HST Primer: Scientific Instrument COS

Overview of the Cosmic Origins Spectrograph (COS)

The Cosmic Origins Spectrograph (COS)

The Cosmic Origins Spectrograph (COS) is an ultraviolet spectrograph designed to optimize observations of point sources. It was installed during Servicing Mission 4 in the instrument bay previously occupied by COSTAR. COS is designed to be a very high throughput instrument, providing medium to low-resolution ultraviolet spectroscopy. The instrument has two channels for ultraviolet spectroscopy: Far-ultraviolet (FUV) and Near-ultraviolet (NUV).

Far-Ultraviolet Channel (COS/FUV)

COS/FUV uses a single optical element to disperse and focus light onto a cross-delay-line (XDL) detector; the result is high ultraviolet sensitivity from about 900 Å to 2050 Å with resolving power for different modes ranging from R ~ 1300 to ~ 17,000 (for detailed information on the resolving power of the FUV channel, please see the COS Instrument Handbook). This detector has heritage from the FUSE spacecraft. The active front surface of the detector is curved; to achieve the length required to capture the entire projected COS spectrum, two detector segments are placed end-to-end with a small gap between them. Each detector segment has an active area of 85 × 10 mm, 16384 × 1024 pixels, and a resolution element of 6 × 10 pixels.

The FUV channel has three gratings: G140L provides nearly complete coverage of the FUV wavelength range in a single exposure with a resolving power R ~ 1300 to 3500. G130M spans wavelengths between 900 Å and 1450 Å, and G160M covers the wavelength range between 1400 Å to 1775 Å; both these gratings provide resolving power R between 13,000 and 17,000 for wavelengths longer than 1150 Å (see the figure below for resolving power at shorter wavelengths). For all three FUV gratings, a small segment of the spectrum is lost to the gap between the two detector segments. This gap can be filled by obtaining two exposures offset in wavelength.

The COS FUV detector is susceptible to gain sag, a reduction in the ability of the detector to convert incoming photons into electrons. One strategy for mitigating this, and subsequently extending the lifetime of the COS/FUV detector, is to occasionally change the location along the cross-dispersion direction where spectra are recorded on the detector, the lifetime position (LP). In the interests of extending the lifetime of COS operations until 2030, from Cycle 30 onwards, COS currently operates under a hybrid LP-mode with the following default LPs: G130M 'blue modes' = LP2, G140L = LP3, G130M/1222 = LP4, G130M standard modes = LP5, G160M = LP6. G160M exposures that are approximately longer than half an orbit use LP6, while shorter G160M exposures may use LP4 to reduce overheads, if requested in the Phase I proposal.

In 2017, a set of restrictions on detector segment, FP-POS position, and target acquisition mode were introduced to maximize the lifetime of the FUV detector. Known as the COS2025 policies, these restrictions are still in effect and are discussed in full on the COS2025 website.

Two new central wavelengths (cenwaves) were first offered in Cycle 26 for COS/FUV observations, named G140L/800 and G160M/1533, and remain supported this cycle. The G140L/800 setting allows for contiguous coverage of the entire spectral region from 800 to 1950 Angstroms on a single detector segment (FUVA) with a low spectral height below 1150 Angstroms, allowing higher S/N for background-limited observations. The G160M/1533 setting extends coverage at the short-wavelength end of G160M by 44 Angstroms to overlap with the longest wavelengths covered by cenwave G130M/1222, and is otherwise very similar to the existing G160M/1577 cenwave. This allows a broad range of FUV wavelengths to be covered at medium resolution by just two cenwave settings (1222+1533). 

The G140L/800 mode offers an absolute flux accuracy of 5%, a relative flux accuracy of 2%, and a wavelength solution accurate to within 150 km/s, comparable to the other G140L cenwaves. The G160M/1533 mode offers an absolute flux accuracy of 5%, a relative flux accuracy of 2%, and a wavelength solution accurate to within 7.5 km/s, comparable to the other G160M cenwaves.

The calibrations of the COS blue modes (cenwaves G130M/1055 and G130M/1096) were improved in 2020 to bring their flux and wavelength uncertainties in line with those of the other G130M cenwaves.

For full details, see the COS Instrument Handbook.

Spectral resolution of the COS/FUV detector as a function of lifetime position for selected cenwaves. These predictions have been derived from optical models and confirmed with on-orbit observations.

Near-Ultraviolet Channel (NUV)

COS/NUV uses a 1024 × 1024 pixel Cs2Te MAMA detector that is essentially identical to the STIS/NUV MAMA except that it has a substantially lower dark count. Four gratings may be used for spectroscopy. Portions of the first-order spectra from the gratings are directed onto the detector by three separate flat mirrors. Each mirror produces a single stripe of the spectrum on the detector. For the low-dispersion grating, G230L, one or two first-order spectrum stripes are available, each covering ~400 Å of the entire 1700 Å to 3200 Å range at resolving power R ~2100 to 3900, depending upon wavelength. For high-dispersion gratings G185M, G225M, and G285M, three non-contiguous stripes of 35 Å are available in each exposure at resolving powers of 16,000 to 24,000. Panchromatic (1650 Å to 3200 Å) images of small fields (less than two arcseconds) may also be obtained at ~0.06 arcsec or greater resolution in imaging mode.

Because of declining throughput, NUV observations with the G285M grating are now designated Available-but-Unsupported.  Users desiring medium-resolution spectra in the 2500-3000 Å region will generally find STIS to be preferable, and, in any case, must explicitly justify use of G285M.


Prolonged exposure to light causes the COS FUV XDL detectors to become less efficient at photon-to-electron conversion, a phenomenon called “gain sag.” When a particular region of the detector is increasingly used, there is a corresponding decrease in the “pulse height” of the charge cloud generated by an individual photon. As long as the pulse heights are above a minimum threshold needed to distinguish real photons from background events, there is no loss in sensitivity. But as the average pulse height in a particular region approaches and then drops below this threshold, real photon pulses are increasingly misidentified as background, causing a decreasing effective throughput. Since the amount of gain sag increases with the total amount of the previous illumination, the effects first appear in regions of the detector that are illuminated by the bright Lyman Alpha airglow line, but eventually, the entire spectrum is affected.

STScI is undertaking a number of actions to mitigate the effects of gain sag and extend the lifetime of the COS FUV XDL detector, including a series of lifetime moves to unsagged portions of the detector. See the general FUV material above for details about the full complement of lifetime positions, and see the COS Instrument Handbook for more information.

A two-zone spectrum extraction algorithm is used by the COS pipeline to calibrate data obtained at LP3 and subsequent lifetime positions. This extraction method is optimized only for observations of point sources using the PSA aperture. FUV spectra of extended sources and FUV spectra obtained with the BOA will be poorly calibrated. As a result, proposers planning to use COS/FUV for science observations with the BOA aperture (whether point source or extended) should consider carefully the effects of poor calibration on their science goals. See the COS webpage for updated information on the calibration of all lifetime positions.

Optimizing the Science Return of COS

Fixed-pattern noise in the COS detectors limits the signal-to-noise that can be achieved in a single exposure (see Section 5.8 of the COS Instrument Handbook). A simple way to remove these detector features is to obtain exposures at multiple FP-POS or CENWAVE settings, both of which shift the spectrum on the detector, and combine them in wavelength space. This is especially important for the COS FUV detector as the fixed pattern noise is larger and more poorly characterized than that of the NUV detector. In addition, the consistent use of multiple FP-POS positions in the G130M, and G140L/800 and G140L/1105 settings will spread the bright geocoronal Lyman Alpha illumination and significantly delay the appearance of gain sag effects. 

Since this simple shift-and-add technique significantly improves the signal-to-noise ratio of the resulting spectrum, and will extend the lifetime of the COS FUV detector, the use of four FP-POS is required for FUV observations except as indicated below. This can be achieved for each CENWAVE with the FP-POS=ALL parameter in APT, or by manually distributing the four FP-POS over one or more visits to the same target.

The following are exceptions to the requirement that all FP-POS be used:

  • When segment B is on for G130M/1291, only two FP-POS (3 and 4) are available, and both must be used.
  • For G160M observations, fewer than four FP-POS may be used if the intended S/N is ≤ 25. See Section 9.5.1 of the COS Instrument Handbook for details.
  • If there is a strong scientific justification not to use all four FP-POS for each CENWAVE, it must be stated in the Phase I proposal. A modest reduction in observational overheads will not normally be considered sufficient justification.

COS exposures may be obtained in either a time-tagged photon address (TIME-TAG) mode, in which the position, arrival time, and pulse height (for FUV observations) of each detected photon are saved in an event stream, or in accumulation (ACCUM) mode, in which only the positions of the photon events are recorded. In TIME-TAG mode, which is the default observing mode, the time resolution is 32 milliseconds. ACCUM mode is designed for bright targets with high count rates that would otherwise overwhelm the detector electronics. Because the lower information content of ACCUM data reduces their utility for archival researchers, its use must be justified for each target. For details, see Section 5.2 of the COS Instrument Handbook.

In both TIME-TAG and ACCUM mode, the Astronomer’s Proposal Tool (APT) automatically schedules wavelength calibration exposures, either during science exposures or between them (see Section 5.7 of the COS Instrument Handbook). The COS data reduction pipeline (CALCOS) uses these data to adjust the zero-point of the wavelength solution for the extracted spectra. It is possible to suppress the taking of wavelength calibration spectra, but since it significantly lessens the archival quality of COS data, it must be justified.

Observers who wish to employ non-optimal observing techniques must strongly justify their observing strategy in the “Description of Observations” section of their Phase I proposal. Non-optimal observing techniques should not normally be adopted solely for the purpose of producing a modest reduction of the observational overheads; in such cases, the observer should normally just request adequate time to use the recommended optimal strategy (see HST Preparation of the PDF Attachment). 

Next: HST Primer: Scientific Instrument STIS

Related Links

COS Instrument Handbook