Abstract : In order to unserstand the fragement production which occurs in heavy ion collisions at intermediate energies we have performed in this work an expérimental and theoretical study of multifragementation. The set of data obtained with the INDRA detector over a wide range of incident energy for the Xe+Sn system allows to perform a study of central collisions which shows that the maximum of fragment production is around 65 MeV/n of incident energy, based on fragment/particle correlation functions, has helped to determine the characteristics of primary fragments produced by the collision.The excitation energy of these fragments shows a saturation at a value of 3 MeV/n beyond 39 MeV/n of beam energy and the evaporated particle represent less than 40% (23% at 50 MeV/n) of all light charged particles, which indicates the importance of the collision dynamics. Calculations with the antisymmetrised molecular dynamics model AMD have been made in order to study the role of dynamics and the time évolution of the collision. The simulations are in good agreement with the experimental data, namely for the most central collisions at 50 and 100 MeV/n of beam energy for the Xe+Sn system. In this scope some developments have been neceesssary so that the diffusion properties of nucleons in nuclear matter can be better taken into account. The calculations allowed to locate the fragment time production between 100 and 200 fm/c in central collisions at 50 MeV/n of beam energy. Furthermore they showed a transparency effect even in the most central collisions. The comparison with the experimental data show that this effect is slightly overestimated in the calculations.