The interstellar medium, filling the volume between the stars of a galaxy, is constituted of two main components: gas and dust. Dust grains are small solid particles, mainly composed of silicate and carbonaceous materials. They play a major role in the physics of the interstellar medium, although accounting for only one percent of its mass. Indeed, they absorb, and reemit in the infrared, an important fraction of the power radiated by stars and accretion disks. In particular, star forming regions are completely opaque in visible light. Only the infrared radiation, emitted at 99% by dust, allows us to study them. Grains are also responsible for the gas heating by photoelectric effect, in photodissociation regions (PDR). Finally, grains are catalysts of numerous chemical reactions, including the formation of dihydrogen, the most abundant molecule in the universe.
The properties of these dust grains (abundance, chemical composition, size distribution, etc.), as well as their evolution, are, however, still poorly known. This is the direct consequence of the complexity of this component and of the lack of observations discriminating different models. These uncertainties are affecting numerous aspects of our knowledge in astrophysics: mass measurements, unreddening (i.e. correction of the extinction along the line of sight), detailed PDR models, etc. Refining our understanding of dust is also crucial to understand the interstellar lifecycle, as grains regulate several processes controlling this cycle. An accurate understanding of grain
physics is thus necessary to understand galaxy evolution.
An approach, to tackle these open questions, consists in studying the way observed grain properties vary with the physical conditions they experience. Such empirical relations, if they are precise enough, allow us to remove some degeneracies on different models. The thesis that we are proposing focuses on the detailed study of the smallest grains (radius < 10 nm) and polycyclic aromatic hydrocarbons (PAH). These components of the interstellar medium radiates out-of-equilibrium in the mid-infrared (5–40 microns). This is the wavelength domain that contains most of the solid state resonance features.
This study will focus on several nearby galaxies, including the Magellanic clouds. The interest of nearby galaxies compared to the interstellar medium of our galaxy resides in the diversity of physical conditions (metallicity, radiation field intensity, etc.)
Several studies have already been published on this topics, especially with the Spitzer space telescope. However, most have been somehow superficial. Numerous aspects remain to be studied: (i) the correlation of the main aromatic bands with the physical conditions; (ii) constraining the evolution of their size distribution; (iii) identifying and modelling several bands of solids in star forming regions. One of the originalities of this thesis will consist in developing a sophisticated method to model the data. Indeed, most previous studies have performed simple linear decompositions. We propose that the student develop a hierarchical Bayesian decomposition code to analyze infrared spectra, with constraints provided by atomic, molecular and solid-state databases. This type of code allows to physically model the sample and to statistically model the distribution of parameters, simultaneously. It allows us to remove several degeneracies and to
extract the maximum information from the data, taking into account the various sources of uncertainties, without overinterpreting the observations. We have recently developed such a code to model spectral energy distributions, and the results are convincing. This new tool and its meticulous application to the data are the warranty of a precise and original interpretation of the physical processes taking place in the studied regions.
The James Webb Space Telescope (JWST), which will be launched in 2019, will observe the mid-infrared domain with
unprecedented sensitivity and saptial resolution. The methods developed during the thesis could be applied to these new data.