A suitable subcategory of affine Azumaya algebras is defined and a functor from this category to the category of Zariski structures is constructed. The rudiments of a theory of presheaves of topological structures is developed and applied to construct examples of structures at a generic parameter. The category of equivariant algebras is defined and a first-order theory is associated to each object. For those theories satisfying a certain technical condition, uncountable categoricity and quantifier elimination results are established. Models are shown to be Zariski structures and a functor from the category of equivariant algebras to Zariski structures is constructed. The two functors obtained in the thesis are shown to agree on a nontrivial class of algebras.

Dans cette thèse, nous examinons comment un assistant de preuve peut être utilise pour étudier les fondements de la géométrie. Nous débutons en nous concentrant sur les façons d’axiomatiser la géométrie euclidienne et leurs relations. Ensuite, nous exposons une nouvelle preuve de l’indépendance de l’axiome des parallèles des autres axiomes de la géométrie euclidienne du premier ordre. Cela nous amène à affiner la classification des plans de Hilbert de Pejas en considérant les propriétés de décidabilité. Mais, notre intuition nous amène souvent à négliger leur utilisation. Un assistant de preuve nous permet d’utiliser un outil parfait qui ne possède aucune intuition : un ordinateur. De plus, les assistants de preuve nous laissent exploiter les capacités de calcul des ordinateurs. Nous démontrons comment utiliser de méthodes algébriques de déduction automatique en géométrie synthétique. Enfin, nous présentons une procédure spécifique destinée à automatiser des preuves d’incidence. / In this thesis, we investigate how a proof assistant can be used to study the foundations of geometry. We start by focusing on ways to axiomatize Euclidean geometry and their relationship to each other. Then, we expose a new proof that Euclid’s parallel postulate is not derivable from the other axioms of first-order Euclidean geometry. This leads us to refine Pejas’ classification of parallel postulates. We do so by considering decidability properties when classifying the postulates. However, our intuition often guides us to overlook uses of such properties. A proof assistant allows us to use a perfect tool which possesses no intuition: a computer. Moreover, proof assistants let us leverage the computational capabilities of computers. We demonstrate how we enable the use of algebraic automated deduction methods thanks to the arithmetization of geometry. Finally, we present a specific procedure designed to automate proofs of incidence properties.