Plasma activated bonding is one of promising candidates in wafer bonding field. In the past decade, much progress in terms of plasma generation, plasma treatment parameters, and annealing conditions has been achieved to allow high strength bonding at low temperatures with minimal defects. Since wafer bonding is often used as a middle process during device manufacturing, subsequent heated (or annealing) processes are inevitable for bonded wafers. However, void generation in annealing process (known as “annealing void”) is a serious problem in plasma activated bonding, especially for Si/Si wafer pairs. In particular, voids more frequently occur and are significantly severer as the wafer sizes further increase. Recently, our group has developed a fluorine activated bonding process enabling strong bonding of Si/Si bonded pairs (~2.2 J/m2 in surface energy) even at room temperature (~25ºC) without requiring annealing by introducing a small amount of carbon tetrafluoride (CF4) into oxygen plasma treatments. It seems that the fluorinated oxide allows water and hydrogen to diffuse out more easily from the interface at room temperature. But we have no idea how voids would be generated in elevated temperatures during subsequent annealing process. In this work, we demonstrate a quantitative evaluation of the void formation. As a comparison, void generation in the samples prepared by O2 plasma activated bonding is also systemically investigated. The void formation model is proposed to gain insight the mechanism of the fluorine containing plasma activated bonding. 8-inch silicon wafers (i.e. 200-mm diameters) were used in the experiments. To inspect void formation, a couple of bonded wafer pairs were prepared with different plasma treatment conditions. All the bonded wafer pairs were heated at 200, 300, 400, 500, 600, 700, and 800ºC for 8 h in ambient air. The heating steps do not aim to improve the bonding strength, but to investigate void formation at high temperatures. The IR imaging system was mainly used to characterize the voids in bonded pairs. To track the void emergence and growth, the digital image processing was employed to calculate void densities over the entire bonded wafer pairs. Water contact angle and x-ray photoelectron spectroscopy (XPS) measurements were made to investigate the plasma effects prior to bonding. In conclusion, we found adding fluorine to oxygen plasma could effectively mitigate void formation in subsequent annealing process. Moreover, the plasma treatment time is a vital parameter to optimize the bonding strength as well as the void generation. Long plasma treatment may produce more porous surfaces and introduce undetected microdefects. Consequently, a short fluorine containing plasma treatment (e.g. ~10 s) is suitable to remove organic contaminants, it is thus nearly void-free at the bonding interfaces during heating from 200 to 800ºC. In practice, an optimum treatment time depends on the application requirements. Figure 1
Read full abstract