September 2020 STAN

Performance of Spatial Scans with the STIS CCD

Spatial scanning with the STIS CCD is an available-but-unsupported mode for obtaining high S/N ratio spectra of relatively bright targets by trailing the target in the spatial direction within one of the long STIS apertures.  Possible scientific applications include the reliable detection of weak stellar and interstellar absorption features (particularly in the red and near-IR where ground-based observations can be severely compromised by strong telluric absorption) and the accurate monitoring of stellar fluxes (both broad-band and in narrower spectral intervals) for characterizing transiting exoplanets.  As discussed in the 2017 March STAN and in section 12.12 of the STIS Instrument Handbook, spatial scanning offers the potential advantages of improved flat fielding, improved fringe removal (beyond about 7000 Å ), and better correction for cosmic rays, compared to the previously used method of deliberately saturating the CCD detector at a fixed pointing.  The 2018 July STAN  presented some initial results regarding the scan geometry and the flux stability that might be obtained, based on data obtained for special calibration program 15383 (PI C. Proffitt), which was designed to test some aspects of this new observing mode.  In this article, we summarize the results of more detailed and extensive analyses of a set of 20 scanned spectra of the bright (V = 5.95) G8 V star 55 Cnc that were obtained for that program, which yield more precise indications of the flux stability that can be achieved.

The second visit of program 15383 obtained a number of scanned spectra of 55 Cnc with the STIS G750L grating and CCD detector, covering the range from about 5240 through 10270 Å.  55 Cnc is a known exoplanet host with numerous prior HST observations, and the visit was scheduled to execute when 55 Cnc e was not in transit -- so that no intrinsic variations in flux would be expected.  During each of the second and third orbits in that visit, a series of ten short (12 arcsec) scans was taken through the 52x2 aperture, in order to assess the repeatability of the fluxes — both within each orbit and from orbit to orbit.  Average counts of order 1-2 x 10⁴ per pixel were obtained for most of the individual columns in each scanned spectrum of the target.  In addition to various stellar absorption lines, the raw spectral images exhibit both the expected interference fringes at the longer wavelengths and a clear horizontal “striping” pattern – which differs from image to image and which is presumably due to some combination of spacecraft jitter and (perhaps) very slight variations in scan speed (Fig. 1).  Deep contemporaneous flat field exposures (for fringe correction) were obtained during the occultations following each orbital visibility period.  While the initial inspections of the spectral images indicated that both the positioning of the scans on the detector and the total integrated (“white light”) fluxes for the scans appeared to be quite consistent (2018 July STAN), those initial assessments did not consider several potentially significant instrumental effects and did not employ the linear regression fitting (“de-trending”) analyses typically used in efforts to remove such instrumental effects from time series observations of exoplanet hosts (e.g., Sing et al. 2011, MNRAS, 416, 1443).

Following those initial investigations, several members of the STIS team and several local (JHU, STScI) exoplanet researchers collaborated on a more comprehensive analysis of the set of scanned spectra of 55 Cnc.  First, cosmic rays and residual bad/hot pixels were identified and corrected in de-striped versions of the images (as the variable striping in the individual images made it difficult to use the usual cosmic ray correction procedures (e.g., OCRREJECT)).  A new de-fringing tool, developed by several members of the STIS team and implemented in python, was then applied to the cosmic-ray-corrected images (with the striping restored, to avoid any possible subtle changes to the total fluxes; Fig. 1).  Finally, a version of the de-trending procedure developed by Sing et al. (2011) was applied to the de-fringed images and spectra.  The de-trending included the customary dependences on the HST orbital phase (up to a 4^th-order polynomial) and on the slight differences in the x,y position of the spectral image on the CCD and in the wavelength zero point determined from the individual images.  The de-trending was applied both to the total integrated fluxes and to the fluxes in more restricted wavelength ranges.  Comparisons were also made with similar analyses of archival pointed/saturated STIS time series spectra of 55 Cnc obtained under program 13665 (PI B. Benneke).

Figure 1 – One of the 20 short scanned G750L spectral images of 55 Cnc.  Wavelength increases from left to right; the scan is performed roughly perpendicular to the dispersion direction.  At the top is the cosmic-ray-corrected image, with the fringing pattern apparent at longer wavelengths.  At the bottom is the de-fringed image.  In both cases, the vertical lines are stellar absorption features, while the horizontal “striping” (which differs from image to image) is likely due to a combination of spacecraft jitter and (perhaps) slight variations in the scan rate.

These more comprehensive analyses of the spatially scanned spectra both confirm the general trends seen in the initial analyses of those spectra and indicate that:

1. The systematic differences between orbits – several hundred ppm for these data – are smaller than those typically seen for pointed/saturated exposures (which can be as high as several thousand ppm) (Fig. 2).
2. Within each orbit, the differences among the scans are also smaller than those often seen for pointed/saturated exposures.
3. The trends in the relative fluxes differ somewhat with wavelength.
4. For the total (“white-light”) flux, the scatter about the de-trending fits is of order 30 ppm (Fig. 3) – comparable to the best values achieved in any previous studies of exoplanet host stars with HST.  Somewhat larger values are found for narrower wavelength regions.
5. While averaging over the larger spatial range in the scanned spectral images does reduce the amplitude of the fringing in 1D spectra extracted from those images, the additional step of de-fringing the images both effectively removes the fringes and appears to reduce the scatter about the de-trending fits by of order 15-20% at the longer wavelengths.

Figure 2 – Relative total (“white light”) fluxes of 55 Cnc – derived from pointed/saturated STIS spectra (black; program 13665) and from spatially scanned spectra (red; program 15383).  The spatially scanned spectra exhibit smaller systematic differences, both from orbit to orbit and within each orbit. (Note, however, that the second orbit for program 13665 included a transit of 55 Cnc e, while both orbits for program 15383 were out of transit.)

Figure 3 – Relative total (“white light”) fluxes of 55 Cnc – derived from spatially scanned spectra (black) and from the linear regression (de-trending) fit to those observed data (red).  The rms deviation between the data and the fit (27 ppm – even before correcting for cosmic rays and fringing) is comparable to the best values obtained from any previous studies of exoplanet host stars with HST.

While these results are based on somewhat limited data (20 scanned spectra obtained during just two orbits in a single HST visit), they do indicate that spatial scanning with the STIS CCD might be advantageous for some applications, compared to other available HST observing modes. For example, while WFC3, with the UVIS and IR grisms, can provide high S/N spectra of fainter targets with broader wavelength coverage, STIS spatial scans may be preferable for obtaining higher resolution spectra of brighter targets (e.g., for V < 7.5) – particularly at wavelengths greater than about 6000 Å, where the UVIS spectra will exhibit saturation and contamination from overlapping orders and the IR spectra would require rather rapid scan rates.

More detailed information regarding observing strategy — e.g., determining the scan parameters for observing a particular target (exposure time, scan rate, initial target positioning) and obtaining custom flat-field exposures — may be found in Sec. 12.12 of the STIS Instrument Handbook; specific questions may be addressed to the STScI HelpDesk.  The spectra of 55 Cnc obtained for calibration program 15383 are publicly available via the MAST archive. We would encourage those who may be interested in using STIS spatial scans to download and examine those data, in order to gauge the potential utility of this technique for their own research projects.