A big yield drop has been observed during the automatic inspection （AO1） after the saw stage. A step by step AOl inspection check and defect review is made to see which step made a big yield drop and which kind of d...A big yield drop has been observed during the automatic inspection （AO1） after the saw stage. A step by step AOl inspection check and defect review is made to see which step made a big yield drop and which kind of defect contributed most to the yield drop. Scanning electron microscope （SEM） and energy dispersive spectrometer （EDS） analysis showed the shape and chemical element of the particle. From the EDS result, particles can be separated into two categories. One was the inorganic related materials, mainly including silicon （Si） element, which came from the saw stage. A design of experiment （DOE） is used to find some reasonable saw relative parameter and optimize it in order to remove the particle from the saw stage. But the quantity of this kind of particle was small. Yield was only improved by less than 5%. Our main effort was to remove another kind of particle which was organic related materials, mainly including carbon （C） and oxygen （O） elernent. This kind of particle was from tape residue. In order to remove the tape residual, one step was added before the saw stage. Almost all of the tape residual was removed. Finally, the final yield was improved by more than 15%.展开更多
文摘晶圆良率是衡量半导体制造系统加工能力的关键指标,对其精准预测有利于排查晶圆制程工艺缺陷、提高晶圆生产率、控制企业生产成本。基于晶圆允收测试(wafer acceptance test,WAT)大数据,提出了基于卷积神经网络和长短期记忆网络(convolutional neural networks and long short-term memory,CNN-LSTM)的晶圆良率预测方法。该方法对WAT数据进行缺失、异常与归一化预处理;构建CNN模型对复杂WAT参数的关键特征进行识别;考虑相邻晶圆间的时序相关性,设计长短期记忆网络进行回归分析,从而实现晶圆良率的准确预测。以某工厂晶圆允收测试过程中采集的实际生产数据进行实验,并与其他传统晶圆良率预测方法的结果进行对比分析,从而验证所提方法的有效性。
文摘A big yield drop has been observed during the automatic inspection （AO1） after the saw stage. A step by step AOl inspection check and defect review is made to see which step made a big yield drop and which kind of defect contributed most to the yield drop. Scanning electron microscope （SEM） and energy dispersive spectrometer （EDS） analysis showed the shape and chemical element of the particle. From the EDS result, particles can be separated into two categories. One was the inorganic related materials, mainly including silicon （Si） element, which came from the saw stage. A design of experiment （DOE） is used to find some reasonable saw relative parameter and optimize it in order to remove the particle from the saw stage. But the quantity of this kind of particle was small. Yield was only improved by less than 5%. Our main effort was to remove another kind of particle which was organic related materials, mainly including carbon （C） and oxygen （O） elernent. This kind of particle was from tape residue. In order to remove the tape residual, one step was added before the saw stage. Almost all of the tape residual was removed. Finally, the final yield was improved by more than 15%.
