绝缘栅双极晶体管
开裂
动力循环
材料科学
功率(物理)
自行车
结构工程
复合材料
电气工程
电压
可靠性(半导体)
工程类
量子力学
历史
物理
考古
作者
Shaojun Zhao,Qi Wang,Tong An,Fei Qin
出处
期刊:IEEE Access
[Institute of Electrical and Electronics Engineers]
日期:2024-01-01
卷期号:12: 196105-196117
标识
DOI:10.1109/access.2024.3519663
摘要
For high-power modules with wire bonding as the interconnection method, fatigue damage and cracking at the bond interface are important forms of module failure. However, the currently used numerical models of the bond interface neglect the influence of microdefects and damage evolution of the interface material and cannot accurately describe the degradation process of the mechanical properties of the bond interface. In this work, the shear strength of the Al-bonded wire-Al metallization layer bond interface of an insulated-gate bipolar transistor (IGBT) module after different numbers of power cycles was measured via shear tests, and force-displacement (F– $\delta $ ) curves and fracture surface morphologies were obtained. The experimental results indicate that the bond interface strength decreases significantly as the number of power cycles increases. To describe this phenomenon, the cohesive zone model-based finite discrete element method (CZM-based FDEM) is introduced in the bonding zone; that is, the bonding zone is discretized via triangular elements, and cohesive elements are inserted between adjacent triangular elements to describe the cracking process of the bond interface. By randomly assigning different material property parameters to the cohesive elements, the microdefects can be characterized, and by adjusting the proportions of cohesive elements with different strengths, the phenomenon whereby the bond interface strength decreases during power cycling can be better demonstrated. Finally, a comparison with the results of shear tests validated that this method can effectively predict fracture processes at the bond interface and is able to describe the degradation of the interfacial mechanical properties observed in the experiments.
科研通智能强力驱动
Strongly Powered by AbleSci AI