Hundreds of different molecules have been observed in space. Some of these are relatively complex (alcohols, acids, peptides, etc.) and may serve as the basis for prebiotic chemistry, the beginnings of which are thought to have been found in the cold environments of the molecular clouds that precede the birth of stars. What is certain is that molecular diversity initially develops in these dense environments (for space: n >10e4 particles/cm3), protected from UV radiation, where the temperature drops to around 10K. In these regimes, molecules and atoms freeze on the surface of interstellar dust (condensates of silicates or carbonaceous soot of submicronic size) and undergo a very profound chemical transformation before returning to the gas phase on numerous occasions (birth of stars, collisions, cometary outgassing, etc.). The transition from gas to grain and back to gas is still poorly constrained, and this is the focus of our team's experimental work. We study the transformation of atoms (O, H, etc.) and the molecules they generate (H2, H2O, etc.) at the surface of interstellar dust grains, without any external energy input (photons, electrons, etc.). We are interested in accretion or sticking, surface diffusion and return to the gas phase, as well as reaction pathways. In dedicated UHV chambers, we expose cold surfaces (10-100K) to atomic (H,O,N..) or molecular (CO, H2O...) gases, monitoring the composition of the surface by infrared spectroscopy, and exchanges with the gas phase by mass spectroscopy. We are particularly interested in surface mechanisms at the molecular layer (or sub-layer) scale, where the most efficient processes take place. In my presentation, after setting out the astrophysical context, I will present our experimental set-up, illustrating the methods we use to study the formation of H2 and H2O, as didactical example, before before moving on briefly to other topics such as the formation of complex molecules, or the study of snow lines displaced by the formation of ammonium salts.​