Atomistic simulations of graphite etching at realistic time scales

Summary¶

Hydrogen plasmas erode graphite plasma-facing components through a combination of prompt collisional damage and slower thermochemical pathways that unfold between ion impacts. Aussems et al. combine ReaxFF chemistry with collective-variable hyperdynamics (CVHD) in LAMMPS to extend the effective time between energetic hydrogen impacts toward roughly one millisecond at a reference flux of order \(10^{20}\) m\(^{-2}\) s\(^{-1}\), matching the abstract’s framing. The central thesis is that dwell time between impacts determines whether erosion is dominated by knock-on sputtering or by thermally activated C–C chemistry—a distinction that collapses when MD schedules impacts too frequently for computational convenience.

The motivation spans fusion devices and related environments where hydrogen–graphite interactions control impurity sources and wall erosion, but where laboratory ion fluxes are orders of magnitude lower than those implicit in picosecond-spaced MD impacts. Bridging that gap requires accelerated dynamics so relaxation, diffusion, and thermochemical bond breaking between impacts enter the same simulation as prompt collisional events.

Methods¶

MD application (LAMMPS + CVHD). All simulations use LAMMPS with a modified Colvars module implementing collective-variable hyperdynamics (CVHD), which builds a self-learning bias potential to accelerate basin-to-basin transitions (papers/ReaxFF_others/Aussems_ReaxFF_graphene_H.pdf, Simulation model section). ReaxFF is used instead of second-generation REBO because long-range van der Waals, Coulomb, and torsion terms matter for H–graphite chemistry under reactive conditions. The work compares unbiased H bombardment with CVHD runs that stretch the effective dwell time between keV-scale H impacts across nine orders of magnitude, reaching ~1 ms between impacts for a reference flux ~10²⁰ m⁻² s⁻¹ as stated in the abstract. Graphite slab dimensions, PBC, thermostat parameters (Berendsen / Nosé–Hoover as applicable), timestep, NVT/NPT staging, production duration, temperature, pressure coupling, and any electric field are not duplicated in the abstract-level note used here—extract them from Methods in pdf_path. CVHD is the enhanced-sampling mechanism; unbiased bombardment supplies controls.

Force-field training: N/A — applies an established ReaxFF description for C/H chemistry cited in the article; the novelty here is CVHD-accelerated dynamics, not a new FF fit.

Findings¶

Erosion yield, H coverage, and emitted species distributions hinge on time between impacts: longer gaps favor thermally activated C–C chemistry, whereas picosecond-spaced impacts emphasize prompt knock-on removal—a regime prior bombardment studies could not reach. Readers must therefore treat simulated flux as a tunable knob, not a literal match to experiment, when chemical erosion matters. CVHD is framed as a bridge rather than a drop-in replacement for unbiased long-time kinetics. Limitations: bias potentials alter dynamics; ReaxFF omits electronic excitations from the plasma environment.

Limitations¶

Hyperdynamics biases dynamics; quantitative erosion yields require unbiased segments and experimental cross-checks. ReaxFF cannot capture electronic excitations from the plasma sheath.

Relevance to group¶

Methodological reference combining ReaxFF with accelerated dynamics for carbon plasma/etching chemistry.

Citations and evidence anchors¶

DOI: 10.1039/C7SC02763J.