Nano Topography Technologies for the Mass Production of Hybrid Bonding Process

Artificial intelligence (AI) semiconductors require ultra-high bandwidth and low power consumption, making three-dimensional (3D) stacking and high-density packaging indispensable. A key enabling technology for such architectures is wafer/chip-level hybrid bonding. Hybrid bonding is a face-to-face direct bonding process in which metal (Cu) and dielectric surfaces (e.g., SiO₂, SiN) with fine patterns are precisely aligned and then bonded through a sequence of surface pre-treatment and activation, low-temperature compression, and subsequent annealing. In this scheme, metal–metal and dielectric–dielectric interfaces are formed simultaneously. The bonding interface must satisfy nanometer-scale gap tolerances and extremely high planarity; otherwise, local roughness, Cu dishing, pattern erosion, and residual particles can induce voids and non-uniform contact, leading directly to degraded electrical connectivity, reliability concerns, and yield loss. Consequently, hybrid bonding processes demand not only accurate control of global wafer warpage but also nanometer-scale characterization of topography and roughness over Cu patterns and dielectric surfaces.  

Despite the rapidly growing demand for nanometer-precision metrology in hybrid bonding, current inspection technologies in mass production are typically limited to sparse sampling measurements on a small fraction of wafers or localized areas. A metrology solution capable of full-field inspection over the entire wafer area for all produced wafers is still lacking. In this study, we investigate and introduce surface-shape and warpage measurement approaches based on optical interferometry, atomic force microscopy (AFM), and deflectometry to address this gap. Through representative case studies, we demonstrate how these techniques can be applied to quantify Cu dishing, local and global flatness, and wafer/chip warpage with nanometer-level sensitivity. The proposed metrology framework aims to enable practical near-full or full-population inspection in hybrid bonding lines, thereby providing quantitative feedback to optimize chemical mechanical polishing (CMP) conditions, surface cleaning and activation processes, and pattern design. Ultimately, such comprehensive and precise surface metrology is expected to play a critical role in securing the high reliability and performance required for next-generation AI semiconductor packages.