1 Introduction

Even after the conclusion of the controversy on the 51 Pegasi planet (Gray 1997; Gray 1998; Mayor & Queloz 1995), the lack of real images remains a major difficulty in interpreting indirect detections of exoplanets. Direct imaging is classically hampered by the huge intensity ratio between a solar-type star and its planets. It requires spatial resolution beyond the atmopsheric limit, together with efficient nulling of the star light (Angel 1994). High-performance adaptive optics are being developed to improve the instrument performance (Angel 1994; Gezari et al. 1997b). New approaches to stellar coronagraphy have also been proposed recently: the phase mask system (Roddier & Roddier 1997) and the achromatic interferometric coronagraph AIC (Gay & Rabbia 1996). Both are expected to achieve a higher sensitivity than the Lyot coronagraph, and should allow the search for the companion closer to the central star (a fraction of the Airy disk) than the Lyot system (a few Airy disks).

Our instrument is a visible Lyot coronagraph which uses the dark-speckle detection (Labeyrie 1995) for exoplanet imaging. It stops the star's Airy peak and rings restored by an adaptive optics system. The method exploits the occurrence of dark speckles (destructive interferences in the focal plane) to clean the halo of scattered light. The statistical analysis of this occurrence shows positions where faint planet images prevent full darkening of the star's speckles. The depletion of zero-photon count in these regions thus indicates the presence of faint companions. By processing a large number of short exposures, faint companions can emerge above the residual noise.

A refined theoretical analysis of dark-speckle imaging is given in Boccaletti et al. (1998), together with numerical and laboratory simulations. Simulations have shown, for the case of the Hubble Space Telescope, that imaging an exoplanet with 10^-9 intensity ratio is possible with a dark-speckle coronagraph in a few hours (Boccaletti 1998).

The first observation, in june 1996, at the Observatoire de Haute-Provence of the binary star HD 144217 ($\Delta m$ = 4.8) has confirmed the simulations (Boccaletti et al. 1998).

Here, we describe recent improvements of our instrument (Sect. 2) and some results obtained during a ten-night observing run in october 1997 at the Observatoire de Haute-Provence (Sect. 3). To test the dark-speckle method and gain experience in high sensitivity coronagraphic imaging, we observed Hipparcos binaries having large magnitude differences. These instrumental developments also prepare the next generation of dark-speckle coronagraphs, to be used on large ground-based or space telescopes (Gezari et al. 1997b) and interferometric arrays (Labeyrie 1998).