|IRAC Handbook PSF||STinyTim model PSF|
The images above show simulated IRAC observations of the HDF-North that were generated in 2000, during the period when the GOODS Spitzer (then, SIRTF) Legacy proposal was being written. They were designed to simulate 100 hours of exposure time in each of the four bandpasses, 3.6, 4.5, 5.8 and 8.0 microns. The RGB color composites were generated using weighted combinations of the four bands. The two simulations use different assumptions about the IRAC point spread function: at left, PSFs meeting the minimum requirements assumed in the SIRTF Observers Manual (edition circa 2000), and at right, a "best case" PSF assuming ideal SIRTF/IRAC point spread functions modeled with the STinyTim software (courtesy J. Krist). The sensitivity limits adopted for the GOODS proposal were based on the "handbook" PSF image at left, as required by the Call for Proposals for Legacy Science programs. The simulations assumed extensive dithering which can be used to recover information lost by the PSF undersampling due to the 1.2 arcsec IRAC pixel scale, e.g., via the "drizzling" technique (the simulation realistically accounts for the effects of undersampling and drizzling on the final image quality). The images were simulated based on NICMOS 1.6 micron observations of the HDF (Dickinson et al. 2000), using photometric redshifts and spectral templates fit to 7-band WFPC2 + NICMOS + K-band photometry and extrapolated to IRAC wavelengths in order to predict object fluxes. Galaxies at z > 2.5 are recognizable by their red IRAC colors, caused by the passage of the rest-frame 1.6 micron spectral inflection through the filter system.
In practice, the in-flight image quality achieved by Spitzer and IRAC is excellent, closer to the STinyTim model above, and substantially better than the pre-launch requirements. This improvement is most noticeable for IRAC channels 1 and 2. The pre-launch specification was for the telescope to be diffraction limited at 6 microns, and therefore there is relatively little improvement in the image quality for channels 3 and 4. The figures below compare the STinyTim PSF simulation shown above with the actual, in-flight GOODS image of the HDF-N. Note that several factors limit the "apples-to-apples" nature of this comparison: (1) The simulation is for 100 hours of IRAC exposure (similar to our ultimate depth with the "ultradeep" GOODS program), whereas the color image shown here is based on the 23 hour exposure time "superdeep" images; (2) The exact recipe used to make the color composite images of the simulations has been forgotten, and therefore the real images were generated differently, with different "stretch" and channel-to-channel balance. The actual image is a composite of the 3.6, 4.5 and 8.0 micron channels (blue, green and red), with the images scaled to uniform intensity in micro-Janskies, such that a flat-spectrum (f(nu)) object would appear to have neutral (white) colors here.
Another fact of note: the proposal simulations did not include the effects of PAH or warm dust thermal emission in the model galaxies. In the real data, objects with apparently red colors in the picture below, which represent enhanced 8 micron emission, are generally low redshift galaxies, with z < 0.5, where the strong 7.7 micron PAH bands are in the 8 micron IRAC channel 4, rather than high redshift galaxies, as was suggested above for the simulations. This is why some relatively bright galaxies appear to have red "halos" around them (the four IRAC channels have not been PSF-matched for this color composite, and therefore the 8 micron PSF is significantly broader than that of the shorter wavelength channels, resulting in this halo effect). Some high-redshift galaxies also appear red, however, for the reasons described above.
|The Real Deal (23 hour exposure)||Pre-launch simulation (100 hour exposure, STinyTim PSF)|