The ionization fraction controls several key processes in giant molecular clouds (GMC): fast ion-neutral reactions driving interstellar chemistry, gas coupling to the magnetic field, excitation of key tracers. Estimating the ionization fraction in the different regions of a GMC, from its diffuse enveloppe to its dense cores, is thus a key step towards understanding its chemico-physical structure, and their link to star formation. However, classical tracers (e.g., DCO+/HCO+) are only detectable in the densest cores.
We propose a statistical approach based on Random Forests, a flexible machine learning model, exploiting large grids of astrochemical models to automatically find the best tracers of the ionization fraction among hundreds of species (Bron et al., 2020). We find several new tracers detectable in the extended enveloppe of the cloud.