The objective of this thesis is to develop the modelling of the nuclear heating of the experimental devices of the future Jules Horowitz material testing reactor (RJH). The strong specific nuclear power produced (460 kW/l) induces intense photonic fluxes which cause heating and large gradients in temperature. The design studies of the experimental devices require an accurate knowledge of these gradients. A priori, we estimate that there are large uncertainties on gamma heating calculations to about 30% (2σ). They are mainly due to the lack of knowledge about gamma emission yields and spectra and secondly come from the mismodelling of the gamma response function. In order to specify this uncertainty, a set of integral dose measurements using different kinds of TLD and a micro-ionization chamber are carried out in the UO2 core of the MINERVE research reactor. The comparison between calculated and measured integral (prompt plus delayed) gamma-ray absorbed doses in TLD shows that calculation constantly underestimates the measurement with C/E = 0.72 ± 15% (k = 2) of total uncertainties. The systematic bias observed is partially attributed to a lack of gamma-ray emission data in basic nuclear data libraries used. A Bayesian method of adjustment is developed in order to re-estimate the principal components of the gamma heating and to transpose the results obtained to the devices of the RJH. This study made it possible to reduce significantly the uncertainties on the determination of the gamma heating from 30% to 15% (2σ).
