GPU-Accelerated PIC-DSMC Simulation for Modeling Cu PVD Process

Plasma-based PVD processes play a crucial role in semiconductor and display manufacturing, requiring precise physical modeling under low-pressure conditions and complex asymmetric magnetic field structures. Under these conditions, the application of fluid equation approaches based on continuum assumptions has limitations, making the introduction of Particle-In-Cell (PIC), a particle-based simulation technique, essential. In this study, we developed a GPU-accelerated PIC simulator and incorporated a Direct Simulation Monte Carlo (DSMC) module that considers molecular collisions for more direct deposition process simulation. Specifically, Ar and Cu were calculated using the DSMC method, and based on this, the distribution of Cu, Ar⁺, and Cu⁺ particles delivered to the wafer was analyzed. As a result, quantitative similarity between the calculated deposition rate distribution and experimental process results was confirmed. The sputtering ratio, angle, and energy information of target Cu play important roles, and their influence on deposition distribution was also analyzed. Through this methodology, it is expected that various PVD processes with different recipes, including sputtering target materials, can be precisely simulated. However, PVD equipment typically has wide discharge regions and often requires three-dimensional analysis, resulting in numerical limitations such as high computational resource requirements and complex boundary condition handling. This study presents the applicability and usefulness of multidimensional PIC-based simulation for optimizing practical PVD equipment and process simulation.