Bombardment of Graphite and Amorphous Carbon Surfaces Using Molecular Dynamics Simulations: Toward A More Realistic Experimental Model

Molecular dynamics (MD) simulations are a useful computational tool in fields such as fusion research. Small but vital portions of fusion reactors are essential to their correct operation and longevity. Using the reactive bond order (REBO) and adaptive intermolecular REBO potentials, it is possible to model carbon-based systems, such as graphite diverter plates, under simulated bombardment. The degradation of these plates due to random bombardments from plasma can eventually incur costly shut downs. To gain a better understanding of the atomic-level dynamics that occur when a graphite and amorphous carbon surface undergo energetic, serial bombardment by atoms such as hydrogen, deuterium, and tritium, these two systems were evolved with the REBO and AIREBO potentials. It was found that the AIREBO potential gave different results with regards to surface evolution, sputter yield, and steady state formation. Graphite surfaces evolved to a much different steady state when compared to amorphous carbon, which lead to varied surface structure and may also lead to differing sputtering yields. An additional round of simulations was performed on graphite surfaces that were deeper in the direction normal to the surface. Based on the previous results, the AIREBO potential and two different bombardment energies were used, and the additional layers added allowed for greater fluences, defined by the number of impacts per unit area, to be achieved. As an additional improvement of the previous work, thermostats were set by using zones of control rather than employing the thermostat on the entire system, achieving atomic layer control of the thermostatted regions during the simulation. After employing these changes and evolving the simulations for only slightly larger fluences than previous simulations, the formation of voids within the graphite layers, or \u27bubbles\u27, was produced. Particle build-up consisting of gaseous D, D2, and other small molecules near the penetration depth caused the formation of these bubbles. It was found for 20 eV impact energies the penetration depth is well defined, because of the lower energy of insertion. The stopping power of the potential on these low energy insertions leads to a noticable build-up of D atoms near the penetration depth. For the 80 eV simulations, the penetration depth is broadened when compared with the 20 eV simulations. The impacts penetrate more layers with increased impact energy, with bubble formation occurring away from the average penetration depth. A comparison of retention ratios is also discussed, and found that the 80 eV simulations retained more D than the 20 eV simulations. To attempt to avoid the issue of bubble formation, and to expand on the capabilities of the MD code, graphite surfaces were expanded in the directions perpendicular to the insertion direction, and the ability to bombard the surface with multiple atom types was implemented. Another improvement was introduced in the code to allow the variable time step algorithm to be used in conjunction with the thermostat. These systems yielded a closer model to experimental conditions, where the energy of interaction between the layers of graphite is larger than the insertion energy of the incident particles. While only smaller fluences compared to previous work have been achieved for these systems, the systems have shown promise in terms of their surface evolution and behavior