Nuclear graphite: a computational and experimental study

Cracking in the Gilsocarbon moderator in nuclear reactors, as a result of irradiation and oxidisation, is still not fully understood due to complex microstructure and properties, where the models are unable to accurately determine where the moderator is in its lifecycle based on the cracks, often leading to premature shutdown.

Models usually treat properties, such as the Coefficient of Thermal Expansion (CTE), as uniform across the moderator, yet the literature has shown the filler particle and binder components in Gilsocarbon have complex microstructures. Routine property measurements are unable to determine specific properties of individual components, hence, in this research a new approach has been developed to measure the CTE at the micro scale, which was previously not possible. The purpose of this CTE characterisation was to build a better understanding with the microstructure, before building physical models based on the results. By better understanding the CTE on a smaller scale, it will allow more accurate larger scale models to be build.

Gilsocarbon was investigated using both optical microscopy and Scanning Electron Microscopy across fractured and polished surfaces to investigate the structure and explore potential Raman Spectroscopy / Focused Ion beam lamella lift-out sites. Raman Spectroscopy was mainly concerned with investigating the disorder and calculating the crystallite size of the binder and filler regions. This was followed by Transmission Electron Microscopy to investigate the microstructure further. FIB lamellae from the filler particle and binder sections had their CTEs measured with the use of a heating chip and tracking expansion with deposited or milled fiducial marks. This was following confirmation that the method worked after characterising the CTE of controlled copper and silicon isotropic samples.

The CTE was modelled using Large Scale Atomic Molecular Massively Parallel Simulations (LAMMPS) using Airebo and ReaxFF potentials. Graphene sheets were simulated to begin with, before using single and twin defect lattices and randomly orientated grain boundaries, consisting of pentagon and heptagon pairs among normal graphene layers, following the observation of Moiré patterns in TEM. Nested fullerenes were also simulated to represent the layered filler particle onion-like structure.

The results show that Gilsocarbon has a complex microstructure, with layered spherical filler particles, surrounded by a coarse binder region. Raman spectroscopy also confirmed this by plotting the crystallite size. This was, however, only achievable on polished surfaces, as fracture faces were rough and no areas could be accurately distinguished. The CTE results of the lamella revealed that the filler particles had a much higher CTE compared to the binder. It is proposed that the difference in CTE means the filler particles contract at a faster rate compared to the binder regions during the cooling in manufacturing, where the difference in contraction causes a pre-tensile stress across the binder, which may have an affect on the response to irradiation.

TEM observations revealed the microstructure across the filler particle and binder to vary, with the latter having more defects across the structure. Numerous features were observed including: Mrozowski cracks; a crazy paving structure, only in the binder; and a never observed "scale network", which later turned out to be a pore network, which gives the white cracks their distinct contrast. The atomistic computation revealed that nested fullerenes were a poor model for the onion structures, as they ripped apart at moderate temperatures, whereas experimentally the structures remained in- tact. The randomly orientated grain boundaries did show some similarities with the lamella CTE results.

The findings on this PhD suggests there is pre-tensile stress existing across the binder regions. This could have a significant effect on the bulk material response to irradiation where different regions change dimensions at different rates, which may explain the cracking mechanism inside nuclear reactors.