5 Discussion

Several optical schemes are possible for the pupil densifier. With respect to the diffractive mode of pupil densification tested by Pedretti et al. (2000), using a single micro-lens array, the present optical scheme adopted here, with two micro-lens arrays, is more flexible and gives a better control of sub-pupil filling. The densified pupil seen from the camera, although only 80% densified in this experiment, can in principle be more completely filled to obtain a rather uniform wavefront with very narrow gaps, of interest for coronagraphy.

Figure 5: Profile of the image of Castor A-B, showing the resolved companion. The Strehl ratio is degraded by seeing.

The alignment however is more difficult than with a single micro-lens array. The pair of micro-lens arrays indeed requires a careful rotational alignment of both arrays relative to the aperture grid within $1^\circ$. Misalignment causes a novel form of aberration: the main rays from sub-pupils define a series of co-axial single-sheet hyperboloids, instead of co-axial cylinders. This is evidenced as a rotation of the image when the camera is moved in and out of focus. Both micro-lens arrays had rotating mounts for this adjustment. The final image must appear (in white light) as a central white peak, surrounded by symmetrical secondary dispersed peaks if densification is incomplete.

We used converging micro-lenses for the first array although diverging ones in principle provide a wider ZOF since the image's diffractive envelope then moves in the same direction as the interference peak. Indeed, with diverging lenses, a point source located at field angle $\alpha$, gives a central interference peak which is off-set as $\alpha(1-1/\gamma_{\rm D})$ (Fig. 1). With converging lenses instead, the off-set goes as $\alpha(1+1/\gamma_{\rm D})$. The difference however becomes vanishingly small with increasing values of the densification factor. With the densification value adopted here, the converging lenses of the first array only cause a minor reduction of field size. Two points are indeed critical for direct imaging. First of all, it is not possible to image more than 4N sources (for our configuation) corresponding to the number of resolution elements contained in the ZOF. Second of all, any object outside the ZOF (28 $\hbox{$^{\prime\prime}$ }$ in diameter), but within the HOF (2 $\hbox{$^\prime$ }$42 $\hbox{$^{\prime\prime}$ }$ in diameter) appears in the ZOF owing to its dispersed higher-order peaks. This effect, called confusion noise, reduces the signal to noise ratio. If the sources density is weak, it becomes possible to reconstruct an image, using multi-spectra exposures (three wavelengths or more) in order to determine the position of objects outside the ZOF. This would increase the field of view of a densified pupil interferometer. The confusion noise induces errors in the determination of the source positions. With 312 resels, we can observe a density of 54 sources per squared arcmin.