Calendar

Supernova remnants (and extended sources in general), are composed of a variety of components from different origins such as the shocked medium, the shocked ejecta or the accelerated electrons. Each component has a spectral signature (bremsstrahlung, emission lines, synchrotron, etc) and a spatial distribution that are projected along the line of sight and the perceived signal is a combination of these components. Spectro-imaging instruments  such as Chandra or XMM-Newton provide a 2D-1D view (X, Y, E) of extended sources. This is both an opportunity and a non-trivial challenge to disentangle the spatial distribution of the spectral components at stake. Whether it is to map the spatial distribution of heavy elements or the plasma properties, current analysis techniques (e.g. Voronoi tiling) treat each region independently and the disentangling process only relies on the spectral signature of the components.

With the current deep archival observations and in preparation for the next generation of telescopes, we need to operate a paradigm shift in the way we analyse X-ray data by drawing from the most advanced signal processing techniques to capture the wealth of information contained in those observations.

Here we propose to apply to X-ray astronomy blind source separation algorithms developed in cosmology to separate the CMB map from the foregrounds in the Planck data. This method exploits both the spectral and spatial signatures of the components yielding more discriminative power to disentangle the different physical components. We will present benchmarks of the methods using toy models and show preliminary results on the Chandra CasA dataset.