Plasma-activated fusion bonding for vacuum encapsulation of microdevices

Joel Soman, Louisiana Tech University


A fabrication process for vacuum-encapsulating PZT microcantilevers was designed in this dissertation. Initially, a low temperature wafer-bonding recipe was optimized with the help of plasma-activation. Conventional direct fusion bonding temperature was reduced from 400°C to 85°C, and final thermal annealing temperature and time of 1000°C for 4 hours (hr) were significantly reduced to 300°C and 1 hr respectively. Tensile tests conducted on dies diced from the bonded wafer stack revealed bond strengths of 22.15 MPa, which was close to the bulk fracture strength of 24 MPa for silicon. Near infrared images of the wafer stack showed no debonded regions at the interface. Surface and interface chemistry of oxygen plasma-activated wafers before, during, and after bonding were investigated. Significance of wet chemical activation technique, like RCA (Radio Corporation of America) cleaning, was studied. The time interval between plasma-activation and fusion bonding was varied, and its effect on the bond quality and bond strength was investigated. Decrease in the bond-quality and strength was observed with an increase in storage time. However, an unexpected increase in the bond quality was observed after 48 hr, and was attributed to the increase in the interfacial oxide layer. Further investigations revealed that the interfacial oxide layer was capable of absorbing gas molecules released as a byproduct of ongoing reactions at the interface of the two wafers. Gettering capability of the interfacial oxide layer was confirmed through the bonding of plasma-activated and 48 hr stored silicon (Si) and silicon dioxide (SiO2) wafers. Infrared images showed a good bond for the wafer stack.

Since designing a fabrication process flow for vacuum-encapsulation of microdevices was the primary objective, lead zirconate titanate microcantilevers were fabricated onto a silicon substrate. The microdevices were actuated in ambient air pressure as well as in a vacuum environment. Broadening of the resonance curve was observed with an increase in the magnitude of ambient pressure, and is a result of increased air-damping. Experimental results obtained were compared to theoretical results from finite element modeling analyses.

Vacuum cavities were fabricated between two Si wafers. Optical lid-deflection method of measuring internal cavity pressure was explored and employed with the help of high aspect ratio pressure diaphragms on a capping wafer. An investigation of seal integrity of the vacuum package revealed real/virtual leaks. The gettering capability of the SiO2 layer was employed in order to preserve the vacuum-level in the cavities. Two types of gettering patterns were investigated. It was concluded that an SiO2 getter layer at the interface increased the seal-integrity of the vacuum packages, while getter rings still showed signs of real leaks. In addition, it was observed that the internal vacuum-level was higher for cavities with getter rings as compared to cavities without getters. It was concluded that getter rings were capable of preventing virtual leaks but not real leaks. A thick interfacial getter layer, however, prevented both the real and virtual leaks.

Finally, a vacuum-packaging fabrication method to encapsulate lead zirconate titanate microcantilevers was proposed. In addition, more accurate methods of measuring package vacuum pressure magnitudes were proposed.