1. Introduction

The reduction of the observations carried out by the Hipparcos satellite came recently to an end and the availability of the final results to the scientific community is due in early 1997. The merging of the two solutions produced independently by the Consortiums FAST and NDAC (Kovalevsky et al. 1992, 1995; Lindegren et al. 1992) provides a single set of astrometric and photometric estimates for each program star, including the analysis of their variability and, of more relevance in the present context, an astrometric solution of the double and multiple stars. These stars were either known as double before the mission, thanks to an intense data compilation of existing information (Dommanget 1985, 1989) or discovered during the analysis of the Hipparcos observations.

It is currently believed that half of all stars are binary, sharing a common proper motion, or put differently that more stars are to be found in association than in isolation. This is a very loose concept since within this idea we cannot rule out that the Sun belongs to such a pair. If we restrict to pairs effectively seen in gravitational interaction from the existence of a non rectilinear absolute motion, the number of true binary objects decreases considerably, may be as most orbital periods are so long that the curvilinear motion has gone unnoticed so far. It is customary to divide the binary stars into broad classes defined by the observational means by which they were recognized, such as visual, eclipsing, spectroscopic or astrometric binaries, rather than by any more physical criterion pertaining to the structure of the couple.

In the case of Hipparcos there were basically two ways to recognize that a star has a nearby companion:

First when the photometric signal of a program star measured on the grid departed significantly from the well calibrated single star signal. Thus statistical tests could be constructed to determine at an early level of the reduction process whether a particular object had to be removed from the single stars basket. This recognition made no use of the absolute motion of the components or, of the photocentre, against the stellar background. It relied only on the recognition that the entry which was being observed was either an extended source or composed of several star-like sources. In this respect this procedure of detection was very akin to the telescopic detection of a visual binary by simple inspection, although in the Hipparcos case this was done by an image processing rather than by the eye. In the following these stars will be referred to as resolved double stars. About such systems have been solved from the Hipparcos observations and an additional 6000 are flagged as suspected non singles.

Alternatively, when the image could not be distinguished from that of a point like source, even though the source might have been a close pair, the combined signal was analyzed in the same way as that of a single star, with the basic assumption that the absolute motion was rectilinear, (see Wielen 1996, for exceptions to this rule). However this assumption failed when the object was a binary star with a period comparable or less than few times the mission length, in which case the "photocentre" had a wavy motion on the sky superimposed to the linear motion of the centre of mass. The detection in this instance was of the same nature as that of astrometric binaries on photographic plates, with the very important reservation that the Hipparcos image of an unresolved object should not be systematically identified with the photocentre.

In this paper we shall be concerned only with this second category of double stars. Our main goal is to characterize the Hipparcos astrometric binaries and to show that from the observed absolute motion on the sky it is possible to derive relevant physical parameters, like the mass ratio and the intensity ratio of the components and not simply their difference as usually done with the astrometric binaries. In fact, the possibility of investigating astrometric binaries with Hipparcos has been considered, even before launch, for example by Morel (1988) or Dommanget & Lampens (1993), who wrote down the basic equations. The actual processing was rather different from what they had imagined, since there is no absolute observation of the motion of the primary component, but a complex signal produced by the two members of the pairs, which can be disentangled and interpreted in term of orbital motion.

To avoid the confusion between the classical astrometric binaries, based on the observation of the photocentre, and the Hipparcos astrometric binaries we coined the word hippacentre to name the physical point attached to the Hipparcos observation of an extended source. The name conveys the idea that it is very specific to Hipparcos, and that at the same time it retains some analogy with the photocentre. It is precisely defined in Sect. 2, while its induced orbital motion is investigated in Sect. 3. Finally its particular relation to the masses and luminosity of the components allows to determine simultaneously the mass ratio and the intensity ratio, provided the separation between the components is larger than about 0 3, as explained in Sect. 4.

A companion paper in preparation will consider the applications of the principles presented here and evaluated on simulated data, to a set of about 100 astrometric binaries observed during the Hipparcos program.