Anodic bonding is a widely used process in research and industry for MEMS Applications. The main advantage is that it is a very reliable process, that results in a strong and hermetically sealed bond. The bonding process is also very stable to perform. These properties are due to the process sequence of heating the wafers, to be bonded, to a temperature in the range of 300 to 450°C. One of these two wafers has to be made of borosilicate glass (e.g., SCHOTT Borofloat33®). In such glasses, sodium ions (network modifier) are dissolved out of the glass network in the temperature range mentioned and thus become free and mobile. If now a negative voltage of typically 1000V is applied to the glass wafer, the free positively charged sodium ions will move away from the contact area of both wafers, the later bonding interface, under the influence of the electric field. This results in a displacement current, which decreases again after a maximum is reached, due to the limited available number of charge carriers (sodium ions). This ultimately leads to a depletion of the charge carriers in the glass. It is generally assumed that a linked effect occurs in the bond field, namely that oxygen ions migrate from the glass to the bond interface, where they cause the actual bond formation via anodic oxidation.Even though this basic concept of the anodic bonding corresponds to state of the knowledge of how the anodic bond is actually formed [1], it is difficult to prove in practice. In order to gain a deeper insight into the formation of anodic bonding and thus open up new applications and even wider utilization of anodic bonding, a first basic experiment was done as follows: A glass wafer was provided with a thin aluminum layer by a sputtering process and another glass wafer was anodically bonded to this aluminum layer. The typical bonding current (depletion and current drop) was observed and the resulting composite of two glass wafers with a very thin intermediate aluminum layer was examined and tested for bonding quality. First of all, it can be stated that bonding formation has taken place locally between the the 6" wafers. Unfortunately, particle contamination could not be completely avoided during processing leading to larger unbonded areas, especially in the center of the wafer. Nevertheless, the bonded areas allow for evaluation of bonding quality. The sputtered aluminum layer has a higher thickness in the center, due to the fact that the glass wafer was eccentrically rotated below a small sputter target. This inhomogeneity effect enables the characterization of different Aluminum layer thicknesses in one sample. The sputter time was set, that at the wafer center an untransparent Aluminum layer was deposited, while at the wafer edge just a few nanometers of aluminum forming a semitransparent layer. After the bonding experiment it was noticed that in all the bonded area the aluminum appears rather homogeneously transparent. According to that it can be concluded that the aluminum layer is oxidized during anodic bonding transforming into transparent Aluminum oxide. This finding is a clear confirmation of the general thesis, that during anodic bonding an anodic oxidation takes place at the bond interface. In further examinations the influence of the bonding parameters on this oxidation process will be investigated. It is expected that with higher temperature and bonding voltage the oxidation will be enhanced, ending up with an optically highly transparent layer of Aluminum oxide. This might be relevant for emerging optical applications requiring anodic bonding as reliable bonding process together with optical transparence for visual light. As blade tests have shown the bonding conditions in the first process are not yet optimal – the bonding strength was rather low, but darker spots in the crack opened area indicate a transfer of the aluminum oxide layer. By variating and improving bonding conditions, a deeper insight into the details of the oxidation process during anodic bonding will be available. So far it is known that easy to oxidize materials (Silicon and Aluminum) can be well anodically bonded while rather inert materials (Silicon Nitride) are hardly or not bondable with this procedure. For that reason, some materials like Platinum, Gold, and Titanium these predictions will be evaluated in practice. With this more profound knowledge it might be possible to predict the anodic bonding behavior of different materials in the future.Literature:[1] Lapadatu, A. C., & Jakobsen, H. (2015). Anodic bonding. In Handbook of Silicon Based MEMS Materials and Technologies (pp. 599-610). William Andrew Publishing. Figure 1
Read full abstract