Source separation algorithms for the analysis of hyper-spectral observations of very small interstellar dust particles

The mid-infrared (mid-IR) spectrum of our galaxy is dominated by continuum and band emission due to carbonaceous very small dust particles amongst which are polycyclic aromatic hydrocarbon (PAH) molecules. Because they absorb the UV photons of massive stars and re-emit this energy in the infrared, IR spectro-imaging of extended interstellar (or circumstellar) regions is a powerful tool to diagnose the nature of these particles together with the local physical conditions. In this paper, we review how the applications of blind/bayesian source separation (BSS) methods applied to mid-IR hyperspectral data can help analyzing these data. We then discuss, in the light of simulations in progress, how BSS methods could be used to identify specific PAH molecules in the interstellar medium when applied to the hyper-spectral data of forthcoming IR telescopes (Herschel, SOFIA, SPICA).