Cosmic Origins Spectrograph Instrument Handbook for Cycle 17

5.1 The Capabilities of COS

---

COS has two channels, one for the Far Ultraviolet (FUV), and one for the Near Ultraviolet (NUV). Both channels use photon-counting detectors, but those detectors are very different, and in many other ways as well the two channels of COS are used in substantially different ways. Both channels also offer a selection of diffraction gratings that you may use to choose either medium- or low resolving power, with good throughput at any ultraviolet wavelength. Only one of the two channels may be in use at any one time.

This section starts with an outline of the spectroscopic capabilities and expected data quality for COS.

Table 5.1: COS Spectroscopic Modes

Grating	Useful wavelength range (Å)1	Bandpass per exposure (Å)	Resolving Power2 R = /	Dispersion (mÅ pixel-1)
FUV Channel
G130M	1150 -1450	2923	20,000 - 24,000	9.97
G160M	1405 - 1775	3604	20,000 - 24,000	12.23
G140L	1230 - 2050	>820	2,000 - 5,000	80.3
NUV Channel
G185M	1700 - 2100	3 × 35	16,000 - 20,000	37
G225M	2100 - 2500	3 × 35	20,000 - 24,000	33
G285M	2500 - 3200	3 × 41	20,000 - 24,000	40
G230L	1650 - 32005	(1 or 2) × 398	1,550 - 2,900	390

1The useful wavelength range is the expected usable range realized in each grating mode. Note that G140L is set so that Lyman- falls in the gap between the two micro-channel plates to minimize the effects of geocoronal glow. With G140L, one half records 1230 - 2050 Å. The other half records whatever spectrum is detected below 1100 Å, but that is expected to be very little in most cases, hence the 820 Å nominal bandpass. The response of COS below Lyman- will be evaluated after launch. 2The lesser value of R is realized for the low-wavelength end of the useful range, and R increases roughly linearly with wavelength. 3The inter-segment gap misses 14.3 Å. 4The inter-segment gap misses 18.1 Å. 5Some shorter wavelengths are recorded in second-order light. These are listed in Table 5.4.

COS also incorporates an imaging capability in its NUV channel (see Chapter 6 and Chapter 7).

5.1.1 Signal-to-noise Considerations

In ground testing, the COS FUV channel was capable of routinely delivering fully reduced spectra with a photon-noise-limited signal-to-noise (S/N) ratio of ~18 per resolution element in a single exposure. In order to achieve higher S/N, COS can move the spectrum in small amounts so that it falls on different parts of the detector. The use of this FP-POS option (see Section 5.8) at four positions can improve the S/N to 35. Photon-limited S/N values as high as 62 have been demonstrated during ground testing for wavelength regions in which good calibrations were available. The COS calibration program will test and confirm these results after installation in HST, with a goal of achieving S/N = 100 by using astrophysical flat-field sources.

In the NUV channel, we expect to achieve S/N comparable to what has been possible with STIS, namely 100:1 or better.

For more about signal-to-noise, see Section 5.8.

5.1.2 Photometric (Flux) Precision

The limits on the precision and accuracy of fluxes measured with COS are expected to be the same as for STIS. COS has the advantage of a fairly large aperture so that there are only small aperture losses (at most 5%; see Section 13.4). The photometric capabilities of COS will be tested after it is installed, but for now we take them to be the same as STIS, namely 5% accuracy on absolute fluxes and 2% on relative fluxes (within a single exposure). The experience with the NUV MAMA of STIS shows that the repeatability of a flux is good to well under 0.5%. The level of repeatability for the FUV detector is not yet known.

5.1.3 Spatial Resolution and Field of View

The spatial resolution of COS is inherently limited by the aberrated Point Spread Function of HST. Ground tests show that COS can separate spectra of two equally bright objects that are 1 arcsec apart in the cross-dispersion direction for either the FUV or NUV channel. The NUV channel's optics can correct the aberrations so that the NUV imaging capability is diffraction limited (see Chapter 6).

The field of view of COS is obviously determined by the entrance apertures that are 2.5 arcsec in diameter, but the aberrated light entering the aperture means that objects up to 2 arcsec from the center of the aperture will be visible in the recorded spectra.

5.1.4 Wavelength Accuracy

The COS specifications for absolute wavelength uncertainties within an exposure are:

• 15 km s^-1 for medium-resolution spectra (the "M" gratings),
• 150 km s^-1 for G140L, and
• 175 km s^-1 for G230L.

The error budget for wavelength accuracy for the various gratings then breaks down as shown in Table 5.2. Note that all quantities are 1. To arrive at the last two columns, the error budget has been divided equally between internal and external sources. The internal sources include the accuracy of the wavelength scale, the dispersion relation, aperture offsets, distortions, and drifts. The external tolerance budget is dominated by target mis-centering in the aperture. For more on this subject, see Section 7.4.3.

Table 5.2: Wavelength Calibration Uncertainties.

Grating	Error goal (1)		Internal error (1)	External error (1)	Plate scale1
	km s-1	pixels	pixels	arcsec	pixel arcsec-1
G130M	15	5.7 - 7.5	3.0 - 4.0	0.09 - 0.12	45.1
G160M	15	5.8 - 7.2	3.1 - 3.8	0.10 - 0.12	44.6
G140L	150	7.5 - 12.5	4.0 - 6.6	0.12 - 0.21	47.1
G185M	15	7.2 - 10.0	1.2 - 1.7	0.03 - 0.04	41.85
G225M	15	9.7 - 13.3	1.6 - 2.3	0.04 - 0.06	41.89
G285M	15	9.7 - 14.7	1.6 - 2.6	0.05 - 0.07	41.80
G230L	175	8.3 - 15.5	1.4 - 2.6	0.03 - 0.07	42.27

1The plate scale is shown to indicate the centering precision needed during acquisition. The values are for the along-dispersion direction.

Tests of COS on the ground before flight showed some motion of the grating mechanisms (OSM1 and OSM2) after they were stopped in their nominal positions. This drift is small but significant enough for the first few minutes to potentially degrade a spectrum in wavelength. It is to properly calibrate this effect that the "TAGFLASH" operating mode was designed. TAGFLASH mode means using TIME-TAG observations with Optional Parameter FLASH=YES (the default), and in this mode the wavelength calibration lamp is exposed periodically during science observations so that any drift can later be removed. Because the wavelength calibration spectra are recorded on the detector well away from the science spectrum, one does not contaminate the other. TAGFLASH is described further in Section 5.7.1.

5.1.5 Scattered Light in COS Spectra

Figure 5.1 shows an example of an observation obtained during ground testing. A CO absorption cell was placed into the input beam while observing in the FUV with a continuum source. The inset shows an enlargement in the bottom of one of the absorption troughs, showing that any light scattered along the dispersion direction is well under 1% of the nearby continuum.

Figure 5.1: Scattered light in the FUV.


 
Shown is a test exposure obtained during ground calibrations using a CO absorption cell. The inset shows an enlargement of the bottom of one of the absorption troughs.
 

5.1.6 Spectroscopic Resolving Power

The available spectroscopic resolving powers (R) available to observers with COS were listed in Table 5.1. Note that no single value of R applies to any one grating, instead it depends on wavelength, with R .

R also depends on the position of the source in the COS aperture; this is shown in Figure 7.2.

Use of the BOA leads to a degradation of R by factors of 3 to 5; see Section 13.1.3 for more information.

5.1.7 Sensitivity

Measurements of the throughputs of the COS optical systems on the ground indicate that COS will be considerably more sensitive than STIS and earlier generation HST instruments at comparable spectral resolutions, particularly in the far ultraviolet. The point source sensitivities (S) for the COS spectroscopic modes are shown in Figure 5.2, Figure 5.3, and Figure 5.4 Note that these are shown per resel, not per pixel. The reader is reminded that the definition of FEFU may be found at Section 1.1.2. A resel is a "resolution element," which is 6 pixels wide for the FUV channel and 3 pixels for the NUV. Thus the per-pixel sensitivities are lower than shown by these factors.

Figure 5.2: Far-ultraviolet Sensitivity Curves for COS.


 
The values shown are counts per resel per second per FEFU, and are for point sources. Please note that these data are plotted for display purposes only and that those planning observations should use the ETC to get accurate estimates. Also note that these data are pre-flight measurements that will be verified after COS is installed. An FUV resel is 6 pixels wide, so the sensitivity per pixel is 6 times lower than the value indicated here.
 

An estimate of the number of counts (N) expected per resolution element in an amount of time (t) for a source flux (F) is given by N = SFt. As an example, with the COS G130M grating at 1300 Å an exposure time of approximately 1900 seconds is required to reach S/N = 15 per 0.066 Å resolution element (R ~ 20,000) for an object with F₁₃₀₀ 10 FEFU.

The sensitivities illustrated here are based on a preliminary analysis of pre-flight test data. Proposers are urged to use the facilities in the COS ETC when planning observations because the ETC uses the most up-to-date information available.

Figure 5.3: Near-ultraviolet sensitivity curves for COS medium-resolution gratings.


 
The values shown are in counts per resel per second per unit FEFU, and are for point sources. The data shown are pre-flight measurements from ground tests. An NUV resel is 3 pixels wide, so the sensitivity per pixel is 3 times lower than the value indicated here.
 

 
Figure 5.4: Sensitivity Curve for Grating G230L.


 
The values shown are in counts per resel per second per unit FEFU, and are for point sources. NUV resels are 3 pixels wide. The data shown are pre-flight measurements from ground tests. An NUV resel is 3 pixels wide, so the sensitivity per pixel is 3 times lower than the value indicated here.
 

5.1.8 Sensitivity to second-order spectra

COS has been designed to avoid contamination of the first-order spectra by any second-order light. In the FUV channel, second-order light is suppressed by the three reflections from optics coated with Al and MgF₂ (two in the HST OTA plus the COS FUV optical element that is chosen). The NUV channel is susceptible, however, because the MAMA detector is sensitive to light down to 1150 Å. This means that NUV settings above about 2300 Å may be vulnerable.

To mitigate this problem, the NUV optics in COS are optimized for NUV wavelengths to provide peak reflectivities between 1600 and 2000 Å. In addition, two of the NUV gratings (G225M and G285M) have bare aluminum surfaces which has poor FUV reflectivity. Given four such reflections, light from below 1250 Å is reduced by 99%. A 2 mm thick fused silica order-sorting filter was placed in front of two of the NUV gratings (G225M and G285M) as well as in front of MIRRORA/MIRRORB so that light passes through it twice, reducing second-order light to very low levels.

The result from ground tests is that only G225M shows measurable second-order throughput, but even then the second-order light was suppressed by factors of 3,000 to 10,000. Because most astrophysical sources decline in flux in going to shorter wavelengths, it is expected that second-order contamination in COS will be insignificant.