Cosmic Origins Spectrograph Instrument Handbook for Cycle 17

7.4 Acquisition Effects on Data Quality

---

7.4.1 The HST PSF at the COS Aperture

The HST Point Spread Function (PSF) has been modeled at the nominal position of the COS Primary Science Aperture (PSA) - see detailed discussion in Section 13.4. These calculated PSFs are based on the known aberrations present in the HST optical design and the surface errors present in the HST primary and secondary mirrors. While the HST primary and secondary mirrors are among the most precise mirrors ever produced, these optics still exhibit a number of zonal surface errors that limit the quality of the PSF, especially at ultraviolet wavelengths.

The models predict that at least 95% of the energy in the HST PSF is contained within the 2.5 arcsec diameter COS PSA at both 1450 and 2550 Å. The contributions of the PSF that fall outside of the PSA are primarily due to the surface errors in the HST optics themselves, not to the low-order aberrations present in the HST optical design.

All of the light passing through the COS aperture is fully corrected for HST aberration by the downstream COS optics. NUV images and spectra are fully corrected for all optical effects. FUV spectra are fully corrected for aberration, but contain a small amount of astigmatism which does not affect spectral resolution or photometric quality. As verified in ground testing, the resultant PSF at the detector for NUV imaging mode is characterized by a FWHM of approximately 2 pixels (~0.05 arcsec) as discussed in Section 6.4. Scattered light and spectral resolution at the FUV and NUV detectors are well within design specifications. The following sections use the results of this modeling to consider the ramifications of target acquisition centering accuracy upon various aspects of data quality.

7.4.2 Centering Accuracy and Photometric Precision

Figure 7.1 shows the relative transmission of the PSA as a function of displacement of a point source from the aperture center, computed using the modeled PSFs described in Section 13.4.2. Obviously any mis-centering of a source leads to some loss of throughput, but that loss is less than 1% if the source is within 0.4 arcsec of aperture center and is less than 5% if the displacement is less than 0.65 arcsec. In other words, the signal-to-noise achieved in an observation is little affected by small centering errors. This means that a target may be displaced by as much as 0.4 arcsec from the aperture center without a significant loss of throughput, and also that two sources separated by as much as ~1 arcsec (i.e., ±0.5 arcsec) will both have essentially full throughput.

Figure 7.1: Relative Transmission of the COS PSA at 1450Å


 
The transmission is shown as a function of displacement from aperture center. The calculation was done for a point source and for the HST PSF at 1450 Å. Note that the absolute transmission with a point source centered is at least 95%. The same curve for 2550 Å is essentially identical.
 

The spatial resolution of COS was measured during ground tests and is at least sufficient to separate spectra of equally bright objects that are 1 arcsec apart in the cross-dispersion direction; see Section 5.1.3.

7.4.3 Centering Accuracy and the Wavelength Scale

If an accurate wavelength-calibrated spectrum is desired, one wants the error contribution from mis-centering to be low compared to other sources of uncertainty. For NUV ACQ/IMAGE acquisitions, a resel (resolution element) is 3 × 3 pixels. In order to not contribute significantly to zero-point uncertainty, then, the centering should be good to about 0.1 resel. The NUV plate scale is 42.3 pixels per arcsec, so the goal for centering is about 0.01 to 0.02 arcsec. For the FUV channel, resels are 6 pixels wide but the plate scale is reduced, with the result that again the desired centering precision is 0.01 to 0.02 arcsec. Simulations of COS imaging acquisitions have been calculated that show that a centering precision of about 0.02 arcsec is, in fact, feasible.

Dispersed-light acquisitions, whether with the FUV or NUV detector, are unlikely to achieve such a high pointing precision without requiring additional peak-ups in both the cross-dispersion and along-dispersion directions. Dispersed-light acquisitions with COS are slower than an imaging acquisition because HST must be moved to get the information needed to determine the object's centroid. The procedure for dispersed-light acquisitions is discussed below.

As just noted, the throughput of COS is little affected by mis-centering of the source, and so a very high centering precision is not necessary if your science goals do not require a good wavelength zero point. For example, the spectra of some objects may include foreground interstellar or inter-galactic absorption lines that can serve to establish relative velocities.

The plate scales for all the COS gratings along the direction of dispersion are listed in Table 5.2.

7.4.4 Centering Accuracy and Spectroscopic Resolution

Figure 7.2 shows the effect on spectroscopic resolving power of displacing a point source in the PSA. The measurements were calculated using ray tracing for grating G130M. The net effect is that there is no loss of spectroscopic resolution with a displacement as large as 0.5 arcsec.

Figure 7.2: Spectroscopic Resolving Power Versus Source Displacement in the Aperture for Grating G130M.