5.5 Image Anomalies
5.5.1 Dust Motes
Several circular patterns, consisting of a dark ring with a bright center, are visible in the ACS flat fields, with typical diameters of ~30 pixels on HRC and ~100 pixels on WFC. These artifacts are shadows of dust on the CCD windows and are weaker on the f/25 WFC than on the f/68 HRC. These motes can be seen in the WFC and HRC flats in Figure 5.13 and Figure 5.14. Since the shapes and depths of these motes are almost independent of wavelength, their effects will be removed by the flats to <<1%, unless any of these particulate contaminants move to different positions on the CCD windows. In case of particulate migrations, the internal lamp flats have a lower f-ratio with a wider angular distribution and cannot be used to patch the flat fields, because they wash out the mote shadows. To correct for new motes, patches to the pipeline flats must be made using the original laboratory flats, corrected for the low-frequency flats derived inflight or, for short wavelengths, using observations of the bright earth.
Larger motes are sometimes present due to dust and blemishes on several ACS filters, including F606W, POL0V, and POL60V. Because the filters are located farther from the detector windows in a converging light beam, imperfections on the filters produce an out of focus image at the detector, where typical mote diameters are ~350-400 pixels on the WFC and ~250-300 pixels on the HRC. One of these large motes can be seen on chip 2 of the WFC F606W flat in Figure 5.13.
Until April 2004, the positioning accuracy of the filter wheels has been within +/- one motor step of the nominal position. This delta corresponds to a distance on the detectors of ~18 HRC pixels and ~20 WFC pixels. Features with sharp transmission gradients at the filter wheels cause a corresponding flat field instability, where errors are 1-3% for a few pixels near the blemishes. If the filter wheel lands in a different place the dust mote will move. Blemish misregistration is an error in the pixel-to-pixel high frequency component of the flat fields and is not related to the low-frequency L-flat correction, which has been applied for all standard and polarizing filters. For details on the L-flat correction, see Section 5.4.2.
This problem was recognized and addressed before launch by a laboratory calibration campaign to obtain flat fields at the nominal position and at plus and minus one step for the F606W+CLEAR on the WFC and for the two POLV filters in combination with the highest priority F475W, F606W, and F775W filters on both HRC and WFC. Since the resolver position uniquely determines the filter wheel step, the ACS pipeline data processing has been enhanced to automatically apply the proper flat for the wheel step position. The keyword FWOFFSET has been added to the ACS image headers to indicate the position of the filter wheel. In April 2004 an update was made to the ACS flight software and the filter wheel is now always positioned at its nominal position. For more information on flat fields for filter wheel offset positions, refer to ACS ISR 03-11.
5.5.2 Optical Ghosts and Scattered Light
The ACS was designed with a requirement that no single straylight feature may contain more than 0.1% of the detected energy in the object producing it. This goal has generally been met, but during the extensive ground and SMOV test programs a few exceptions have been identified (Hartig et al. 2002, Proc SPIE 4854) such as the WFC elliptical haloes and the F660N ghosts. While these ghosts exceed the specified intensity, their origin and characteristics are well defined and they should have minimal impact on the ACS science program. We describe below in more detail some of the optical ghosts relevant for ACS.
- WFC elliptical haloes: These ghosts are caused by reflection from the CCD surface (which lies at an ~20 degree angle to the chief ray) up to the detector windows and back to the CCD. They show up as pairs of elliptical annuli, aligned along the negative diagonal of the FOV (see Figure 5.15), and are observed when bright sources are placed on the lower right (D amplifier) quadrant of the WFC detector. The surface brightness of the annuli increases and size decreases with proximity to the corner. Two pairs of ghosts are seen, produced by reflection from the four window surfaces. The total energy fraction in each ghost may exceed 0.2% of the target signal.
- F660N ghosts: The F660N narrow band filter produces pairs of relatively bright circular annuli stationed near to (but radially outward from) the target image (see Figure 5.16). This is due to reflection from the two surfaces of the second "blocker" substrate back to the many-layer dielectric stack on the first substrate, which in turn reflects at high efficiency at the filter wavelength range. These haloes contain ~2% of the detected target energy and are always ~10 and 20 pixels in diameter.
- Annular ghosts: Large annular ghosts near their parent images are caused by reflection from the detector windows, back to the filters, then returning to the CCD. The fringes arise from interference between the HeNe laser light reflections from the two surfaces of the windows (Figure 5.15). Another type of annular ghost arises from reflections between the inner and outer window surfaces; these are much smaller in diameter, relatively low in intensity (well within the specification) and are displaced radially from the parent image by a small amount.
In general, little can be done about ghosts and blooming in the post-observation data processing phase. Instead, some judicious planning of the actual observations, particularly if bright sources are expected in the field of view, is recommended. For instance, the impact of diffraction spikes (which for ACS lie along X and Y axes) and of CCD blooming (which occurs along the Y direction) due to a bright star, can be reduced by choosing an ORIENT which prevents the source of interest from being connected to the bright star along either of these axes. Alternatively, a suitable ORIENT could move the bright star(s) into the interchip gap or off the field of view altogether. Similarly, the impact of WFC elliptical haloes can be minimized by avoiding a bright star in the quadrant associated with amplifier D.
Figure 5.15: Optical ghosts, diffraction spikes, and blooming. This image was obtained during ground calibration at Ball Aerospace using a HeNe laser (633nm) through F625W. It is a factor of ~50,000 saturated with 9 highly saturated point sources in and just off of the field of view. Note the diffraction spikes along the X and Y axes of the WFC, and the severe blooming of the charge along the Y axis. Several optical ghosts with different origins and intensities (see text for details) are visible: (a) WFC elliptical haloes show up in the lower left as pairs of elliptical annuli aligned along the negative diagonal of the FOV; (b) Large annular ghosts are seen near their parent images; (c) To the lower right, there exist smaller fainter annular ghosts which tend to be displaced radially from the parent image by a small amount. 
Figure 5.16: F660N optical ghosts. This 700 s F660N exposure illustrates the F660N ghosts, namely pairs of relatively bright circular annuli stationed near to (but radially outward from) the target image.These haloes contain ~2% of the detected target energy and are always ~ 10 and 20 px in diameter.
5.5.3 Cross Talk
Images obtained with the WFC are affected by a small amount of electronic cross talk between the four CCD quadrants that correspond to the four amplifiers of the two detectors. The effect produces electronic "ghost" (i.e., negative) images in a given quadrant that mirror real images recorded on other quadrants. Figure 5.17 shows an example of the cross talk in one of the images obtained during the Great Observatories Origins Deep Survey (GOODS) program, a pointing in the GOODS-S field taken through the F850LP passband. Source images a, b, c and d, which are recorded on quadrant C, produce negative ghost images a', b', c', d', a'', b'', c'', d'', and a''', b''', c''' and d''' in quadrants A, B and D, respectively.
Inspection of images affected by cross talk reveals a number of interesting characteristics of this effect. The cross talk ghosts have a surface brightness that is a few electron lower than those of the average background and there does not appear to be a simple proportionality between the light profile of the sources and the ghosts. Rather, as Figure 5.18 shows, the strength of the cross talk, measured as the average surface brightness of ghosts, seems to be roughly proportional to the logarithm of the sources.
The intensity of the cross talk ghost also appears to vary depending on which quadrant hosts the source and which hosts a ghost. For example, sources in quadrant D of Figure 5.17 produce clearly visible ghosts in quadrant C, but much fainter ghosts (if any at all) in quadrant A and B. Also, sources in quadrant A and B do not seem to produce any obvious ghosts in the other quadrants.
Figure 5.17: Cross Talk in the WFC Detectors. Due to the cross talk, galaxies a, b, c and d recorded in quadrant C have mirror ghost images in quadrants A, B, and D. Note the effect appears stronger in quadrants C and B than in quadrant A, and that sources in quadrant D produce ghosts that follow a similar trend. However, sources in quadrants A and B do not seem as effective in producing ghost images. 
Figure 5.18: The average surface brightness of ghost images plotted as a function of the flux of the sources. All sources are in quadrant C. There is a correlation between the surface brightness of the ghosts and the flux of the sources.
The magnitude of the cross talk depends on the background of the victim quadrant, the sense of the correlation being that the cross talk is stronger when the background is lower. Figure 5.19 shows that in images with low background the cross talk creates ghosts whose surface brightness is ~40% lower than the average background. In images with high background the ghost surface brightness is only ~3% lower.
Tests on existing images show that the cross talk has minimal impact on flux measurements of sources superimposed on ghost images and thus mostly presents a cosmetic issue. The bottom panel of Figure 5.20 shows six growth curves of the same star derived from six individual frames, all with identical exposure time and filter. In three of the frames, due to the adopted dithering strategy, the star is located within a major ghost image (the annulus around the photometric aperture is also within the ghost image), while in the remaining three frames the star is outside the ghost image. The observed photometric scatter is typical for a star with this brightness of post-pipeline processed data (FLT files) with no correction for geometrical distortions and no cosmic ray removal. As a comparison, the upper panel shows the growth curves of a star of comparable apparent magnitude in the same frames but located in an region unaffected by cross talk. The photometry of the latter star has the same amount of scatter as the former one, showing in particular that the cross talk acts as an additive effect and not as a multiplicative one.
Figure 5.19: The dependence of the cross talk, measured as the difference of the sky background in affected and non affected areas, as a function of the sky background. The strength of the cross talk diminishes as the background increases.
Figure 5.20: BOTTOM - Six growth curves of the same star derived from six individual frames, all with identical exposure time and filter. In three of the frames, due to the adopted dithering strategy for the observations, the star is located within a major ghost image (the annulus around the photometric aperture is also within the ghost image), while in the remaining three frames the star is outside the ghost image. The observed photometric scatter is typical, for a star with this brightness, of post-pipeline processed data (FLT files) with no correction for geometrical distortions and no CR removal. TOP - As above but for a star of comparable apparent magnitude and in the same frame that is located in a region unaffected by cross talk. The photometric scatter of the two stars is very similar.
There is evidence that the strength of the cross talk is significantly weaker in images acquired with GAIN=2, than in images taken with GAIN=1. Figure 5.21 shows two exposures through the same filter and with the same exposure time, the former acquired with GAIN=1 (upper panel), the second with GAIN=2 (lower panel). Cross talk ghosts are clearly observed in the former image, but not in the latter one. A more quantitative comparison supports the visual inspection. Figure 5.22 shows the median value of pixels in ghost images as a function of the source pixel flux, i.e., of pixels located in mirror symmetric locations relative to the pixels in the source quadrant.
Observers who do not need to sample the readout noise of the CCD with the accuracy afforded by the GAIN=1 setting, might want to consider to use GAIN=2 in order to minimize the cross talk.
At this time we have not yet determined a robust way to predict the size of a cross talk residual and the exposure parameters on which it depends. Efforts are ongoing to identify algorithms to remove the effect of cross talk from the WFC images and we encourage GO's to look for updates on the ACS STScI web page.
Figure 5.21: Two exposures through the same filter (F814W) and with the same exposure time (22.5 sec), the former acquired with GAIN=1 (TOP), the second with GAIN=2 (BOTTOM). Cross talk ghost images are observed in the former image, but not in the latter one.
Figure 5.22: The median value of pixels in ghost images as a function of the source pixel flux, i.e. of pixels located in mirror symmetric locations relative to the pixels in the source quadrant, derived from the CCD frames shown in Figure 5.17. The images are sky subtracted. Source pixel flux is binned in logarithmic bins 0.2 in size. Using the median ensures that mostly background pixels, as opposed to pixels affected by sources, are being considered in the analysis. Filled symbols are for GAIN=2, hollow one for GAIN=1. The pixels in the image with GAIN=1 have systematically lower values than those of the image with GAIN=2, an indication that the cross talk is stronger in this case. Their values also decrease as a function of the source pixel value more steeply than in the GAIN=2 case.
5.5.4 Scattered Earth Light
Most observers will not experience significant issues with scattered Earth light in their observations. Normally observations are scheduled only when the bright Earth limb is more than 20 degrees from the HST pointing direction. This is sufficient to eliminate serious impacts from scattered Earth light -- the most severe impact will be for observers with targets in the CVZ who may notice the sky background increased by a factor of 2 or 3.
It is possible to make arrangements for observations at smaller bright Earth limb angles, and these images have a potential for serious impacts from scattered light. There are two types of impact: elevated background and non-uniformity in the background. For example, at a bright Earth avoidance angle of 14 degrees it is possible for the sky level to be increased by a factor of 100 compared to normal pointings away from the Earth; this will of course have a serious impact on the background noise and detection of faint targets. Also, non-uniformity can arise since the scattered light is taking an increasingly non-standard path through the HST optics, and hence the flat fielding becomes corrupted. At this same angle of 14 degrees, it is possible to have both large scale gradients across the field of view (up to ~20% amplitude in the WFC) and small scale features in the background (up to ~12% in WFC and ~30% in the HRC). See ACS ISR 2003-05 for more details.
