Wafer bonding is one of the key technologies for the 3D integrations, and 300-mm-wafer-level bonding is applied for image sensors, memories, and so on. Fusion bonding using plasma activation as wafer-level bonding process for such devices should require high-yield and high-reliabilities, therefore the interfacial analyses are significant. There are several interfacial analyses for bonded interfaces such as scanning acoustic microscopy (SAM), TEM and so on. These methods are indispensable in the evaluation of the bonded interfaces, therefore, the specimens would be affected by the measurement preparations or circumstances. Generally, SAT images of the interface of the bonded wafers show voids among the interface caused by redundant hydrophilization, and interfacial structure can be observed by cross-sectional TEM. However, the cross-sectional TEM image of the interfacial structure of dioxide was difficult to analyze, because the contrast of the reacted (or water contained) layer is very weak. As the reaction occurs at the interface of substrates and analyzing targets are light elements (H2O, -OH) and other deuterated substitutes. Therefore, in previous study, we demonstrated the interfacial analysis of wafer-level bonded dioxide wafers by neutron reflectivity (NR), and we confirmed that NR would have possibilities to distinguish conditions of bonded interfaces.[1] In this research, 300 mm Si wafers are prepared and they are oxidized by wet thermal oxidation at 900ºC and a 100-nm-thick SiO2 layer is deposited. Some of dioxide wafers were polished by chemical mechanical polishing before introduced to a bonding apparatus, and other wafers were introduced as prepared. In the bonding apparatus, first, wafers are activated by plasma, that plasma activation is optimized as a sequential plasma process same as our previous work[2]. Activated wafers are cleaned with megasonic water, then they are transferred and bonded in relatively high vacuum (20 Pa), or are transferred in N2 ambient under atomic pressure and bonded in relatively low vacuum (1000 Pa). These wafers are bonded with a 1000 N load. After that, the bonded wafers are annealed at 200 ºC for 7h in an infrared (IR) image furnace within 3-5 hrs (0 days) or after c.a. 2 days storage in the air. (Table 1)The bonded wafers are evaluated by blade test for bonded strength measurement and NR. NR measurements were performed at BL17 SHARAKU, which is the vertical type time-of-flight neutron reflectometer installed in Japan Proton Accelerator Research Complex (J-PARC). Neutron pulses of 25 Hz are generated with a wavelength band of 0.2 - 0.88 nm. Neutrons penetrate into the bonded interface from the side with incident angles of 0.2º, 0.7º and 1.6º. The obtained NR data was analyzed by Motofit program within the IGOR software package and the analytical program developed for NR data to evaluate neutron scattering length density (SLD) profiles in the film thickness direction of interfacial structures.The results are shown in Table 1. Bonding strength was increased under condition of N2 ambient. And there are a few improvements of bonding strength by storing before post annealing. NR results revealed interfacial structures. Bonded interfaces under low-vacuum conditions have thicker and higher dense interfaces than bonded interfaces under high-vacuum conditions. It seems that the interfacial structures are affected by the bonding circumstances.At the conference, we would like to explain these bonded interfacial structures more precisely with other analyses.[1] M. Fujino, et. al., WaferBond’19, Halle, Germany, Dec. 2019, p.45[2] M. Fujino, et. Al., Jpn. J. Appl. Phys., 59, SBBA02, 2020 Figure 1
Read full abstract