2.1 Choosing between COS and STIS

With the installation of COS and the repair of the Space Telescope Imaging Spectrograph (STIS), HST has two spectrographs with significant overlap in spectral range and resolving power. Each has unique capabilities, and the decision of which to use will be driven by the science goals of the program and the nature of the target to be observed.

In the far-UV (from about 1100 to 1800 Å), the throughput of the COS FUV channel exceeds that of the STIS FUV MAMA by factors of 10 to 30, and the combination of the spectroscopic resolving power (~17,000) and wavelength coverage (300 to 370 Å per setting) of the medium-resolution COS FUV modes, as well as the extremely low detector dark rate of the XDL detector, results in a discovery space (throughput times wavelength coverage) for observations of faint FUV point sources that is at least 10 times greater for most targets than that of STIS modes with comparable resolution, and as much as 70 times greater for faint, background-limited point sources.

Because of its MgF₂ windows, the STIS FUV MAMA is insensitive to wavelengths below about 1150 Å. The COS FUV XDL detector is windowless and provides useful throughput to at least 900 Å. See Section 5.1.2 for details.

In the near-UV (~1700 to 3200 Å), COS and STIS have complementary capabilities. To accommodate the NUV detector format the COS NUV spectrum is split into three non-contiguous stripes, each of which covers a relatively small range in wavelength. Obtaining a full NUV spectrum of an object requires several set-ups and exposures (six or more for the medium-resolution gratings and three for G230L grating). When broad NUV wavelength coverage is needed obtaining a single STIS spectrum will often be more efficient than taking separate COS spectra. Users should note that for the shorter NUV wavelengths COS modes often have a substantial throughput advantage over the comparable STIS modes, while at the longer NUV wavelengths it can be STIS that has the advantage. They should also consider that the STIS NUV modes have produced a large set of existing observations, while the COS NUV modes have so far seen limited use. As a result, the calibration of the STIS NUV modes is likely to be superior to that of comparable COS modes for the foreseeable future.

After installation into HST in 2009, the dark rate of the COS NUV detector had initially been substantially lower than that of STIS NUV detector. However, the dark rates for the two detectors have converged over the years. Observers are advised to perform detailed calculations using both the COS and STIS ETCs and to consider carefully the relative instrument overheads to determine which combination of instruments and modes is best for their science.

For observations of extended sources, the spatial resolution offered by STIS must be weighed against the superior sensitivity of COS. One of the primary design goals of STIS was to provide spatially resolved spectra in the UV, optical, and near-IR. The STIS long slits, when used with the first-order gratings, allow spatially resolved observations that exploit the intrinsically high resolution of HST over the full width of the detectors (approximately 0.05 arcsec per 2-pixel spatial resolution element over a length of 25 arcsec with the NUV and FUV MAMAs, and approximately 0.1 arcsec per 2-pixel spatial-resolution element over a length of 52 arcsec with the CCD).

COS was optimized for point-source observations. While COS has relatively large entrance apertures (2.5 arcsec diameter), flux from regions more than 0.4 arcsec from the aperture center is significantly vignetted. These large apertures also mean that objects extended in the dispersion direction will yield spectra with lower spectral resolution. In addition, the optical design of the FUV channel limits the achievable spatial resolution; measured values of the FWHM in the spatial dimension vary between about 0.25 and 1.5 arcsec, depending on grating and wavelength (Section 5.1.9). The COS NUV channel uses a different optical design and has a spatial resolution comparable to that of the STIS first-order NUV modes (~0.05 arcsec), with somewhat better sampling. However, for sources extending more than 1 arcsec in the spatial direction, the various NUV spectral segments will begin to overlap.

The line-spread functions (LSFs) of both instruments exhibit non-Gaussian wings due to mid-frequency zonal (polishing) errors in the Optical Telescope Assembly (OTA). Using STIS, one can minimize their effects through the use of narrow apertures. Narrow apertures are not available on COS. The broad wings of the LSF, especially in the short wavelengths of the FUV band, can limit the ability of COS to resolve faint, narrow features, and blend closely-spaced lines. Studies that require accurate knowledge of the line profile will require full consideration of the COS LSF (Section 3.3). The non-Gaussian wings of the COS LSF should have only modest impact on science programs targeting broad lines and continuum sources.

Both COS detectors and the STIS MAMA detectors are prohibited from observing objects that exceed specific brightness levels (see Chapter 10 in this handbook and Sections 13.8 and 14.8 of the STIS Instrument Handbook). Some brightness limits have been established for the health and safety of the instrument, while others are practical limits that are set to ensure good data quality. Because STIS is less sensitive than COS, the brightness limits for STIS tend to be significantly less stringent. In the NUV range, the STIS G230LB and G230MB gratings can also be used with the STIS CCD, which has no bright-object limitations. STIS also has a number of small and neutral-density apertures that can be used with the MAMA detectors to attenuate the light of a too-bright object. COS has only a single neutral-density filter that attenuates by a factor of about 200, but also degrades the spectral resolution by a factor of 3 to 5. In most cases some combination of STIS gratings and apertures will be a better choice for observing a UV-bright object than COS with its neutral-density filter. Users are advised to compare results from the COS and STIS ETCs when deciding on an appropriate strategy for their target.

The STIS high-dispersion echelle modes E140H and E230H have resolving powers of ~114,000 (or even R ~ 200,000 with the 0.1 × 0.03 aperture and specialized data reduction; see Section 12.6, "Improving the Sampling of the Line Spread Function," of the STIS Instrument Handbook), significantly higher than the best COS resolution. Also, STIS can obtain spectra in the optical and near-IR at wavelengths up to 10,200 Å, while the maximum wavelength observable by COS is about 3200 Å.

Both STIS and COS can perform observations in TIME-TAG mode, whereby the time of each photon's arrival is recorded. STIS is capable of a much finer time resolution (125 μs vs. 32 ms for COS), although few programs require such a high sampling rate. Due to its lower sensitivity, STIS may be able to observe a target in TIME-TAG mode that is too bright for TIME-TAG observations with COS. On the other hand, TIME-TAG data acquired with the COS FUV detector includes information on the pulse-height distribution, while TIME-TAG data acquired with the STIS and COS MAMAs do not. Pulse-height information can be valuable in identifying and rejecting background counts in the spectra of faint sources.

Lifetime Impact Considerations

The COS bright-object safety screening limits (Section 10.2) for the FUV were chosen to protect the detector over short time periods. However, very UV bright sources or emission lines can still deplete a significant percentage of lifetime for a given detector Lifetime Position (LP) without violating the screening limits. Approximately 27,000 counts can fall on a pixel before the reduced gain at that location results in significant flux loss. Historically, the brightest COS observations of continuum sources placed as much as 0.1 counts sec^-1 pixel^-1 across FUVA or FUVB for >10,000 sec. In such rare cases, the integrated counts on the detector consumed as much as 5% of the total available lifetime at these regions. Exhausting the gain at a particular lifetime position in an accelerated fashion is inconsistent with the goal of retaining the functionality of the COS FUV detector for as long as the lifetime of Hubble, and we therefore provide guidance on observations that may be better suited for observations with STIS than COS. 

Of primary concern are sources that have a significant continuum over one or both segments with a high enough count rate to negatively impact the lifetime at a particular position when observed for a long duration. This is most likely to be the case for observations of nearby hot stars that last for one orbit or more. Additionally, some emission lines that are as bright or brighter than geocoronal Lyman alpha could potentially cause localized gain holes. More quantitatively, FUV observations where the estimated total counts in the brightest pixel exceed 1% of a pixel's lifetime, or 270 counts in a given pixel, may be more suitable for STIS if the science can be accomplished in that way within a reasonable observing time. For a source where the continuum is responsible for the brightest flux on a pixel, this threshold refers to the total exposure time on a target per grating, irrespective of the number of FP-POS used. For a source dominated by an emission line, the threshold refers to the total exposure time per FP-POS, since the emission line will fall on different parts of the detector with each FP-POS.

The procedure for determining whether an observation could exceed the above threshold is to perform an ETC calculation for the science target and make note of the brightest pixel for each segment. That count rate should be multiplied by the total exposure time per visit if the brightest pixel is on the continuum, or the exposure time per FP-POS per visit if the brightest pixel is an emission line. Finally, multiply the count rate by the number of visits used for the science target to get the total counts for that pixel. For Segment B of G130M/1291 and Segment A of G140L/800 or G140L/1105, the brightest pixel could be geocoronal Lyman alpha. In these cases, users can estimate the brightest pixel due to the target by selecting the "No airglow" option in Question 5c of the ETC. If the resulting calculations shows the possibility of exceeding 270 counts in a given pixel and the science can be accomplished with STIS, it is highly recommended that PI's consider moving to an equivalent mode of STIS (i.e., E140M in place of G130M). COS contact scientists (CSs) will perform similar checks while reviewing Phase II proposals, and reach out to PIs if lifetime impact concerns are identified.