材料科学
互连
直接结合
化学机械平面化
堆积
引线键合
电迁移
纳米技术
抛光
半导体工业
三维集成电路
集成电路封装
可靠性(半导体)
微观结构
半导体
过程(计算)
热压连接
机械工程
表面改性
电子包装
共金键结
工程物理
半导体器件
炸薯条
粘结强度
表面完整性
扩散焊
倒装芯片
电子设备和系统的热管理
热的
微电子
过程开发
阳极连接
工作(物理)
混合动力系统
作者
Yueting Zheng,Hao Cui,Anyang Yu
标识
DOI:10.1016/j.jsamd.2026.101155
摘要
The semiconductor industry faces both opportunities and challenges in the post-Moore era. As artificial intelligence and high–performance computing advance, conventional 2D interconnects have become a performance bottleneck. Three–dimensional integration has thus emerged as a vital solution. Copper is the preferred interconnect material due to its low resistivity, excellent electromigration resistance, and cost–effectiveness. However, conventional thermocompression bonding requires high temperatures, posing thermal damage risks to sensitive devices. This has spurred the development of low–temperature hybrid bonding, which enables direct Cu–Cu and oxide–oxide connections without micro-bumps, permitting closer chip stacking and denser interconnects. This review outlines advancements in Cu–Cu bonding for 3D packaging. It subsequently presents various techniques–including surface pretreatment (e.g., plasma treatment, Chemical Mechanical Polishing (CMP) Process Optimization), hybrid bonding methods (e.g., surface activated bonding), and microstructure control (e.g., preferential (1 1 1) orientation)—that enhance low-temperature bonding performance. Key challenges and corresponding solutions are examined. Finally, the paper summarizes the critical technologies and processes essential for achieving high–quality, low-temperature hybrid bonding. • A comprehensive analysis is presented on surface pretreatment, microstructural control, and multi-process integration for hybrid bonding. • Emphasis is placed on (1 1 1)-oriented nanotwinned copper,which enables high- strength bonding at low temperatures while enhancing interfacial reliability. • This work provides critical technical support for cutting-edge applications such as HBM and AI chips.
科研通智能强力驱动
Strongly Powered by AbleSci AI