STIS Update

Van Dixon, dixon@stsci.edu

Effects of Imperfect CTE on Scientific Observations with STIS

After more than 13 years on orbit, the charge-coupled device (CCD) detector in the Space Telescope Imaging Spectrograph (STIS) shows the effects of accumulated radiation damage. In particular, its charge-transfer efficiency (CTE)—the fraction of charge successfully moved between adjacent pixels during readout—continues to decline. When this number is less than unity, some fraction of its counts trail behind a bright pixel in one or both dimensions of the image. CTE trails have three effects. First, they remove counts from the data, reducing the observed flux in both spectroscopic and imaging observations. Second, they distort the images of stars and other targets, introducing a systematic bias into astrometric measurements. Third, the trails from cosmic rays and hot pixels that lie between the target and the read-out amplifier (down-stream trails) add noise to the target’s spectrum or image. Tools to correct the photometry and astrometry of point sources are available, but they are not applicable to extended sources and do not reduce the noise imparted by CTE trails. In this article, we explore the effects of CTE degradation on imaging and spectroscopic observations with STIS. Additional details are provided in STIS ISR 2011-02.

Figure 1In recent observing cycles, most STIS CCD imaging has used the coronagraphic mask to study circumstellar disks and planets. When a bright star is placed behind the mask, the stellar point-spread function (PSF) extends across the field, filling the charge traps like a pre-flash. Near the star, CTE effects are negligible; farther away, they begin to reappear. Figure 1 presents a coronagraphic image of the field around the bright (V = 1.16) star Fomalhaut. The image is constructed from seven separate exposures, combined with a median filter to reject cosmic rays (CR-SPLIT = 7). The star lies to the right, about 33″ from the image center. Hot pixels within about 20″ of the star do not show CTE trails, but those farther away do. (For reference, the median pixel value approximately 20″ from the star is ~200 electrons in each CR-SPLIT image.) The extent of the CTE-free zone varies with the stellar brightness and exposure length.

To explore the effects of CTE trails on faint spectra, consider the pair of faint (V = 18) supernova spectra (program 11721, PI: R. Ellis) plotted in Figure 2. Both were observed in late 2009 with grating G430L for a total of 2300 seconds using CR-SPLIT = 3, but one was placed at the center of the CCD (Y ~ 512; red curve), while the other was placed at the E1 position (Y ~ 900; black curve), closer to the read-out amplifier (Y = 1024). The spikes in both spectra are down-stream CTE trails. Though the two exposures were obtained under similar conditions, the red spectrum shows many more trails than does the black.

Figure 2

Some simple statistics can help to quantify the noise imparted by CTE trails. We normalize both spectra (using a smoothed version of the E1 spectrum as a template) and consider the region between 4300 and 5600 Å.  The mean and standard deviation of the central (red) spectrum are µ = 1.16 and σ = 0.46, corresponding to a mean signal-to-noise ratio (S/N) per two-pixel resolution element of 3.07 (µ/σ × √2). The values for the E1 (black) spectrum are µ = 1.00, σ = 0.09, and S/N = 16.6. The use of the E1 position significantly reduces contamination by CTE trails, but it is not a panacea. When given our template spectrum as input, the STIS exposure-time calculator (ETC) estimates that a 2300-second exposure with CR-SPLIT = 3 would yield a spectrum with S/N between 25 and 45 per two-pixel resolution element in this wavelength range. The mean S/N of the E1 spectrum is less than half this value. (Vignetting at the E1 position is responsible for roughly 10% of this discrepancy.) Faint CTE trails thus represent an important souce of noise that is not included in the ETC.

A pixel-based CTE correction, similar to that under development for ACS, will eventually be available for STIS.  Unlike current CTE correction algorithms, it will be applicable to both point and extended sources, and it will reduce the noise imparted by down-stream CTE trails to both spectroscopic and imaging data.  Until then, juducious use of scattered light in coronagraphic images and the E1 position for spectroscopic observations can reduce the impact of CTE effects on STIS CCD data.