The ACS filter wheels include four dispersing elements for low resolution slitless spectroscopy over the field of view of the three ACS channels. One grism (G800L) provides low resolution spectra from 5500 Å to 10,500 Å for both the WFC and HRC. A prism (PR200L) in the HRC covered 1700 Å to beyond 3900 Å, although reliable wavelength and flux calibration was guaranteed only up to 3500 Å. In the SBC a LiF prism (PR110L) covers the range 1150 Å to ~1800 Å and a CaF
2 prism (PR130L) is useful from 1250 Å to ~1800. The grism provides first order spectra with almost constant dispersion as a function of wavelength but with second order overlap beyond ~10,000 Å. The prisms have non-linear dispersion with maximum resolution at shorter wavelengths and much lower resolving power at longer wavelengths.
Table 6.3 summarizes the essential features of the four ACS dispersers in the five supported modes.
Figure 6.17 shows the full G800L spectrum of the white dwarf GD153 (V=13.35 mag) obtained in one 60 second exposure. The first order is contaminated by the second order beyond ~10,000 Å. The total flux in the zeroth order is 2.5% of that in the first order, so locating the zeroth order is a less effective method of obtaining the wavelength zero point of weak spectra than using a matched pair of direct and grism images. The third and fourth orders contain about 1% of the flux in the first order, and the negative orders contain about 0.5% of that flux. When bright objects are observed, the signal in fainter orders may be mistaken for the spectra of fainter objects. In crowded fields, many spectral orders from different objects may overlap.
Table 6.3 lists the linear dispersion for the first and second order spectra, but their dispersions are better described with second order fits (ISR ACS 2005-08). Because the grism is tilted with respect to the optical axis, the wavelength solutions are field dependent. This dependence has been calibrated within 0.5 pixels over the whole field; the linear dispersion varies by ±11% from center to corner. The full extent of the spectrum of a bright source (orders -2, -1, 0, 1, 2, 3) is 1200 pixels (60 arcseconds). The higher spectral orders are not in focus, so their spectral resolutions are smaller than expected from their nominally higher dispersions.
G800L provided higher spatial resolution with the HRC than with the WFC, but the spectra were tilted at −38
° with respect to the HRC’s X axis.
Figure 6.18 and
Figure 6.19 show the ranges and sensitivities of the zeroth, first, and second orders as a function of wavelength and pixel, respectively.
Figure 6.20 shows the spectrum of the standard star GD153. Orders -1 through +2 span about 70% of the 1024 detector columns, and the +2 order overlaps the +1 order beyond ~9500 Å. The dispersion varies by ±2% from the center to the corners of the detector. Because of the HRC’s limited FOV, many G800L spectra were truncated by the edges of the detector or originated from objects located outside the corresponding direct image.
Figure 6.21 shows the sensitivity versus wavelength and the wavelength range of the HRC pixels for prism PR200L. The dispersion peaked at 5.3 Å/pix at 1800 Å, but dropped to 105 Å/pix at 3500 Å and then to 563 Å/pix at 5000 Å. Consequently, the spectrum piled up at longer wavelengths, where 1500 Å of spectrum was spanned by only 8 pixels. For bright objects, this effect could lead to saturation and blooming of the CCD, which could affect other spectra. The dispersion also varied by ±4% at 2000 Å between opposite corners of the detector. The tilt of the prism caused an offset of up to ~250 pixels between the CCD positions of the direct image and the PR200L spectrum of an object (
Figure 6.21), and caused a similar amount of vignetting along the low-x side of the HRC image. Consequently, an additional prism aperture was defined with a reference point offset by 7.4 arc seconds from the geometric center of the CCD. The wavelength solution used by the aXe data reduction software (see
Section 6.3.7) accounts for this aperture offset.
Figure 6.22 shows the sensitivity with wavelength and the wavelength range of the pixels for PR110L. This prism is sensitive below 1200 Å and includes the geocoronal Lyman
α line, so it is subject to large background signal. The dispersion is 2.6 Å/pix at Lyman
α and decreases to 21.6 Å/pix at 1800 Å. The declining efficiency of the CsI MAMA detector beyond ~1800 Å occurs before the long wavelength pile-up, but observations of standard stars indicate that the throughput is ~1000 times higher at 4000 Å than indicated in
Figure 4.11. Observations of stars redder than spectral type F experience significantly higher counts between 2000 Å and 6000 Å, with a peak count rate at ~3500 Å. For G stars, this peak can be ~3 times larger than the maximum UV count rate. This red leak and the geo-coronal Lyman
α must not exceed the MAMA Bright Object Protection (BOP) limits (see
Section 4.6). The prism’s optical tilt causes an offset of up to ~250 pixels between the positions of the direct image and the PR110L spectrum of an object (
Figure 6.22), and causes a similar amount of vignetting along the high-x side of the SBC image.
Figure 6.23 demonstrates this offset with the summed direct image and PR110L image of the standard star WD1657+343. An appropriately offset aperture is automatically implemented by the planning software for all PR110L observations. The wavelength solution used by aXe (
Section 6.3.7) accounts for this aperture offset.
The short wavelength cut-off of the PR130L prism at 1250 Å excludes the geo-coronal Lyman α, so PR130L is the preferred disperser for faint object detection between 1250 Å and 1800 Å. The dispersion varies from 1.65 Å/pixel at 1250 Å to 20.2 Å/pixel at 1800 Å.
Figure 6.24 shows the sensitivity versus wavelength, along with the resolving power (R) and the offset from the direct image in pixels (
Δx) as functions of wavelength. As for PR110L, Bright Object Protection must be considered when using this prism, even though the background count rate is lower (see
Section 4.6). As for the other prisms, the direct and dispersed images use different apertures with a small angle maneuver between them.
Table 6.4 lists the V detection limits for the ACS grism/prism modes for unreddened O5 V, A0 V, and G2 V stars generated by the
ETC. These limits were computed for WFC and HRC using the parameters CR-SPLIT=2 and GAIN=2. An average sky background was used, but users should be aware that limiting magnitudes are sensitive to background levels, e.g., the limiting magnitude of an A0 V star in the WFC using the F606W filter changes by ±0.4 magnitudes at the background extremes.
Chapter 9 provides details of the calculations. Depending on the wavelength region, the background must also be taken into account in computing the signal to noise ratio. The background at each pixel consists of the sum of all the dispersed light in all the orders from the background source. For complex fields, the background consists of the dispersed spectrum of the unresolved sources; for crowded fields, overlap in the spectral direction and confusion in the direction perpendicular to the dispersion may limit the utility of the spectra.
For more detailed, two-dimensional simulations of ACS slitless spectra, an IRAF/PyRAF package called aXeSIM is available.
aXeSIM generates slitless images and their associated direct images using object shapes and spectra given as input. In the most primitive form,
aXeSIM uses Gaussians as object shapes and direct imaging magnitudes as “spectra”, however, more realistic object shapes (e.g. a PSF from
TinyTim) and high resolution spectra can be provided. aXeSIM is described in Kuemmel, Kuntschner & Walsh (
2007, ST-ECF Newsletter 43, 8) and Kuemmel et al. (
2009, PASP 121, 59).
Because ACS spectroscopy is slitless, the point spread function of the target modulates the spectral resolution. For extended sources, the size of the target in the dispersion direction limits the achievable resolution (ACS ISR 2001-02). The dispersions of the grism and prisms are well characterized, but the zeroth order of grism spectra are generally too weak to reliably set the wavelength zero point. For typical spacecraft jitter, wavelength zero points to ±0.4 pixels should be routinely achievable using a direct image taken just before or after the grism or prism image. The jitter information can be used to obtain more accurate coordinates for the center of the FOV. These coordinates allow one to determine better relative offsets between the direct and the spectroscopic images. The red wavelength range of each pixel in G800L images is small enough that fringing can modulate the spectra. The peak-to-peak fringe amplitude was about 30% at 9500 Å for the HRC, and it is about 25% for the WFC. Models of the fringing in the WFC and HRC are described in
ACS ISR 2003-12. In practice, the fringing is significantly reduced by the smoothing effects of the PSF and intrinsic object size in the dispersion direction.
ACS ISR 2008-01 shows that the errors due to fringing are less than 0.1% for continuum sources and can therefore be neglected. For narrow emission lines, however, fringing can cause line flux variations of 12% and more. For realistic scenarios like Wolf Rayet emission lines, variations of ~4% are seen.
The STSCI pipeline does not provide an extracted spectral count rate vs. wavelength, but the software package aXe is available to extract, wavelength calibrate, flat field, and flux calibrate ACS grism and prism spectra. Full details are presented by Kuemmel et al.
2009, PASP 121, 59 and at:
