Solar and stellar magnetic polarisation

Light coming from magnetic, chemically peculiar (CP) stars of the upper Main Sequence shows a small but measurable linear polarisation which depends on the geometry of the magnetic field on the stellar surface, and which varies with the rotation of the star. The realistic modelling of this polarisation, the investigation of its properties, and the question of whether it can be employed in the diagnostics of stellar magnetic fields, all this was the major goal of the project in question. For this purpose it was necessary to develop an entirely new computer code named COSSAM (Codice per la sintesi spettrale nelle atmosfere magnetiche) which allows the synthesis of the spectrum of a rotating magnetic stars in linearly and circularly polarised light. Doing this for the visual domain of the spectrum, it is necessary to take into account hundreds of thousands of lines and a many times larger amount of components resulting from Zeeman splitting in the stellar magnetic field. Employing the object oriented programming language Ada95 and its parallel constructs, it has been possible to attain this goal, viz. the calculation of polarised spectra over large wavelength intervals as a function of magnetic field strength and direction, and the systematic study of the dependence of the integrated polarisation on the magnetic field parameters. On account of the heavy numerics involved in polarised radiative transfer, parallel computing turned out absolutely essential. For all practical purposes, object-oriented parallel computing has never been employed in (astro)physics. Whereas most spectral line synthesis codes are written in Fortran und therefore have to use Message Passing Interfaces (external libraries) in order to run in parallel, it is possible to parallelise codes written in Ada95 within the strict ISO standard, with all ensuing advantages concerning transportability and safety. It could be demonstrated, within the framework of this project, that Ada95 is not only a viable alternative to Fortran in the field of (astro)physical high performance computing, but that in many fields it is clearly superior to Fortran90. The newly developed code COSSAM also makes it possible - in an iterative approach - to directly recover the stellar magnetic field from observational spectropolarimetric data in intensity, linear and circular polarisation. As a first step towards this direct inversion a method has been developed and improved which is based on the modelling of various magnetic quantities under the assumption of the superposition of a dipole and a quadrupole component. It appears that in a number of CP stars such a dipole-quadrupole field configuration is a good approximation to the magnetic topology on the stellar surface. Finally, we have investigated the influence of the splitting of spectral lines in the magnetic field (the Zeeman effect) on the radiative acceleration of chemical elements in the stellar atmosphere. It turned out that, contrary to hitherto published assumptions, there can be an up to 10-fold increase in the acceleration due to the Zeeman effect.