12.12 Spatial Scans with the STIS CCD

Spatial scans, which are a supported mode for WFC3 observations (see WFC3 ISR 2012-08: Considerations for using Spatial Scans with WFC3), can also be used with the STIS CCD as an available-but-unsupported mode. For a bright target this can potentially allow signal-to-noise levels of at least several hundred or perhaps as high as a thousand to one to be obtained, even at wavelengths where the STIS CCD is affected by strong IR fringing. This can be especially valuable at wavelengths that are also strongly affected by telluric features in ground-based spectra. Many such regions can be observed with the STIS G750M at wavelengths up to 10140 Å (1.014 μm), with a resolving power close to 10000, allowing weak features such as diffuse interstellar bands to be observed with an accuracy not achievable by any other means.

Spatial scans are currently not available with STIS MAMA observations.

12.12.1 Why Use Spatial Scans with STIS?

Trailing an external point source in the spatial direction during a first order STIS CCD spectral observation can spread the target's light over a much larger area of the detector than would be the case for a simple pointed observation. This has several potential advantages:

• Many more photons can be collected before reaching the detector full well saturation limit.
• Spreading the light over a larger fraction of the detector area will better average over flat field variations.
• Trailing along a long slit can result in IR fringing patterns that are significantly closer to the fringing patterns produced by the available contemporaneous tungsten lamp flats, leading to much improved near-IR fringe removal.
• A wider variety of algorithms can potentially be used to detect and remove hot pixels and cosmic rays.

However, spreading the source light over a larger area of the detector can also significantly increase the dark current and read-noise in the extraction region, so the advantages of spatially scanned spectra will be most significant for very bright targets.

12.12.2 Possible STIS Spatial Scan Use Cases

Trailing along Narrow Slit to Maximize S/N and Fringe Removal

The initial on-orbit use of this technique was performed by HST program 14705 and reported in Cordiner et al. 2017 (ApJ, 843L, 2). Signal-to-noise in excess of 600:1 per resolution element was achieved for G750M spectral observations near 9300 Å. In this case, the narrow 52 × 0.1 aperture was used for both the trailed external observations and the internal lamp flats. As the STIS PSF near this wavelength is relatively wide compared to the slit, the fringing pattern from the lamp and the trailed external target are very similar, and this greatly facilitates the fringing correction.

In Figure 12.8, we show an example of one of these trailed spectra observations before and after a simple division by the contemporaneous fringe flats. The very effective removal of the fringing pattern, allows detection of much weaker spectral features (vertical dark bands in this image), than would be possible for a simple pointed image.

Trailing along Wide Slit to Maximize Photometric Repeatability

While use of a narrow slit does ensure the best fringe removal and is recommended for programs measuring the absolute equivalent width of weak features, it does compromise the absolute flux calibration of the observation.

STIS/CAL program 15383 recently obtained scanned G750L observations of the G8 V star 55 Cnc, and included both scans in the narrow 52 × 0.1 and the wider 52 × 2 apertures. This visit included a series of repeated 12" long scans along the wider aperture over the course of two adjacent orbits that should allow good limits to be set on the absolute repeatability of flux measurements with scanned G750L observations. This will answer the question of how the repeatability of scanned observations compare to the more traditional technique of taking saturated spectra at a fixed pointing (see STIS ISR 1999-05), which has commonly been used for transiting planet observations with STIS. Detailed analysis of the 15383 data is described in the September 2020 STAN and a paper is in progress.

Variations in the trail rate are also visible during the wide aperture scans, and this will probably prevent any use of trailed STIS observations to model source flux variations during an individual scan; it remains to be determined what limits this may place on the accuracy and repeatability of the integrated flux for individual scanned exposures.

12.12.3 Using STIS Spatial Scans

As STIS spatial scans are an "available-but-unsupported" mode, only a limited amount of tuning has been done to optimize this mode. In brief, to scan along one of the long apertures, one must specify the starting location (in arcsec, relative to the aperture fiducial point; using the POS TARG special requirement), the scan rate (in arcsec s^–1; see below), the scan orientation (nominally 90 degrees), the scan direction, and the total exposure time. The scan length is then given by the product of the scan rate and the exposure time; the Y coordinate of the starting location should be specified to position the scan as desired on the detector. Currently, "forward" scans place the spectra at close to the intended position, but there is apparently a timing error in the implementation of "reverse" scans that result in them being offset on the detector to higher than intended Y locations, with larger offsets at higher scan rates. As a result, we currently recommend that STIS observers use only "forward" scans and avoid both the "reverse" and "round-trip" options. Because different apertures can map to slightly different positions on the detector, when scanning up the 52 × 2 and doing fringe flats with the 52 × 0.1, the procedure should be to first peak up in the 52X0.1, and then add an X component to the POS TARG for the 52 × 2 scan to correct for the –0.049241721" difference in SIAF X locations between the two apertures.

Visit 01 of program 15383 took scanned imaging observations to check the alignment of the default scan angle with the science apertures. This visit suggested a small correction to the default scan angle of ~+0.065 degrees, but the repeatability of this correction has yet to be verified.

For more information about proposing for spatial scanning using STIS, please see Section 7.3. in the Phase II Proposal Instructions.

Using the ETC to Estimate Scan Parameters

Currently the STIS ETC does not support calculations for scanned observations. However, to achieve the expected flat-fielding improvements, it is essential to avoid saturating the detector.

To estimate the expected peak flux levels, we recommend beginning with a spectroscopic ETC calculation of a pointed observation with CR-SPLIT=1 for the requested total exposure time of an individual scan with all other parameters set as they will be for the actual observation. Then go to the "Table of Source and Noise Counts per Pixel" on the ETC results page. In this table, "pixel" refers to a pixel in the extracted 1D spectrum which is actually a sum in the original 2D image over 1 pixel in the dispersion direction and the extraction height (7 pixels by default for the STIS CCD), in the spatial direction, while the "counts" given in this table are always in units of electrons, regardless of the gain setting. The source counts/per pixel in this table have been scaled by the wavelength dependent fraction, E, of the total source flux at that wavelength that is expected to fall within the extraction height, and this excluded light in the wings of the PSF will contribute to the counts accumulated during the scan. For the G750L grating the encircled energy fraction in the 7 pixel region varies from ~0.8 at 6000 Å, to about 0.6 at 10000 Å. If, S, is the "Source Per Pixel" in the ETC table, and L is the length of the scan in pixels (the scan length in arcseconds divided by 0.0508"/pixel), then the approximate local source count rate in electrons per second per pixel in the scanned 2D image will simply be S/E/L; this is the number that should be compared to the saturation limits.

Because the spectrum will be spread over a larger region of the detector than was assumed in the ETC, when obtaining a good S/N estimate for fainter targets it will also be necessary to scale the dark current, sky background, and the read-noise squared from the values calculated for the default extraction height to the actual area to be used in the final extraction. If these are significant compared to the source count rate, then the final signal-to-noise should be manually computed using the equations in Section 6.4. If the read-noise and dark current are negligible compared to the Poisson noise of the source counts, then the S/N estimate in the original ETC calculation is probably as good an approximation as any available.

We can take as a concrete example one of the series of 218.1 s exposures done in the 2^nd and 3^rd orbits of visit 11 of 15383. The scan rate there was set to 0.055 arcsec/sec, which yielded a scan length of 12 arcseconds or about 236 pixels. An ETC calculation for a pointed exposure with the same parameters is given by http://etc.stsci.edu/etc/results/STIS.sp.1044073/. At 5828.9 Å, the tabular data page for this ETC calculation gives about 1.48e7 electrons as the "source per pixel". For the scanned observation, we would then predict about 1.48e7/0.8/(12/0.0508) = 78317 e^–, or for GAIN=4 about 19500 DN in each pixel of the FLT image. The contribution from the dark current is negligible, and examination of the FLT file for exposure ODQF11UOQ shows that the mean count rate in the column corresponding to this wavelength is indeed very close to 19500 DN, although the actual value fluctuates from row-to-row by up to 10% due to variations in the scanning rate during the observation. To account for scan rate variations, fringing effects, and other approximations in the ETC calculations, we recommend keeping the planned peak source counts for GAIN=4 STIS CCD observations below about 100,000 e^– in each pixel of the 2D image.

Setting Fringe-Flat Exposures for Spatial Scans

The supported TARGET=CCDFLAT exposures available in APT to provide tungsten lamp IR fringe flats, defaults to relatively modest exposure times intended to support the S/N requirements of pointed observations. Using this option also prohibits exposure times greater than these default values from being used.

Since scanned observations can collect many more photons per resolution element than is possible in pointed observations, it may sometimes be useful to take deeper tungsten lamp images than are supported with TARGET=CCDFLAT. This can be done when available modes are enabled by setting TARGET=NONE, and adding the special requirement LAMP=TUNGSTEN. It is however then the responsibility of the user to set the exposure and instrument parameters, as the default settings differ from those used with TARGET=CCDFLAT. Note that APT does not recognize TARGET=NONE exposures as CCDFLAT exposures and will complain of missing fringe flats whether or not the parameters are set correctly. However, sometimes this warning will be spurious, as the appropriate TARGET=NONE, LAMP=TUNGSTEN exposures are actually there. It is important to note that setting TARGET=NONE is an available-but-unsupported mode and will require scientific justification.

The overhead on each individual tungsten lamp exposure can be significant, and it will often be significantly more efficient to take a few deep lamp exposures rather than a number of shorter ones using the CCDFLAT default times.

The fringing correction is very sensitive to the exact distribution of wavelengths falling on each pixel of the detector, and shifts in instrument alignment can cause noticeable changes over the course of a single visibility period. For this reason, the science and fringe-flat exposures should be accurately registered (e.g., if taken through different apertures), and it may sometimes be useful to intermingle fringe-flat lamp exposures with the external science exposures, even if this reduces the time available to observe the external target.