1 Introduction

Optical long baseline interferometry is one of the upcoming techniques that will undoubtedly provide compelling, high angular resolution observations in optical astronomy. The first attempt to use interferometry in astronomy was proposed by [Fizeau (1868)] and achieved by [Stefan (1874)] on a single telescope with a pupil mask. [Michelson & Pease (1921)] first succeeded in measuring stellar diameters, but their interferometer was not sensitive enough to enlarge their investigation. Interferometry is a rather complex technique which needs extreme accuracies directly proportional to the foreseen spatial resolution: 1 milliarcsecond on the sky translates to 0.5 $\mu$m in optical delay on a 100-m baseline. That is why modern direct interferometry started only in the 70's with [Labeyrie (1975)] who produced stellar interference with 2 separated telescopes. Also interferometric experiments require very low noise detectors which became available only recently. In addition, the atmosphere makes the work even more difficult and dramatically limits the sensitivity of ground-based interferometers. Space-based interferometric missions are therefore being prepared, like the NASA Space Interferometric Mission (SIM) or the interferometry corner stone in the ESA Horizon 2000+ program: Infrared Spatial Interferometer, (IRSI) .

Long baseline interferometry is based on the combination of several stellar beams collected from different apertures and is aimed to either aperture synthesis imaging [Roddier & Léna (1984)] or astrometry [Shao & Staelin (1977)]. A number of interferometers are currently working with only two apertures: SUSI [Davis et al. (1994)], GI2T [Mourard et al. (1994)], IOTA [Carleton et al. (1994)], PTI [Colavita et al. (1994)]. COAST [Baldwin et al. (1996)] and NPOI [Benson et al. (1997)] have started to perform optical aperture synthesis with three apertures by using phase closure techniques. The increase in the number of apertures is one of the major feature of new generation interferometers, like CHARA with up to 7 apertures [McAlister et al. (1994)] or NPOI with 5 siderostats [White et al. (1994)]. We are on the verge of new breakthroughs with the construction of giant interferometers like the VLTI (Very Large Telescope Interferometer) by the European community which will use four 8-m unit telescopes and three 1.8-m auxiliary telescopes [Mariotti (1998)], or the Keck Interferometer [Colavita et al. (1998)] which will have two 10-m telescopes and four 1.5-m outriggers. They will both achieve high sensitivity thanks to their large apertures and allow the combination of more than three input beams.

We propose in this article a new technology for beam combination that is inherited from the telecom field and micro-sensor applications. This technology will answer many issues related to astronomical interferometry. The technology is called integrated optics on planar substrate, or, in short, integrated optics. The principle is similar to that of fiber optics since the light propagates in optical waveguides, except that the latter propagates inside a planar substrate. In many aspects, integrated optics can be considered like the analog of integrated circuits in electronics.

We describe in Sect. 2 the optical functions required by an interferometer. We present in Sect. 3 the principle of integrated optics, the technology and the available optical functions. Section 4 presents the concept for an interferometric instrument made in integrated optics, and touches upon future possibilities. Section 5 discusses the different technical and scientific issues of this new way of doing interferometry. Results with a first component coming from micro-sensor application will be presented in Paper II [Berger et al. (1999)]. They demonstrate the validity and feasability of the integrated optics technology for astronomical interferometry.