The parallax data gathered by Hipparcos
for field subdwarfs allow a muchmore precise
determination of globular cluster distances by
direct sequencefitting than was previously
possible. We determined the distance and age ofthe
old, representative globular cluster M92 from a
set of more than 500 subdwarf candidates with
Hipparcos parallaxes. Precise [Fe/H] values were
derived using the equivalent width of the CORAVEL
cross-correlation function.
Our best estimate of the distance of M92
is (m —
M)v = 14.67 ± 0.08 (including
binaries with a statistical correction) or
(m —
M)v = 14.74 ± 0.08 (classic
treatment, i.e. excluding binaries). The agreement
of the cluster sequence with the position of
extreme metal-poor field subdwarfs is remarkable
[figure, left]. The distance found is slightly
higher than previously thought. The corresponding
ages, derived by comparing the luminosity of the
turnoff and subgiant-branchstars with up-to-date
evolution models [figure, right], are 14±1.2 Gyr
or 13±1.2 Gyr respectively, implying a minimum age
of 13 Gyr for the Universe.
Other authors have claimed that larger
distances, and smaller ages, resulted from
Hipparcos subdwarf parallaxes. Although there are
some differences in the data sets used by each
author, the main difference resides in the
treatment of systematic and selection biases. We
have examined at some length the biases affecting
the determination of the mean luminosity of a set
of subdwarfs selected in the usual way. By means
of Monte Carlo simulations, we show that selection
biases act in a direction opposite to the classic
Lutz-Kelker bias affecting parallax data, and that
they can be dominant. The biases introduced are of
the order of 0.1 mag.
It is pointed out that, now that the whole
Hipparcos catalogue is available, we shall be able
to greatly reduce the systematic biases by
refining the fitting procedure. Firstly, by
imposing no [Fe/H] limits and fitting the subdwarf
locus for all [Fe/H] values at the same time.
Secondly, by imposing no parallax error limit and
fitting the data in parallax space instead of
magnitude space (comparing observed parallaxes
directly with parallaxes predicted from the
models).