transfer Inefficiency (CTI=1 - CTE) of STIS has changed over the years. For very faint sources near the center of the detector, a substantial fraction of the counts collected by the detector can easily be lost during the readout.
Figure 1 - Measured CTI values over the history of STIS are shown. Each color gives the results for a different source count level. Source counts are in units of electrons per pixel along the dispersion direction, but integrated over a box perpendicular to the dispersion.
Most of the charge lost during a transfer will reappear during a later transfer. This leads to the appearance of "charge tails" below the sources that can by themselves be a considerable source of noise.
The best way to ameliorate this problem is to move the source closer to the readout of the detector. To this end, a number of years ago, new "E1" aperture positions were defined in the long STIS slits to place the target near row 900 of the CCD, much closer to the readout register. This reduces the number of transfers during the readout by about a factor of four. In addition to reducing CTI losses, placing the source spectrum closer to the readout register substantially reduces contamination by the CTE "Tails" of cosmic rays and hot pixels. Also, the dark current is noticeably lower near the top of the detector.
While the CALSTIS software does include an algorithm to correct the extracted flux of a point source spectrum for CTE effects, it cannot restore the lost signal-to-noise. Use of the E1 aperture positions is strongly recommended for all but the very brightest or most extended spectroscopic point sources. Note also that, while the STIS Spectroscopic Exposure Time Calculator does not currently include correction for CTE effects, the effects can be estimated using an iraf script available at http://www.stsci.edu/hst/stis/software/analyzing/scripts/cteloss_descrip.html
Additional information about the STIS CCD CTE and the algorithm for correcting the fluxes of point-source spectra can be found in Goudfrooij et al. (2006 PASP, 118, 1455).
CCD Dark Current
The CCD dark current has also continued to increase over time as radiation damage has accumulated. Since the STIS CCD lacks a working temperature controller or sensor, the dark current also fluctuates with temperature (the CCD housing temperature is used as a surrogate for the unavailable CCD chip temperature). Figure 2 shows the dark current scaled to a housing temperature of 18 C as a function of time at both the center of the CCD and near row 900 at the E1 aperture position.
Figure 2 - The measured median dark current in units of e-/pixel/s is shown as a function of time for a position near the center of the detector (left plot) and near the E1 aperture positions (right plot) at row 900. The red line shows an extrapolation of the late 2004 values to the current epoch.
Time-Dependent Sensitivity Changes
All STIS modes have shown wavelength-dependent changes in sensitivity over time. These changes are believed to result from changes in contaminants on the optical surfaces within STIS. Initial throughput measurements show the post-repair optical throughput are within 2% of values expected from the extrapolation of trends observed in 2004.