Abstract

Spectral analysis of B stars: An application of Bayesian statistics.

Jean-Michel Mugnes (Université Laval & CRAQ)

Carmelle Robert (Université Laval & CRAQ)

B stars are, because of their short lifetime, bound to their birth environment. Also, as very luminous objects, B stars allow us to indirectly probe near and far regions of the Universe. Furthermore, the most massive of them produce the heavy elements needed to enrich the Universe. All these points make the study of B stars (and massive stars in general) a center of interest for many areas of astrophysics such as galactic structure and dynamic, interstellar medium, or star forming region. Spectral analysis is a powerful tool for investigating stars, but the actual results often reveal important uncertainties. It is then vital to develop technics and methodologies that could reduce these uncertainties.

During this talk, I will show how I use the Bayesian theory along with the stellar atmospheric code TLUSTY and the synthetic spectrum program SYNSPEC (made by I.Hubeny and T. Lanz) to perform spectral analysis of B stars. The Bayesian probability theory is used as an extension of the logic to the case where, because of incomplete information, Aristotelian deductive reasoning is unavailable. Set aside for a long period of time in favor of a classical statistical approach, the Bayesian theory had to wait the 1940s and 1950s to get a complete mathematical demonstration and the 1980s for sufficiently powerful computers to demonstrate that it could outperform standard technics in many areas of science.

I will also present the first results obtained for a few B stars (observed at the Mont-Mégantic Observatory) which fundamental parameters (such as effective temperature, surface gravity…) have been classically determined by other authors. Comparison of these results will indicate the methodology reliability and efficiency or at least will help to improve it.