Atomistic-scale Simulations on Surface Activation Process of Dielectric Oxides for Hybrid Bonding Applications

Hybrid bonding has emerged as a promising 3D integration packaging technology for the next generation stacking devices, with the advantages of higher performance and smaller form factor over the conventional micro-bump interconnects. Despite the current development and understanding of such methodology so far, detailed mechanisms underlying the dielectric surface activation and bonding processes have yet been fully elucidated at the atomic level. Here, we present the full atomistic-scale simulation results that demonstrate surface activation process of dielectric oxide materials using reactive molecular dynamics (MD) simulation. We modeled a substrate material of silicon oxide $(\mathrm{S}\mathrm{i}\mathrm{O}_{x})$ and investigated the chemical and physical modifications introduced by N 2 plasma which is one of industrial-wise widely adopted gases for surface activation process. The plasma ion species and their impact energy along with flux data were collected from the plasma chamber simulation and were directly reflected into the MD framework. On the basis of simulation results, we first discuss the surface chemical state change of dielectric oxides after the plasma treatment. The degree of changes was quantified by evaluating the penetration depth, number density, and bond information (e.g. dangling, bridging bonds) which can potentially be correlated with the XPS (X-ray Photoelectron Spectroscopy) data. Another major finding which is a surface reconstruction mechanism is further discussed. We propose that ion bombardment plays an important role of breaking the chemical bonds, but at the same time, it physically reconstruct the surface to produce surface inactive sites. These sites hinder the formation of interfacial bonds during the hybrid bonding process thereby deteriorate the bonding strength of dielectric materials after the annealing process. The atomistic insight presented in this work can provide thorough understanding of plasma activated surface topology at the resolution scale that is often hard to reach with the experimental characterization techniques. In addition, the computational workflow represented here may provide a useful guideline for the surface modification process and overcome the time and expense resource constraints by supplementing empirical decisions made from the trial-and-error based full design of experiment endeavors.