Theoretical and experimental investigation on topside fracture formation mechanisms in semiconductor silicon dicing: Modeling of fracture area based on blade structure

Mechanical dicing is the final step in silicon wafer processing and directly determines the quality of the chips. Topside fracture readily occurs during this process, increasing the risk of chip failure. However, there is currently no research on the mechanism of topside fracture formation and control methods specifically for dicing. To address this challenge, this study investigates the formation mechanism of topside fracture during silicon wafer dicing. The dicing blade is divided into two grinding zones based on material removal: the primary grinding zone (PGZ) on the outer circumference, which performs plunge grinding, and the secondary grinding zone (SGZ) on the side of the blade, which performs face griding. Topside fracture is classified into three types—the primary grinding zone, the secondary grinding zone, and mixed fracture—based on the interaction between these zones. The formation processes of these fracture types are discussed in detail. A topside fracture area model, which accounts for blade characteristics, was developed by integrating the three fracture formation mechanisms with crack propagation conditions, addressing the theoretical gap in the study of fracture during the dicing of brittle materials. Subsequently, a series of silicon dicing parameter experiments were conducted, accurately identifying the three types of fracture. It was found that the proportion of secondary grinding zone fracture was inversely proportional to the average fracture area. Therefore, by referring to the model and controlling the processing parameters, it is possible to reduce the proportion of secondary grinding zone fracture and, in turn, control the fracture area. Additionally, the segmented model established in this study, which takes different types of fracture into consideration, reduced the average error rate of fracture area calculations from 19.5 % to 12.51 %, compared to a model that does not account for fracture types. This research provides valuable insights into reducing silicon dicing fracture areas and improving dicing quality, serving as a reference for the study of fracture in the dicing of other hard-brittle materials.