1 Introduction

At its 23rd General Assembly the International Astronomical Union (IAU) decided (Appenzeller 1998) to introduce from January 1 1998 the new International Celestial Reference System, ICRS (Arias et al. 1995). In accordance with the IAU 21st General Assembly Resolution on Reference Systems (Bergeron 1992), the ICRS is realized by extragalactic radio sources forming the International Celestial Reference Frame, ICRF (Ma & Feissel 1997; Ma et al. 1998). ICRS and ICRF have thus replaced the previous reference system as realized by the FK5 fundamental catalogue.

The replacement of the FK5 system by the ICRS/ICRF has much deeper effects than a mere gain of precision. For the first time a reference system has been realized in the radio band rather than in the optical domain, which had been standard practice of astrometry for about a century. This decision was justified by the obvious advantages of using radio sources as reference objects, including higher random and systematic accuracy, long-time stability due to the lack of detectable proper motion, and independence of the equinox definition by means of the celestial equator and the ecliptic. In the optical domain, however, where the majority of all astronomical observations is still being performed, no comparable reference system realization existed.

The success of the European Space Agency (ESA) Hipparcos astrometric mission removed this deficiency. The Hipparcos mission yielded a catalogue of astrometric parameters of $118\,000$ stars (ESA 1997), derived with milliarcsecond (mas) precision, and indirectly linked to ICRS with sub-milliarcsecond accuracy (Kovalevsky et al. 1997). Consequently, the IAU recommended the Hipparcos Catalogue as the primary realization of the ICRS in the optical domain.

However, for many applications, the ICRS realization by Hipparcos suffers from the excessive brightness of the reference stars (the median magnitude of the Hipparcos stars is close to V=8.5 mag) and their low number density. Thus the primary reference system realization is practically inaccessible to many modern astronomical observations. This aggravates the problem of extending the reference system to faint stars, well-known to astrometry for decades. In the ICRS/Hipparcos context, that problem may be solved by the construction of a reference catalogue of many more fainter stars, which is directly linked to the Hipparcos system.

The most promising candidate for the purpose is the catalogue of more than $1\,000\,000$ stars which resulted from the Tycho experiment on board the Hipparcos satellite (ESA 1997). The uniformly high systematic and random accuracy of the Tycho positions, its native link to the Hipparcos system, its relatively high star density and the availability of homogeneous two-colour photometry make the Tycho Catalogue an ideal means for extending the ICRS/Hipparcos frame to fainter stars.

The only drawback of Tycho as an astrometric reference catalogue is the poor quality of the proper motions. The proper motions of Tycho stars derived within the experiment have low precision, typically 30-40 mas per year (mas/yr), which will deteriorate the reference frame within a decade.

The derivation of high-precision proper motions of Tycho stars as a way to construct a denser reference catalogue constituted the aim of the Tycho Reference Catalogue project, as proposed by the present authors in 1992 (Röser & Høg 1993). The high-quality early-epoch observations necessary to derive proper motions with an accuracy of a few mas/yr, were to be provided by the Astrographic Catalogue (AC), a collection of twenty catalogues of position measurements at mean epoch near B1907.0, containing observations of circa 4.5 million stars brighter than $m_\mathrm{pg}=12$ mag. Due to the early epoch and high intrinsic accuracy (0$.\!\!^{\prime\prime}$2-0$.\!\!^{\prime\prime}$4), AC constitutes an excellent first epoch for the massive derivation of proper motions.

The TRC pilot project (Kuzmin et al. 1997b), performed in 1996-97 using the early access to Hipparcos and Tycho data, confirmed that the AC makes it possible to derive proper motions with 2-3 mas/yr precision for nearly all Tycho stars. The pilot project also demonstrated the necessity to put AC onto the Hipparcos system and the crucial importance of this procedure for the quality of the resulting proper motions.

The publication of the Hipparcos and Tycho catalogues (ESA 1997) allowed full-scale realization of the TRC project, resulting in the recently published Tycho Reference Catalogue which provides highly-accurate positions at epoch J2000.0 and proper motions of 990182 stars of the Tycho Catalogue. The TRC has been made available through astronomical data centres; the catalogue characteristics and derivation procedure were summarized in a recent paper (Høg et al. 1998a). In the present paper we describe various aspects of the TRC construction in detail, aiming to provide the catalogue user with an in-depth knowledge of the observations and procedures used to derive the Tycho Reference Catalogue.