Bibcode
DOI
Allende Prieto, Carlos; Barklem, Paul S.; Asplund, Martin; Ruiz Cobo, Basilio
Bibliographical reference
The Astrophysical Journal, Volume 558, Issue 2, pp. 830-851.
Advertised on:
9
2001
Journal
Citations
43
Refereed citations
39
Description
Spectra of late-type stars are usually analyzed with static model
atmospheres in local thermodynamic equilibrium (LTE) and a homogeneous
plane-parallel or spherically symmetric geometry. The energy balance
requires particular attention, as two elements that are particularly
difficult to model play an important role: line blanketing and
convection. Inversion techniques are able to bypass the difficulties of
a detailed description of the energy balance. Assuming that the
atmosphere is in hydrostatic equilibrium and LTE, it is possible to
constrain its structure from spectroscopic observations. Among the most
serious approximations still implicit in the method is a static and
homogeneous geometry. In this paper, we take advantage of a realistic
three-dimensional radiative hydrodynamical simulation of the solar
surface to check the systematic errors incurred by an inversion assuming
a plane-parallel horizontally-homogeneous atmosphere. The thermal
structure recovered resembles the spatial and time average of the
three-dimensional atmosphere. Furthermore, the abundances retrieved are
typically within 10% (0.04 dex) of the abundances used to construct the
simulation. The application to a fairly complete data set from the solar
spectrum provides further confidence in previous analyses of the solar
composition. There is only a narrow range of one-dimensional thermal
structures able to fit the absorption lines in the spectrum of the Sun.
With our carefully selected data set, random errors are about a factor
of 2 smaller than systematic errors. A small number of strong metal
lines can provide very reliable results. We foresee no major
difficulties in applying the technique to other similar stars, and
obtaining similar accuracies, using spectra with
λ/δλ~5×104 and a signal-to-noise
ratio as low as 30.