Fast ballistic transport algorithm for direct simulation Monte Carlo applied to thin-film process simulation

The thin-film deposition process is critical to semiconductor manufacturing. New device structures and advanced process technology have prompted growing industrial interest in developing novel methods and models to predict process outcomes. Capturing morphological features such as pinhole defects and void closure time demands a simulation with both a large domain and fine resolution, which is, however, severely limited by the high computational cost of ballistic transport. To improve the simulation efficiency of thin-film deposition process, we propose a fast vacuum ballistic transport algorithm based on a two-dimensional segment tree and finite state machine for direct simulation Monte Carlo. We demonstrate that this approach achieves lower time complexity and offers better compatibility than some existing accelerate algorithm, thereby enabling tighter coupling with thin-film process simulations. The model is calibrated against experimental results of silicon nitride grown by plasma-enhanced chemical vapor deposition (PECVD). Benefiting from the improved runtime efficiency, enabling larger simulation domains with finer resolution, the method reveals nanometer-scale pinhole and void formation in deep trenches. The proposed algorithm yields high prediction accuracy when integrated with a PECVD profile simulation model. Moreover, this algorithm can be integrated into simulation models involving complex surface events as an acceleration module.