Molecular Dynamics Study on Plasma-Surface Interactions of SiCN Dielectrics for Wafer-to-Wafer Hybrid Bonding Process

Abstract

Recently, Cu/dielectric hybrid bonding process is receiving a significant amount of attention as a novel technology to three-dimensionally (3D) integrate next generation devices of fine pitch (sub- $\mu\mathrm{m}$) interconnects. Among various dielectric material candidates for hybrid bonding, SiCN has been greatly considered due to its applicabilty as Cu diffusion barrier layer and reliable mechanical strength that endures chemical mechanical planarization (CMP) process. While the purpose of hybrid bonding is to establish a robust connection between the metal pads, dielectric surface layer plays a crucial role in providing reliable bonding strength to sustain Cu pads until adequate grain growths are achieved during the thermal annealing process. Thus, understanding the plasma-surface interactions of the dielectrics is essential to elucidate the surface activation mechanisms that consequently leads to the bonding quality. In this work, we present a new approach of investigating the plasma-surface interactions using atomic-scale simulation. We established a procedure to derive SiCN surface model based on the surface information and generated three different surface models according to the C/N atomic composition ratio of 0.5, 1 and 2. A series of molecular dynamics (MD) simulations that imitates plasma treatment process were performed on these surface structures to determine the state of O2 plasma activated surfaces and to figure out if there are any correlation between atomic compositions of SiCN dielectrics with the degree of surface activation, which in this work quantified by the surface areal density of SiOH. Another sets of MD simulations were performed to propose a method for promoting surface hydroxylation, which was to include OH species during the plasma treatment process step. Our simulation results indicate that surface activation can successfully be facilitated by the implication of OH species without deteriorating the surface roughness which is also an important surface feature that is known to affect the overall bonding quality. The atomistic insight we presented in this work can provide thorough understanding of SiCN dielectric surface activation process. In addition, the computational approach and procedures we proposed could be an effective measure to examine novel concepts for process development without the time and expense constraints of trial-and-error based experimental approach.