Modeling of dense carbons using a polygranular image-guided approach and characterization of the structure-elasticity relationship

The ability of C/C composites to maintain high mechanical properties up to very high temperatures (above 3000 K), combined with their low density, justifies their use in extreme conditions, especially in aerospace applications. However, due to the highly anisotropic nature of their constituents, partial crystalline order, and the challenges in conducting certain tests under operational conditions, the relationship between the structure of these materials and their mechanical behavior is not well-established. Particularly, predicting the individual behavior of different constituents (fibers, matrices) at a small scale is a fundamental step in the development of virtual materials at the composite level. In this thesis, 210 atomistic models of dense carbons have been reconstructed using a Molecular Dynamics-based Poly-Granular Image-Guided Atomistic Reconstruction (PG-IGAR) method. These models are generated using a parametric approach that varies the grain size and their distribution of the 002 orientation directions. The models are then characterized by analyzing local atomic environments, calculating X, neutron, and electron diffraction properties, and HRTEM images. The mechanical characterization is performed by calculating isothermal and adiabatic (i.e., at infinitely rapid deformation rate) elasticity tensors. A structure-elasticity relationship is established through machine learning. Within the generated database, six models of pyrolytic carbons (pyCs) are identified by closely matching an experimental database, and their structural and mechanical properties are compared. Ultimately, the elastic properties of these models contribute to a continuous model of non-elastic behavior. (author)