detectors During the ground calibration of STIS, large scattered light halos were identified in images of point sources and long slit spectral images at long wavelength (>750nm). The long wavelength scattering was traced on the SITe 1024 × 1024 CCD and its header package. Since ACS employes SITe CCD, concerns for the performance of ACS led the ACS team to study the long wavelength scattering, in collaboration with SITe, to assest the impact to the ACS science and find out a solution.
The long wavelength halos discovered in CCD images during STIS ground calibration are typically observed to extend a few hundred pixels from the core of the image (figure 1). The fraction of light scattered in the halo increases as a function of wavelength. An optical bench to image pinhole mask and narrow slits has been assembled at Ball laboratories to test a 1k × 1k ACS CCD. The CCD was operated at -80°C and images were processed using standard analysis procedures. Pinhole and long slit images have been taken through five broadband filters at wavelength between 500 and 1000nm.
In order to highlight the extent of the halo, all the images were exposed to saturate the full well in the image core by a factor of 10. Consequently, images show charge bleeding along columns.
|
|
| Figure 1: Images of a point sources at 500 nm (left) and 1000 nm (right) [Click on the image to see an enlarged version] | |
The 535 nm image exhibits only low level diffraction artifacts associated with the imaging optics. The diffraction seen in the 535 nm image is also present in the 1030 nm image, superposed on the diffuse halo. From the long slit images we derived the radial distribution of light in the halo as a function of wavelength and the encircled energy . At 780 nm the fraction of light in the halo is ~4%, while at 1000 nm the halo comprises ~38% of the light.
|
|
| Figure 2: Profiles (left) and encircled energy (right) of a long slit at several wavelenghts | |
In addition to the diffuse extended halo, the 1000 nm images also exhibits another artifact, which appears superposed on the halo in image at wavelengths longer than 1000 nm. This artifact has the appearance of a cross or diffraction spike with a strong horizontal spike and a weaker vertical spike. This artifact appears in non saturated images where the halo is limited and charge blooming does not cover the spikes . Shorter wavelengths images show a circular image, as it would be expected.
|
| Figure 3: Diffraction spikes in the 1000 nm point source image |
The long wavelength halos observed in STIS and ACS CCD images are a
direct result of the technique used to package the SITe CCDs. These
CCDs are thinned and backside illuminated with a typical device
thickness of 15 micorn. To provide a solid support, the CCD structure
is bonded to a soda glass substrate. Thinned CCD are relatively
transparent to longer wavelength photons. For example, a CCD thinned
to 16 micron will transmit ~5% of 800 nm photons and ~85% of 1000nm
photons. In case of SITe CCDs long-wavelength photons pass through the
thinned silicon and enter the soda glass substrate. Photons are
scattered in the substrate and then undergo a diffuse reflection at
the rear metalized surface, where some fraction will re-enter the CCD
from underneath. This schematic diagram illustrates this mechanism:
|
| Figure 4: Mechanism for long wavelength scattering halos in SITe CCDs |
Rasemberg (1997a,b) has modeled the system at Ball Aerospace, for the SAGE program, by considering the soda glass as a volume diffuser with uniform scattering properties, and the rear surface of the header as a white lambertian reflecting diffuser. The model results show that the substrate material is strongly forward scattering, and that the halo is produced by this forward scattered light reflecting from the rear surface of the substrate.
The ACS team carefully considered all these solutions: the first three possibilities would have required significant changes to CCD processing at SITe, effort incompatible with the ACS delivery schedule.
The option of preventing long wavelength light entering the header has been accomplished by placing a reflective layer between the CCD and the header. This thin metal layer (~200 A of aluminium) has been evaporated over the CCD's frontside gate structure prior to backside thinning. The thin Al layer is highly reflective at all the wavelengths and avoids, almost completely, the interaction of the red leaked light with the diffusing substrate, by reflecting it straight back into the CCD :
|
| Figure 5: Mechanism for removing the halos in ACS CCDs |
A number of test devices were fabricated at SITe using this additional metal layer on th CCD's frontside and then tested at Ball Aerospace. The addition of this barrier layer between the CCD and the glass header, has proved to be very effective in reducing the long wavelength halo. This is shown in the following figures where the averaged profiles of a long slit imaged with one of these "halo-fix" CCDs are presented along with the encircled energy.
|
|
| Figure 6: Profiles (left) and encircled energy(right) for a CCD with the halo-fix | |
Comparison tests show that the metal layer effectively removes long wavelength scattering in the 800nm and 900 nm profiles. The 1030 nm profile still exhibits some halo, although at a lower level than previously.
This schematic comparison shows the significative reduction of the fraction of light scattered in the halos
|
The presence of the metal layer also serves to increase the CCD's quantum efficiency at long wavelength, since the light reflected back into the CCD and detected, would otherwise have passed through the CCD. The relative improvement has been measured as ~8% at 800nm and ~65% at 1000nm (see figure). Another concern associated with metalizing the CCD's frontside is the effect it might have on the fringing at near-IR wavelengths. We have measured the fringing amplitudes in metalized and unmetalized devices and found no difference in the two CCDs (see figure).
|
|
The residual scattering shown in the 1030nm profile in figure 6 can be explained by the inclusion, in the profiles, of the diffraction patters structure seen in long wavelength images, although the extended diffuse halo has been suppressed. The diffraction pattern (see figure 3) has a strong horizontal spike and a weaker vertical spike. The diffraction-like structure appears to be due to an interaction between the long wavelength light and the gate structures on the CCD's frontside. These gate structures and pixel channel stops have typical scale size of a few microns. Profiles and encircled energies in the previous figures were obtained averaging rows in long slit images where the strong horizontal spike was predominant. From an analogue analysis of a point source image in a metalized CCD we obtained the encircled energy profiles along the directions of the two diffraction spikes and along two directions at 45 degrees from the vertical. clearly shows that the residual scattered halo, suggested from , corresponds, in reality, to the horizontal diffraction spike and that the halo-fix effectively removes long wavelength scattering at all wavelengths, as it can be seen from the profiles along diagonal directions. The CCDs will be oriented,in the HST FOV, to match the CCDs intrinsic diffractions spike with spikes produced by the secondary mirror holding supports.
|
|
|
|
| Encircled Energy at 1000 nm along the diffraction spikes | |
In order to analyze the scientific impact of the long wavelength halos on ACS science, we have generated simulated images by using point spread functions generated from the laboratory long slit images. Simulated images were generated to correspond to a 1000 second WFC observation. The aim of this simulation was to determine the limit magnitude with the spreading of the star flux in an extended halo and after the CCD metallization. The stellar photometry package DAOPHOT was used to measure the photometry of the simulated sources. shows the measured limiting magnitudes for a metalized "halo-fix" 1024×1024 CCD and a standard (un-metalized) ACS 1024×1024 CCD. The limiting magnitude were determined from a cutoff defined as the magnitudes where measured photometric errors exceed 0.2 magnitudes. The graph clearly shows that in the absence of a fix to the long wavelength scattering halos, the detection limits of the ACS WFC would be compromised by up to 1.7 magnitudes at 800nm.
|
| Comparison of limiting V magnitudes from simulated images |
