Intel and CEA-Leti accelerate D2W bonding

Intel and CEA-Leti have optimized the hybrid direct bonding, self-organizing process for D2W (Die-to-Wafer) bonding. This can improve alignment accuracy and manufacturing throughput by thousands of dies per hour.

This approach uses the capillary force of water droplets to align the die on the target wafer.

The D2W hybrid bonding process is considered essential to the success of future memory, HPC, and photonic devices, but it is much more complex, less accurate, and die-assembled than bonding between wafers. The throughput is low.

CEA-Leti has been developing self-organizing techniques for several years with the goal of significantly improving throughput and placement accuracy.

“Commercial-scale throughput from D2W self-organization has two major challenges associated with die handling,” said Emilie Bourjot of CEA-Leti. “Combining the self-organizing process with the pick-and-place tool can improve the throughput as follows: Fine alignment is performed by the droplets, which reduces alignment time. Combined with a collective die handling solution, all dies are coupled at the same time without any precise placement along the process flow, resulting in higher throughput. “

Process optimization is also an important part of this task to increase process maturity and target industrial requirements. “With this alignment and throughput performance, it’s definitely a promising step to ally the magic of physics with a simple drop of water,” says Bourjot.

The research paper states that “capillary force arises from the principle of surface minimization and, in the case of liquids, is acted upon by surface tension.” From a macroscopic point of view, liquids tend to minimize their liquid / air interface and reach equilibrium with minimized energy. This mechanism allows the die to self-align at the binding site.

The liquid selected as the rearrangement vector must exhibit high surface tension and must be compatible with direct coupling. With the exception of water, which has a surface tension of 72.1 mN / m, most liquids have a surface tension of 20-50 mN / m, making them a good candidate for a self-assembling process using hydrophilic bonds where water is already important. Will be. Mechanism parameters. “

According to the paper, “water-dispensing techniques and surface treatments to regulate surface hydrophilicity seemed to be important for the proper implementation of the self-assembling process.” “In this way, excellent alignment performance was achieved with a homemade collective self-assembled bonding bench. As a result, the average misalignment was less than 150 nm and 3σ was less than 500 nm. Finally. , Self-assembly process and compatibility with a wide range of die dimensions (8 x 8 mm², 2.7 x 2.7 mm)21.3 × 11.8mm2 And 2.2 x 11.8 mm2) Has been demonstrated. “

By comparison, the state-of-the-art alignment is 1 µm for pick-and-place tool post-bonding and the best case is 700 nm, but the self-alignment process allows alignment for post-bonding below 500 nm and even less than 200 nm. ..

CEA-Leti describes the “homemade collective self-organizing bonding bench” and states: “Because there are no industrial tools for a self-organizing approach, the team has created a unique lab bench that enables collective self-organization. Non-reproducible manual process control is still aligned below 500 nm. This strongly suggests that this process-specific industrial tool provides higher reproducibility, robustness, and accuracy. “

The conclusions of this paper emphasize this point. Despite these breakthroughs, “many aspects of self-organization need to be investigated, and significant improvements can only be made if the tool supplier develops a conforming tool to automate this process. Become.”

For this collaboration, CEA-Leti designed the process flow and used bonding physics, process, and process integration expertise to perform wafer processing and self-organizing bonding. We also evaluated the characteristics of nanotopography, scanning acoustic microscope, alignment, etc. Intel’s participation included providing specification, modeling, and pre- and post-bonding process integration expertise to be compatible with the foundries of the self-organizing process.

https://www.electronicsweekly.com/news/business/intel-cea-leti-accelerate-d2w-bonding-2022-06/ Intel and CEA-Leti accelerate D2W bonding

Back to top button