Résumé : About 60 percent of the electricity production in Belgium originates from nuclear power plants. Belgium owns 7 nuclear pressurized water reactors, which are located in two sites: 4 reactors in Doel and 3 reactors in Tihange. Together they have a capacity of approximately 5900 MWe. All these reactors use classical uranium oxide fuel assemblies. Two of them (Doel3, Tihange2) have also accepted a limited number of mixed (uranium and plutonium) oxide fuel assemblies. These mixed fuel assemblies came from the reprocessing of spent uranium oxide fuel assemblies in La Hague (France). The reprocessing of spent fuel gives birth to vitrified high-level waste, and to different isotopes of uranium and plutonium, which can be used in the manufacture of mixed oxide fuel assemblies. Each country producing radioactive waste must find a solution to dispose them safely. The internationally accepted solution is to dispose high-level radioactive waste in a deep and stable geological layer. This seems to be the most secure and environment-friendly way to get rid of the high-level radioactive waste. One of the few stable geological layers, which could accept radioactive waste in Belgium, is the Boom clay layer. Another possible layer is the Ypresian clay layer, but it is not the reference option for the moment. The Boom clay layer is quite thin (about 100 m thick) and is not at a large depth (about 240 m below the ground surface) at the proposed disposal site, beneath the SCK CEN Nuclear Research Centre in Mol. A large number of studies have already been performed on the Boom clay layer, and on the possibility of building a high-level radioactive waste repository in this geological medium. Since 1993, the Belgian government has promulgated a moratorium on the reprocessing of spent uranium oxide fuels in La Hague. Since then, spent fuel assemblies are considered as waste, and ONDRAF/NIRAS (the Belgium Agency for Radioactive Waste and Enriched Fissile Materials) has thus to deal with them as waste. This rises a number of questions on how to deal with this new kind of waste. A solution is to directly dispose these spent fuel assemblies in containers in a repository, just like the other high-level radioactive waste. This repository would be build in the Boom clay layer at a depth of about 240 m beneath the SCK CEN. One of the questions raised by this new kind of waste is: "could the direct disposal of the spent nuclear fuel assemblies lead to nuclear criticality risks in the future?". Nuclear criticality is the ability of a system to sustain a nuclear fission chain reaction. This question was not a key issue with vitrified high-level waste because these do not include fissile uranium and plutonium isotopes, which could lead to a criticality event. The spent fuel repository will be designed in order to totally avoid the occurrence of a criticality event at the closure time. But in the future history of the repository, external events could possibly affect this. These events could maybe lead to criticality inside the repository, and this has also to be avoided. This work tries to answer this question, and to determine how to avoid a long-term criticality event inside the repository. The only complete research work answering this question has been performed in the U.S. for the Yucca Mountain repository but this design is fully different from the Belgian one studied here: for example, the waste are not only spent fuel waste, and the geological layer is volcanic tuff.

The main achievements of this work are:

*A first set of in-package criticality scenarios for different design options for a Belgian spent fuel repository in the Boom clay layer.

*A large number of criticality calculations with different parameters (fuel type, fuel burnup, fuel enrichment, distance between the fuel assemblies, distance between the fuel rods, water fraction inside the overpack) for the different design options.

*A preliminary study of the effects of the spent fuel assemblies isotopic evolution with time on the multiplication factor.

*For the first time, a coupling between the in-package criticality scenarios and the criticality calculations has been performed.