The interest for hadrontherapy, the use of ion beams for the treatments of cancerous tumors, is increasing. This can be attributed to the great accuracy of ion beams to target the tumor while sparing the surrounding healthy tissues (due to the high dose deposition in the Bragg peak and the small angular scattering of ions) as well as the potential biological advantage of ions for some tumor types compared to photons. To keep the benefits of carbon ions in radiotherapy, a very high accuracy on the dose location is required. The dose deposition is affected by the fragmentation of the incident ion that leads to a consumption of the projectiles with their penetration depth in the tissues (up to 70% for 400MeV/u 12C in water), to the creation of lighter fragments having different biological effectiveness (RBE) and to the apparition of a fragmentation tail behind the tumor. The constraints on nuclear models and fragmentation cross sections in 116 the energy range used in hadrontherapy (80 to 400MeV/u) are not yet sufficient to reproduce the fragmentation processes with the required accuracy for clinical treatments. A first integral experiment, on thick water equivalent targets has been performed by our collaboration on May 2008 at GANIL (France) [1]. The goals were the measurements of energy and angular distributions of the fragments due to nuclear reactions of 95MeV/u 12C with thick PMMA targets. Comparisons between experimental data and Geant4 simulations using different physics processes show discrepancies up to one order of magnitude for the production rates of fragments. The shapes of the angular and energy distributions are also not well reproduced. To improve the models and reach the accuracy required for a reference simulation code for hadrontherapy, a second experiment has been performed on thin targets on May 2011 at GANIL. The experimental set-up was made of five three stages E-E telescopes, each composed of two Si detectors and one CsI scintillator mounted on rotating stages to cover angles from 4 to 45 . The energy calibration of the detector and the identification of the fragments have been achieved. The double differential cross sections 2 /( E ) have been obtained for all fragment isotopes from p to 12C for nuclear reactions of 12C with H, C. Al, O and natTi. Simulations are under completion to evaluate the systematic uncertainties (first estimations lead to an accuracy of 10% for Z>2 fragments and up to 20-30% for protons at forward angles). The cross sections increase with the mass of the target. The results also show a more forward-focused angular distributions for heavier fragments and a predominance of Z=2 fragments at forward angles (< 10 ), compatible with the three alpha structure of the 12C. The energy distributions of the fragments at forward angles are peaked close to the beam energy showing an emission dominated by the quasiprojectile. The angular and energy cross sections obtained with a thin PMMA target have been reproduced with a 10% accuracy thanks to the elemental cross sections obtained on C, H and O targets. The experimental setup and the results for the different targets will be presented.