1 Introduction

The search for extra-solar planets is one of the most exciting fields in present-day astrophysics. We are living at an epoch where progress in instrumentation is beginning to allow the detection of other planets around nearby stars.

The first step was reached when Wolszczan and Frail announced the discovery of a planetary system around a pulsar (PSR 1257+12) (Wolsczczan 1992) but a real advance occurred when Mayor and Queloz announced the discovery of a planet around the solar-type star 51Peg (Mayor & Queloz 1995). At present, more than 50 other planets have been discovered (Schneider 2000), mainly using a radial velocity technique. Even though this method has shown its efficiency (determination of the period, semi-major axis, lower limit of the planet mass, multiplicity of the system), it is an indirect method of detection, and cannot give spectral information about the planet itself.

Direct methods of detection such as imaging or spectroscopy should in the near future make it possible to get information about the planets' composition. Several space missions are presently under study to reach that goal: DARWIN in Europe (Léger et al. 1993; Léger et al. 1996; Fridlund et al. 2000), and TPF in the United States (Woolf et al. 1998). The instrumental concept for such missions should allow high angular resolution as well as high dynamic range observations. For example the observation of an Earth-like planet orbiting around a solar type star located at 10 pc from the sun subtends an angle of 0.1 arcsec as well as dynamic range $\Delta m$ around 17 in the N band.

One possible instrument that fulfills these requirements is the Bracewell interferometric coronograph (Bracewell 1978), or instruments based on similar concepts (Angel et al. 1986; Léger et al. 1993). The specificity of all these instruments is to cancel the flux in the direction they are pointed in (stellar direction) while the flux from an off-axis direction is transmitted. Thus, they are well suited for the detection of faint star companions.

In order to be efficient, the on-axis extinction should be comparable to the contrast between the star and its companion (about 7 10⁶ at 10 $\mu$m). The extinction is quantified by the rejection rate, written $\rho$ and defined as the ratio of the constructive to the destructive interference flux. An ideal interferometer has thus an infinite rejection rate for a point source at infinity, because the on-axis extinction is total. Due to optical ripples and coating inhomogeneities, the extinction of a real interferometer is never perfect and thus it has a limited rejection rate (the dark fringe is never completely dark).

Several experiments to demonstrate the validity of this concept are under way (Serabyn 1999; Hinz et al. 2000; Morgan et al. 2000) but at present, no experimental demonstration of high rejection nulling interferometry ( $\rho > 10^{4}$) has been proposed in the thermal infrared. We already have quantified the requirements for a 10⁶ rejection rate nulling interferometer (Ollivier & Mariotti 1997; Ollivier 1999). In this paper, we describe an experimental setup that should reach that level of extinction in a monochromatic case at $\lambda \sim$ 10 $\mu$m. In the first section, we summarize the requirements needed to reach such a high rejection rate. In the second section, we describe the experimental interferometer we have built, the principles behind it and the fabrication and performance of its individual components. In the last section, we present the first results obtained with it.