半导体晶圆键合加压精确控制方法  

Method for Precise Control of Semiconductor Wafer Bonding Pressure

作　　者：张慧[1] 周字涛 鲁统伟[2] 王成君 张辉[1] 李早阳 张志胜[1] ZHANG Hui;ZHOU Zitao;LU Tongwei;WANG Chengjun;ZHANG Hui;LI Zaoyang;ZHANG Zhisheng(Southeast University,Nanjing 211189,China;Hubei Key Laboratory of Intelligent Robot(Wuhan Institute of Technology),Wuhan 430205,China;The 2nd Research Institute of CETC,Taiyuan 030024,China;Xi'an Jiaotong University,Xi'an 710049,China)

机构地区：[1]东南大学,南京211189 [2]智能机器人湖北省重点实验室(武汉工程大学),武汉430200 [3]中国电子科技集团公司第二研究所,太原030024 [4]西安交通大学,西安710049

出　　处：《电子工艺技术》2024年第6期46-49,共4页Electronics Process Technology

基　　金：国家重点研发计划项目(2022YFB3404300);智能机器人湖北省重点实验室基金项目(HBIR202303)。

摘　　要：晶圆键合质量受键合压力控制精度显著影响。对于本身有一定翘曲度的晶圆,容易在键合过程中产生非预期位移,造成键合工艺偏差,甚至碎片。因此,提出了一种基于模糊PID(比例-积分-微分)和RBF(Radial Basis Function)径向基函数神经网络算法的半导体晶圆键合压力控制方法。由于加入了模糊PID和深度学习算法控制,使得键合工艺过程响应更迅速,控制更精准,且安全性更高。采用铝样片进行加压试验,结果表明RBF神经网络模糊PID控制算法能够满足压力精度为±1%的控制需求。The bonding pressure control precision have an important influence on the bonding quality of wafers.The wafer with a certain warping degree is easy to produce unexpected displacement during the bonding process,which results in the deviation,and even fragmentation.A semiconductor wafer bonding force-control method is proposed based on fuzzy PID (proportional integral-differential) and RBF (Radial Basis Function) neural network algorithms.Due to the addition of fuzzy PID and deep learning algorithm control,the bonding process response is faster and more accurate,and the safety is higher.The force-testing experiments are done by using the aluminum samples.The results show that the RBF neural network fuzzy PID control algorithm can meet the control requirements of force accuracy with±1%.

关 键 词：晶圆键合压力 RBF神经网络 模糊PID控制 压力控制精度 

分 类 号：TN605[电子电信—电路与系统]