Light and heavy ions particle therapy, mainly by means of protons and carbon ions, represents an advantageous treatment modality for deep-seated and/or radioresistant tumours. An in-beam quality assurance principle is based on the detection of secondary particles induced by nuclear fragmentations between projectile and target nuclei. Three different strategies are currently under investigation: prompt γ rays imaging, proton interaction vertex imaging and in-beam positron emission tomography. Geant4 simulations have been performed first in order to assess the accuracy of some hadronic models to reproduce experimental data. Two different kinds of data have been considered: β+-emitting isotopes and prompt γ-ray production rates. On the one hand simulations reproduce experimental β+ emitting isotopes production rates to an accuracy of 24%. Moreover simulated β+ emitting nuclei production rate as a function of depth reproduce well the peak-to-plateau ratio of experimental data. On the other hand by tuning the tolerance factor of the photon evaporation model available in Geant4, we reduce significantly prompt γ-ray production rates until a very good agreement is reached with experimental data. Then we have estimated the total amount of induced annihilation photons and prompt γ rays for a simple treatment plan of ∼1 physical Gy in a homogenous equivalent soft tissue tumour (6 cm depth, 4 cm radius and 2 cm wide). The average annihilation photons emitted during a 45 s irradiation in a 4 π solid angle are ∼2 × 106 annihilation photon pairs and 108 single prompt γ whose energy ranges from a few keV to 10 MeV.