Hydrophilic direct bonding is nowadays widely used in microelectronics for SOI substrate fabrication and 3D integration. Trapped water at the bonding interface plays a key role in the adhesion and adherence mechanisms [1]. Previous studies also reported that water could penetrate the bonding interface. This may modify the radial inhomogeneity of water [2]. However, this reported technique requires specific Silicon-to-Silicon bonding, lacks precision and needs additional characterizations (XRR, FTIR) to quantify the amount of trapped water at the bonding interface. In this work, we present an original methodology that can both accurately characterize and quantify water imbibition.Pulsed nanosecond laser annealing enables to reach extremely high temperatures, above the silicon melting point (1414°C), for very short durations (from a few tens of ns to a few hundred ns). These annealing conditions result in silicon surface melt. The beginning of melting is heterogeneous and leads to a particular surface topology with the formation of molten islands during laser annealing [3]. Recrystallization of these molten islands results in a localized increase of roughness (see fig. 1a). This roughness can exceed the direct bonding critical roughness. Non-bonded areas, i.e. bonding defects, can then be intentionally and precisely placed at the direct bonding interface.In a first step, bonding defects were thus formed along a wafer diameter with a 2 mm pitch. Then, using high-resolution acoustic microscopy, the bonding defects radii evolution was measured over time after direct bonding and without annealing (see fig. 1b). A progressive growth of defect radii was evidenced when getting closer and closer to the wafer edge. An additional water amount coming from the clean room atmosphere resulted in silicon oxidation at room temperature and di-hydrogen generation. Di-hydrogen was then trapped by bonding defects. In other words, bonding defects acted as humidity sensors. We could then record the penetration length and plot it as a function of the square root of time (fig. 1c). The penetration rate is the equivalent of a diffusion coefficient that could be calculated for both Silicon-to-Silicon and Silicon-to-Oxide bonding. We noted a very good agreement with Tedjini’s results for Silicon-to-Silicon bonding [2]. Meanwhile the calculated diffusion coefficient for Silicon-to-Oxide bonding was surprisingly about twice higher.Second, a bonding defect network with the same 2 mm pitch in both X and Y direction was created. After direct bonding, various duration 500°C temperature anneals were performed. The defects radii evolution was once again measured. As it increased over time, we could write the Griffith equilibrium between elastic (due to defect pressure) and surface (γ) energies. Describing defects as circular blisters, the amount of di-hydrogen trapped molecules could then be calculated. Plotting this number of H2 molecules, N, as a function of the annealing time (fig. 1d), we observe that a plateau was reached, meaning that all the di-hydrogen produced in the network was collected in the bonding defects. The water amount responsible for this di-hydrogen generation could then be evaluated and compared to data from other measurements. REFERENCES [1] F. Fournel, C. Martin-Cocher, D. Radisson, V. Larrey, E. Beche, C. Morales, P. A. Delean, F. Rieutord, and H. Moriceau, “Water Stress Corrosion in Bonded Structures”, ECS Journal of Solid State Science and Technology, 4 (5) P124-P130 (2015)[2] Tedjini M, Fournel F, Moriceau H, Larrey V, Landru D, Kononchuk O, et al. Interface water diffusion in silicon direct bonding. Appl Phys Lett. 12 sept 2016;109(11):111603.[3] L. Dagault, S. Kerdilès, P. Acosta Alba, J.-M. Hartmann, J.-P. Barnes, P. Gergaud, E. Scheid, and F. Cristiano, Investigation of Recrystallization and Stress Relaxation in Nanosecond Laser Annealed Si1−xGex/Si Epilayers, Appl. Surf. Sci. 527, 146752 (2020). Figure 1
Read full abstract