Key innovations in Advanced Packaging include the use of chiplets, improved heat dissipation techniques, enhanced miniaturization, material integration, and power optimization strategies. These technological advances are fundamental for scaling computing capabilities while maintaining the efficiency and reliability required by next-generation AI applications. Among these innovations, Hybrid Bonding has emerged as a pivotal technology. Hybrid Bonding facilitates superior AI performance by enabling efficient data transfer and reducing latency. This technology is critical for the seamless integration of various components, such as logic and memory, into a compact and high-performance system-on-chip (SoC) architecture. The ability of Hybrid Bonding to support high bandwidth and low latency interconnections is essential for handling the massive parallel processing workloads inherent in AI computations. Part one of this presentation will delve into the recent advancements in Hybrid Bonding, with a specific focus on the advantages of sputtered silicon nitride (SiN) and silicon carbon nitride (SiCN) for enhancing the performance of Hybrid Bonding. SiN and SiCN are recognized for their excellent dielectric properties, mechanical strength, and thermal stability, making them ideal candidates for use in Hybrid Bonding. Our recent investigations at Evatec have demonstrated that sputtered SiN and SiCN layers can significantly improve the reliability and performance of Hybrid Bonding interfaces, offering a robust solution for the demanding requirements of AI and HPC systems. Another critical aspect of wafer processes, particularly in Hybrid Bonding, is the flatness of the wafer. Process-induced warpage can severely impact the yield and performance of these processes. To address this challenge, the deposition of sacrificial layers that counteract warpage is essential. Reactive sputtering of materials such as Silicon Nitride (SiN) and Silicon OxyNitride (SiON) has emerged as a promising technique for creating these sacrificial layers. These layers serve to correct the warpage induced during various processing steps, thereby ensuring optimal wafer flatness and improving the overall yield of the Hybrid Bonding process. In the second part of this presentation, we will present our findings on the use of reactive sputtered SiN and SiON as sacrificial layers for warpage correction. Our studies reveal that these materials can be effectively used to achieve the desired wafer flatness, which is crucial for the success of Hybrid Bonding and other wafer-level processes. By fine-tuning the deposition parameters, we have been able to optimize the properties of these layers, thus enhancing the performance and yield of the overall process. In conclusion, the advancements in Hybrid Bonding, coupled with innovations in material engineering for wafer flatness correction, are key enablers for the continued evolution of AI and HPC technologies. The insights presented in this study underscore the importance of these developments and their potential to meet the growing computational demands of future AI applications.
ElGhazzali et al. (Wed,) studied this question.