Analyses of photospheric spectra of stars are vital tools for our
 understanding of issues from the origin of the Universe, the
 evolution of the Galaxy to the lives of stars. Contemporary
 analyses of spectra depend on model stellar atmospheres computed
 usually from a set of simplifying assumptions: plane-parallel
 homogeneous layers in radiative/convective hydrostatic and local
 themodynamic equilibrium. Only few attempts have been made to
 derive atmospheric structures as direct as possible from
 observations, i.e. to construct empirical atmospheres. With new
 observational methods, and new wavelength regions opening up, it
 is very important to try such methods. In this respect the infrared
 spectral region offers new and unique possibilities.

 This proposal considers such atmospheres of cool dwarf stars whose
 chemical compositions are unique clues to the early evolution of
 the Galaxy. The study is motivated by the power of ISO photometry
 to measure the temperature in the upper photospheric layers: the
 opacity in the infrared is provided by the free-free transitions of
 the H- ion and scales as wavelength-squared. Hence, a large range
 of depths is sampled by the flux between 3 to 120 micron. Previous
 studies of optical line and continuous spectra suggest that
 theoretical model atmospheres do not yet adequately represent the
 real atmospheres of cool dwarf stars. This is the case, for
 instance, of the very metal-poor star HD140283 for which standard
 model atmospheres with usual mixing length values seem to be unable
 to provide simultaneous fits to the H Alpha and Beta lines, and
 show a temperature distribution which is considerably hoter than
 that predicted by theoretical models in the lower layers. These
 inadequacies also motivate the study.