Automated Determination of Stellar Parameters from Simulated Dispersed Images for DIVA

Abstract

We have assessed how well stellar parameters (T_eff, logg and [Fe/H]) can be retrieved from low-resolution dispersed images to be obtained by the DIVA satellite. Although DIVA is primarily an all-sky astrometric mission, it will also obtain spectrophotometric information for about 13 million stars (operational limiting magnitude V ~ 13.5 mag). Constructional studies foresee a grating system yielding a dispersion of ~200nm/mm on the focal plane (first spectral order). For astrometric reasons there will be no cross dispersion which results in the overlapping of the first to third diffraction orders. The one-dimensional, position related intensity function is called a DISPI (DISPersed Intensity). We simulated DISPIS from synthetic spectra (...) for a limited range of metallicites i.e. our results are for [Fe/H] in the range -0.3 to 1 dex. We show that there is no need to deconvolve these low resolution signals in order to obtain basic stellar parameters. Using neural network methods and by including simulated data of DIVA's UV telescope, we can determine T_eff to an average accuracy of about 2% for DISPIS from stars with 2000 K < T_eff < 20000 K and visual magnitudes of V=13 mag (end of mission data). logg can be determined for all temperatures with an accuracy better than 0.25 dex for magnitudes brighter than V=12 mag. For low temperature stars with 2000 K < T_eff < 5000 K and for metallicities in the range -0.3 to +1 dex a determination of [Fe/H] is possible (to better than 0.2 dex) for these magnitudes. Additionally we examined the effects of extinction E(B-V) on DISPIS and found that it can be determined to better than 0.07 mag for magnitudes brighter than V=14 mag if the UV information is included.