For those unfamilar with PUEO, it is based on a 19 subaperture curvature wavefront system (Roddier et al. 1991). A complete description of the system is given by Arsenault et al. (1994), and a comprehensive discussion of its performance by Rigaut et al. (1997). If the target is to be observed in the infrared, a dichroic is used to reflect the visible light to the wavefront sensor (WFS) while transmitting the IR light to the science detector, but if the target is to be observed in the visible, a beamsplitter must be used to send a percentage of the light to the WFS. The delivered image quality depends on the brightness of the reference star, its distance from the target, the seeing at the time of observation and the wavelength of observation. To gain an appreciation of how these variables affect the images and what can be expected, the reader is encouraged to visit the "performance meter" at http://www.cfht.hawaii.edu/Instruments/Imaging/AOB/psf.html@. To achieve FWHM ~ 0.12" imaging in the near-IR (diffraction-limited) and at J and I under median seeing conditions, the reference star must be brighter than R~14 and within ~30" of the target. However, the system will provide good correction on stars as faint as R = 17 under good seeing conditions, and the reference star can be located anywhere within the 90" diameter field. AOB also offers considerable advantages in terms of operational efficiency, since it is literally a push-button operation that eliminates all the overhead of focussing and guiding associated with conventional observing.
Since WFPC-II on HST delivers excellent images in the visible spectral region, most of the observations with AOB to date have been carried out in the near-IR where there is less competition and also where it produces the best images. Hence, the observations reported here were mostly made with the University of Montreal infrared camera MONICA (Nadeau et al. 1994) which was modified to give a pixel scale of 0.034" and hence a field of 8.8"x 8.8". This very small field is a significant handicap for many observations, as is the fact that the detector suffers from persistence problems, i.e., a bright source leaves a residual image that slowly decays with time. Since the site seeing is variable, observations of relatively bright stars usually have to be carried out in order to monitor the PSF (point-spread function). The sequence of observations of very faint sources thus has to be carefully planned to minimize such problems while frequently observing the PSF star to attain the scientific goals. Veran et al. (1997) have recently demonstrated that statistics of the wavefront sensor signals can be used to derive an excellent model of the average PSF (as long as the guide star is m = 13 or brighter). This is obviously ideal in that the PSF estimate is simultaneous with the target observation, and it obviates the requirement of directly observing the bright star (saving time and avoiding persistence problems). A comparison demonstrating how well the model reproduces the observed PSF is reproduced in Figure 18. For each observation, statistical data on conditions at the time of observation are recorded in "psf" files and the atmospheric-corrected PSF is stored in "dph" files. The PSF corresponding to each exposure can then be calculated with the CFHT "DPH2PSF" package (see http://www.cfht.hawaii.edu/~veran/cfht_dph2psf.html@)
Examples of Self-referencing Targets
Objects such as the nuclei of nearby galaxies, Seyferts and AGN can be usually guided on directly, i.e., the nucleus is often sufficiently point-like to enable the WFS to function properly as long as there is sufficient flux. The cores of NGC 6946 and M31 are examples of bright spirals which were successfully observed. It is well-known that the nucleus of M31 is double, with the brightest component, P1, being offset from the fainter component (and dynamical centre) by ~0.5". In spite of the duplicity, guiding on the nucleus of M31 with AOB was straightforward, resulting in diffraction-limited images (FWHM = 0.12") of the stars in the nuclear bulge region (Figure 19) The near-IR image quality is comparable to those of WFPC, enabling accurate magnitudes and colors to be determined for a significant number of stars in the bulge (Davidge et al. 1997). Spectroscopy of the double nucleus of M31 with AOB + OASIS in the visible, or in the J and H bands with OSIS linked to AOB with an integral field unit (Felenbok & Crampton, in progress) should finally unravel the mystery of this double nucleus (see e.g., Bacon et al. 1994; Kormendy & Richstone 1995; Tremaine 1995).
AGN are obviously also easy to guide on if they are sufficiently bright. Several AGN and Seyfert 1 type galaxies with point-like nuclei and magnitudes in the range m = 11-14 (e.g., bright ones like NGC 4151 to fainter ones like NGC 6814) were observed during commissioning. However, guiding was not successful on more diffuse objects such as the V = 13 nucleus of the Seyfert 2 galaxy Mrk 266, or the large bright elliptical galaxy which hosts 3C296. Numerous "AGN" projects are currently underway and a few results have already been published or are in press, e.g., NGC 7469 (Lai, 1997), Mrk 273 (Knapen et al. 1997) and NGC 1068 (Rouan et al. 1997)
Targets with guide stars
Nearby guide stars must be used for targets which are not sufficiently concentrated or bright enough for the WFS to perform adequately. In this case, not only are observations of the guide star itself required to monitor the PSF variations, but measurements of starfields (e.g., globular cluster fields) are also required to estimate the degradation of the PSF due to isoplanatic effects. Steinbring (1997) has developed semi-empirical software that models this degradation over the field as a function of the seeing and produces an "off-axis PSF" that can be used to help analyse the target data.
Registering invisible targets
Targets which are very faint, particularly those that are diffuse, often present additional problems since the MONICA field is so small that there aren't any objects or features that can be used to register the images. Near-IR images always have to be dithered to remove the effects of bad pixels and to improve the flat-field, offsets are often required to enlarge the area surveyed, and differential flexure between the WFS and the detector may produce additional shifts if the total exposure is long. Consequently, registering and superimposing the images is not trivial. Fortunately, this was forseen in the initial design of AOB and every effort was made to minimize the flexure between the WFS and the detector so that the WFS coordinates (recorded in the FITS header) can be used to determine the offsets between exposures. Observations of sequences of overlapping fields in globular clusters can be used to calibrate WFS offsets in terms of pixel location on the detector and/or equatorial coordinates. Since MONICA and the "AOB optics" have to be mounted/dismounted before/after each run these calibrations should be checked, but so far they appear to have been very repeatable. Through comparison of AOB + MONICA coordinates with positions of stars derived from HST images, the pixel scale is ~0.0342" per pixel at H and the image is rotated such that N is at ~ -10.6º from the top of the image and E is at -100.6º (this rotation varies slightly from run to run). Comparison of the WFS X and Y offsets with offsets of the frames derived from coordinates of stars in globular cluster fields gives the following calibration:
Further examples of "invisible targets" are observations of high redshift galaxies. A pair of CFRS galaxies near a R = 15.3 guide star were observed during commissioning, and one of the z = 3.0 Steidel et al. (1996) galaxies was also attempted. In both cases, the goal was to examine the morphology of high redshift galaxies at more normal restframe wavelengths than is possible in the visible (H band corresponds to restframe I and B at z = 1 and 3 respectively). Observations of these very faint targets were significantly hampered by the detector remanence effect discussed above and also by the reduced sensitivity (three times lower than anticipated) of MONICA during the latter half of 1996. For example, it was initially estimated that each z = 3 galaxy (R ~ 24) would require a total exposure of ~8h, but when it became obvious that the object was barely visible after a few hours (because of the reduced sensitivity), the project had to be abandoned.
Many extragalactic projects will benefit enormously from the improved spatial resolution that is now possible from the ground, provided that the targets are bright and sufficiently concentrated or can be selected near suitable guide stars. Since a large number of extragalactic projects don't involve point sources that can benefit from the ultimate resolution that an instrument like AOB can deliver and/or probably don't have a sufficiently bright reference source to achieve that resolution, critically sampling the diffraction core is less important than a reasonable field size (so that sky and PSF stars are included in the field). Hence a pixel scale of the detector for AOB of ~0.07" rather than 0.034" per pixel would be advantageous for most such projects. In addition to imaging, the forthcoming combination of "adaptive optic image quality" and spectrographs designed to capitalize on it will be very exciting. I, for one, was continually frustrated during commissioning by not being able to get spectra of, e.g., the 0.1" images of the nucleus of M31 to find out what was really going on.
Given a suitable guide star, AOB delivered near-IR images worse than ~0.1" on only 1 of the ~30 AOB nights that I've been involved with so far. Figure 22 shows typical results, from ~350 observations taken on the first 10 commissioning nights (Rigaut, Lai & Veran, 1996). It feels like we're really entering a new era: from now on ground-based astronomers will complain when the images are significantly worse than 0.1" rather than 1"!
Acknowledgment: While the above discussion is the result of my personal experience with AOB, the fantastic performance of Pueo is due to the combined efforts of teams from CFHT, Meudon, Laserdot, UH and DAO. Francois Rigaut deserves much of the credit for this superb implementation of Francois Roddier's curvature concept. It is a very exciting instrument to use.
Arsenault, R., Salmon, D.A., Kerr, J., Rigaut, F., Crampton, D., & Grundmann, W. 1994, SPIE, 2201, 883