Reductive CH₄ Plasma-Induced Modification of Cu Surfaces for Low-Temperature Hybrid Bonding

As artificial intelligence (AI) and high-performance computing demand greater parallel data processing, next-generation memory architectures must also evolve. High bandwidth memory (HBM) supports this with its 3-D-stacked structure and wide I/O interface, but further improvements in power efficiency and thermal management are still required. In addition, microbumps are currently used for chip stacking face scalability limitations. To address these challenges, hybrid copper (Cu) bonding has been proposed as a promising solution. This study presents a surface treatment strategy for low-temperature Cu-to-Cu direct bonding using CH₄-based reductive plasma: 1) native Cu oxides were removed and 2) re-oxidation was suppressed by forming a hydride carbon passivation layer. A design of experiments (DOEs) approach was used to optimize inductively coupled plasma (ICP) power, CH₄ flow rate, and pressure. Surface analysis was performed using X-ray photoelectron spectroscopy (XPS), TOF-SIMS, transmission electron microscopy (TEM), and sheet resistance measurements, while postbonding evaluation included TEM, SEM, and shear strength analysis. The results demonstrate that the energy of radicals and ions plays a decisive role in Cu oxide reduction, contributing to effective surface activation. The resulting passivation layer was confirmed to be a hydrogenated amorphous carbon (a-C:H) film rich in C, CH, and C₂H species. Bonding was conducted at 260 °C and 15 MPa for 1 h, followed by postannealing at 200 °C. TEM analysis revealed a void-free and oxygen-free bonding interface with the presence of carbon, suggesting that the carbon layer acted as a passivating layer. These findings confirm that reductuve CH₄ plasma treatment is highly effective for enabling low-temperature Cu bonding and highlight the strong potential of this approach for next-generation hybrid bonding technologies.