1 Introduction

The chemical anomalies observed in several classes of late-type stars are due to mass transfer across a binary system. This scenario, first suggested by McClure (1983), holds for dwarf and giant barium stars (Pop. I G and K stars with overabundance of carbon and heavy elements like Sr and Ba produced by the s-process of nucleosynthesis), CH stars (the Pop. II analogs of Ba stars), and extrinsic S stars (late-type giants characterized by ZrO bands and no Tc lines, an element with no stable isotopes). The set of spectroscopic orbits available for those chemically-peculiar red stars (CPRS) has been considerably enlarged thanks to a decade-long radial-velocity monitoring. Such orbits are now available for 63 giant barium stars and 19 extrinsic S stars (Jorissen et al. 1998), 21 dwarf barium and subgiant CH stars (McClure 1997; North, priv. comm.) and 8 CH stars (McClure & Woodsworth 1990). Most of them were not available to the Hipparcos reduction consortia. The astrometric parameters (parallaxes and proper motions) published in the Hipparcos catalogue (ESA 1997) were therefore derived from a single star solution (the so-called "5-parameter model''). Since many of these CPRS have orbital periods in the range 1 - 5 yr, their unaccounted orbital motion may seriously confuse the parallax and proper motion as determined by the Hipparcos consortia. Thus, unrecognized orbital motions might introduce a so-called "cosmic error'' (of a systematic nature) in the proper motion (Wielen 1997; Wielen et al. 1999). Parallaxes may also possibly be in error, especially for binaries with orbital periods close to 1 yr.

The conclusions of previous studies inferring the kinematics and absolute magnitudes of CPRS (Bergeat & Knapik 1997; Mennessier et al. 1997; Van Eck et al. 1998) from the data of the Hipparcos catalogue may possibly be affected by these systematic errors. It is therefore of importance to re-evaluate the astrometric parameters of CPRS with a model correctly accounting for the orbital motion (the so-called "12-parameter model'', reducing to 9 parameters if the spectroscopic orbital parameters P, T and e are fixed), and to compare these values with those from the Hipparcos catalogue to identify the stars most affected by these systematic errors.

This comparison should allow to assess the reliability of the kinematic properties and absolute magnitudes of CPRS formerly derived from the Hipparcos-catalogue values. As a by-product, the astrometric orbit has been obtained in many cases, leading to an estimate of the masses through the ratio $M_2/[1+(M_1/M_2)]^2\equiv f(M) /\sin^3i$ (derived by combining the orbital inclination from the astrometric orbit with the mass function f(M) from the spectroscopic orbit when only one spectrum is observable).

From the operational point of view, the numerical re-processing of the raw Hipparcos data (Intermediate Astrometric Data or Transit Data; see Sect. 2.1) presented in this paper is innovative on the following points: (i) a global optimization technique is used to minimize the objective function in the 12-parameter (or, for spectroscopic binaries, 9-parameter) space, and (ii) a change of independent variable allows to derive always positive parallaxes, contrarily to the situation prevailing in the Hipparcos catalogue. The tool developed here for CPRS may in the future be applied to any other binary star to re-process the Hipparcos data when new orbits appear. The second paper of this series will be devoted to the re-processing of spectroscopic binaries involving a giant component (Boffin et al. 1993). Another is devoted to Procyon (Girard et al. 2000). We encourage interested readers to communicate to the authors other spectroscopic binaries whose Hipparcos data would need to be reprocessed.