3 Experimental arrangement

Following the initial sky testing achieved by Pedretti et al. (2000), using a diffractive mode of pupil densification, we have tried to improve the densification technique. Among a wide range of possibilities, we have chosen to use a pair of micro-lens arrays. The hypertelescope's scheme is represented in Fig. 2. Instead of a true array of mirror elements, we have used a 40 cm Newtonian telescope at Observatoire de Haute Provence as collecting optics. In order to have nearly diffraction-limited image quality without adaptive optics, the hypertelescope exploits only 10 cm of the aperture. The rest of the aperture served for auto-guiding with a SBIG ST4 CCD camera.

Figure 2: Hypertelescope experimental setup. The incoming light beam from a Newtonian telescope is collimated by lens L1. A Fizeau mask installed for convenience in the pupil plane following L1, rather than at the primary mirror, has N=78 holes of $100~\mu$m size each. It defines in the entrance aperture a virtual "diluted giant mirror'' of 10 cm size with s=1 mm sub-apertures. The densification is achieved with two micro-lens arrays (ML$_{\rm 1}$ and ML$_{\rm 2}$).

The lens L1 produces a pupil image, 10 times smaller than the entrance aperture, which is masked by a grid of 78 holes of 0.1 mm size, centered 1 mm apart. The virtual grid thus defined in the entrance aperture has 1 mm holes spaced 10 mm apart. Two arrays of convergent and confocal micro-lenses (ML1and ML2), having a short and a long focal length respectively (20 mm and 120 mm), achieve the pupil densification. The front focal plane of the first array is located close to the grid, so as to provide a pupil plane close to the second array. Collimated beams from each sub-pupil become recollimated and widened when transmitted through the facing pair of micro-lenses. The densification factor, ratio of ML2 and ML1's focal lengths, amounts to 6, providing 80% filling (diameter) in the exit pupil. The micro-lens arrays utilized were fabricated by one of us (DH) at Observatoire de Paris (Bensammar et al. 2000), with enough lens-to-lens uniformity of thickness to keep piston errors within Rayleigh's tolerance, as required for a highly constructive interference, providing a high Stehl ratio, in the star's "high-resolution'' image. The rather faint star images thus formed with less than 1 cm² of total collecting area were recorded on a CCD camera with 9 $\mu$m pixels (0.62 pixel/ $\hbox{$^{\prime\prime}$ }$image sampling). With its aperture size of 10 cm and equivalent mirrors of 1 mm diameter, the ZOF extent is $1.22(\lambda /d)/\gamma _{{\rm D}}=28 \hbox{$^{\prime\prime}$ }$ at $\lambda=675$ nm and the angular resolution is $1.22\lambda /B=1.6\hbox{$^{\prime\prime}$ }$.

Figure 3: a) Image of Castor, showing the resolved binary A-B, spaced 3.8 arcsec. The half ZOF is about $14 \pm 0.6$ arcsec wide. b) Image of Pollux, obtained with a 10 min exposure. It matches the theoretical pattern, with the residual side-peaks due to incomplete pupil densification. With respect to the laboratory images and the numerical simulation, the peaks are however somewhat widened by seeing and exceed the theoretical arc-second resolution limit of the 10 cm array. c) Numerical simulation of a point source's monochromatic image with the 78-aperture hypertelescope.