The purpose of the investigation is the determination of accurate
(one per cent) temperatures for selected stars for use in a flux
calibration for ISO, as described by van der Bliek et al.
(1992) using the Infrared Flux Method. The method and its
applications have been discussed in detail by Blackwell et al.
(1980, 1990, 1991),
Blackwell & Lynas-Gray (1994), Saxner & Hammarbäck
(1985),
Bell & Gustafsson (1989), Smalley
(1993), Glushneva et al. (1993)
and Alonso et al. (1995). Because of known limitations
of the IRFM, the range of stars is restricted to those with temperatures
between 10000 K and 4200 K, and surface gravities between
and 1.0.
The method is based on the measurement of two stellar quantities:
the absolute monochromatic continuum flux density, 
, at a chosen
wavelength 
in the near infrared, and the absolute integrated
flux F.
The ratio of these two quantities is related to the stellar temperature
T through the equation
where 
gives the monochromatic flux from the star as a
function of effective temperature, the surface gravity, the wavelength
at which the monochromatic flux is determined, and the atomic abundances.
The method is applied by calculating the right-hand side of the equation
as a function of temperature for a chosen value of
on the Rayleigh-Jeans
tail, and for selected values of 
and A, using a range of model
stellar atmospheres. This procedure gives a calibration relating
temperatures to measured values of R. Later calibrations, for example those
of Blackwell & Lynas-Gray (1994)
have been made using more recent models
by Kurucz (1991, 1992).
Other calibrations have been published by
Alonso et al. (1996a)
for dwarf and subdwarf stars, and by Mégessier
(1994). The accuracy of flux calculations using model
atmospheres has been considered by van der Bliek et al.
(1996), whilst Mégessier (1994) has
assessed the influence of the model atmosphere on derivations of temperature.
The dependence of the IRFM on stellar metallicity has been
discussed by Mégessier (1994) and by
Smalley (1993). However, an accurate
correction for metallicity cannot always be made for the present program
stars because of a lack of precise measurements of [Fe/H]. In addition,
even the solar iron abundance is now uncertain to the extent of 0.12 dex
following the work of, for example,
Holweger et al. (1991, 1995) and
Blackwell et al. (1995).
Most of the basic [Fe/H] values adopted for the
stars in this paper have been taken from the survey of measurements by
Cayrel de Strobel et al. (1992),
supplemented by those given by
Edvardsson et al. (1993). Where there are no direct
measures, Schuster & Nissen (1989) and Carney
(1979) have suggested methods of
deriving [Fe/H] depending on stellar photometry, but the results are
probably uncertain to 0.6 dex. In these circumstances, we have assumed a
mean value of 
dex for stars for which there are no
direct measures of [Fe/H], a value which corresponds to the mean measured
value of [Fe/H] for all of the stars in this program. For the majority
of stars, these procedures should give results with a mean temperature
error of less than 0.5 per cent through uncertainty in metallicity.
The IRFM requires values of interstellar extinction so that corrections can be made to observed infrared and integrated fluxes. Blackwell & Lynas-Gray (1994) give a table showing the effect of interstellar extinction on the determination of temperature. In the present work, individual extinction values have been obtained using parallaxes measured by the Hipparcos satellite (ESA 1997). The Hipparcos parallaxes for the more distant stars have been of the utmost importance, for they have enabled the the IRFM to be greatly extended in its range of application. Following previous work we assume an average interstellar extinction of AV = 0.8 mag/kpc, with Allen's (1955) wavelength dependence, but also consider the results of the complex mapping of extinction made by Arenou et al. (1992).