1 Introduction

The Hipparcos Catalogue (ESA 1997) includes two sets of Cousins V-Icolor indices - a functional V-I (entry H75 in the main Hipparcos Catalogue) and a best available V-I at the time of the Catalogue's release (entry H40). This color index is an important temperature indicator for late-type stars (Dumm & Schild 1998; Bessell et al. 1998). Since only 2989 Hipparcos stars are listed as having direct measurements of the Cousins V-I index, nineteen different methods of variable accuracy were used to obtain V-I photometry (see ESA 1997, Sect. 1.3, Appendix 5). In numerous cases the reductions of Hipparcos V-I photometry relied heavily upon the satellite's star mapper photometry - the Tycho $B_{\rm T}-V_{\rm T}$ color indices. However, the Tycho photometric system alone is not well-suited for the studies of fainter red stars. A combination of intrinsically low fluxes from these stars in the $B_{\rm T}$ bandpass and a short crossing time ($\sim$22 ms) of the star mapper's four vertical slits resulted in low S/N ratios. This, in combination with the residual bias that was not fully corrected by the de-censoring analysis (Halbwachs et al. 1997) in deriving the Tycho photometry for faint stars, diminishes the reliability of much of the published Hipparcos V-I indices for stars with $V-I\ga1.5$. As demonstrated by Koen et al. (2002), the listed Hipparcos V-I photometry of red stars shows a disappointingly large scatter with respect to the ground-based photoelectric V-Imeasurements. In extreme cases the disagreement can reach up to 2-3 mag.

Our interest in the V-I photometry of red stars is primarily motivated by the potential effect of incorrect V-I color indices on the chromaticity corrections in Hipparcos astrometry. On average, a one magnitude offset in the V-I value could introduce a $\sim$1 mas bias in the star's position (abscissa) along the scan direction. Besides grossly incorrect V-Iindices for some red stars (Koen et al. 2002), there is a systematic color bias related to neglecting in the Hipparcos reductions the intrinsic color variation in large amplitude variables such as Miras.

In retrospect, the Hipparcos V-I photometry would have gained considerably from the parallel-in-time ground-based V-I observations of stars with extreme colors and/or considerable color variability. For a number of reasons, most importantly, a prorogated decision to choose the V-I index, this opportunity was lost. Is it possible to improve the Hipparcos V-I photometry now? Here we attempt to answer this question. It appears that high-grade V-I photometry for red stars is possible down to $V\approx8$ and may even be used to obtain an estimate of effective temperatures. In general, the re-calibrated V-I photometry is useful in identifying some difficult cases in the Hipparcos Catalogue, such as red and variable stars in binary systems. Throughout the paper we refer to Cousins V-I color indices, unless it is explicitly stated otherwise.