材料科学
剥离(纤维)
薄脆饼
剥脱关节
复合材料
冶金
光电子学
破损
沉积(地质)
湿法清洗
作者
Hongmei Li,Hongwei Wang,Yuxin Li,Wenyang Cen,Liqun Wang,Lin Li,Yinzhou Yan,Kewen Pan,Wei Guo
标识
DOI:10.1016/j.jmatprotec.2026.119226
摘要
High-efficiency, low-damage exfoliation of 4H-SiC wafers is critical for their application in the power electronics sector and for reducing manufacturing costs. However, the high exfoliation force required after conventional laser modification is a core challenge for this technology. This study systematically reveals two primary physical causes for the high exfoliation force. First, the laser induces a complex modified layer, comprising an amorphous phase and 3C-SiC, resulting in strong interfacial bonding with the 4H-SiC substrate. Second, the crystal’s inherent ∼4° off-axis growth mode results in a stepped and staggered crack distribution during propagation, which prevents the cracks from becoming fully interconnected. To address these issues, an innovative hybrid exfoliation technique is proposed and validated. The proposed technique combines ultrasonic cavitation with selective chemical etching using a KOH solution. Ultrasonic vibration and cavitation physically drive microcrack propagation, while selective chemical etching weakens the interfacial bonding strength of the modified layer. This synergistic method reduces the exfoliation force to 26.85 N (representing a reduction of 66.01 %). Building on this, we further optimize the laser modification process by designing a dual-modification strategy that introduces interconnection points to proactively control the crack path and ensure complete crack-network connectivity. Ultimately, this fully optimized strategy reduces the exfoliation force to 19.85 N and achieves efficient exfoliation of a complete 6-inch 4H-SiC wafer in approximately 3 min. This study elucidates the specific origins of the high exfoliation force through microstructural and elemental analysis of the modified layer and provides a new feasible and scalable solution for low-damage wafer exfoliation.
科研通智能强力驱动
Strongly Powered by AbleSci AI