To increase the yield of the wafer-level Cu-Cu thermo-compression bonding method, certain surface pre-treatment methods for Cu are studied which can be exposed to the atmosphere before bonding. To inhibit re-oxidation under atmospheric conditions, the reduced pure Cu surface is treated by H₂/Ar plasma, NH₃ plasma and thiol solution, respectively, and is covered by Cu hydride, Cu nitride and a self-assembled monolayer (SAM) accordingly. A pair of the treated wafers is then bonded by the thermo-compression bonding method, and evaluated by the tensile test. Results show that the bond strengths of the wafers treated by NH₃ plasma and SAM are not sufficient due to the remaining surface protection layers such as Cu nitride and SAMs resulting from the pre-treatment. In contrast, the H₂/Ar plasma–treated wafer showed the same strength as the one with formic acid vapor treatment, even when exposed to the atmosphere for 30 min. In the thermal desorption spectroscopy (TDS) measurement of the H₂/Ar plasma–treated Cu sample, the total number of the detected H₂ was 3.1 times more than the citric acid–treated one. Results of the TDS measurement indicate that the modified Cu surface is terminated by chemisorbed hydrogen atoms, which leads to high bonding strength.

A novel surface activation technology for Cu-Cu bonding-based wafer-level vacuum packaging using hot-wire-generated atomic hydrogen treatment was developed. Vacuum sealing temperature at 300 °C was achieved by atomic hydrogen pre-treatment for Cu native oxide reduction, while 350 °C was needed by the conventional wet chemical oxide reduction procedure. A remote-type hot-wire tool was employed to minimize substrate overheating by thermal emission from the hot-wire. The maximum substrate temperature during the pre-treatment is lower than the temperature of Cu nano-grain re-crystallization, which enhances Cu atomic diffusion during the bonding process. Even after 24 h wafer storage in atmospheric conditions after atomic hydrogen irradiation, low-temperature vacuum sealing was achieved because surface hydrogen species grown by the atomic hydrogen treatment suppressed re-oxidation. Vacuum sealing yield, pressure in the sealed cavity and bonding shear strength by atomic hydrogen pre-treated Cu-Cu bonding are 90%, 5 kPa and 100 MPa, respectively, which are equivalent to conventional Cu-Cu bonding at higher temperature. Leak rate of the bonded device is less than 10⁻¹⁴ Pa m³ s⁻¹ order, which is applicable for practical use. The developed technology can contribute to low-temperature hermetic packaging.

The sealed cavity pressure of a micro-package often becomes the determining factor of the performance of an encapsulated micro-electro mechanical system (MEMS). In this study, a sealed cavity pressure evaluation method for a micro-package using a thin diaphragm is studied comprehensively. The sealed cavity pressure is identical to the surrounding pressure when the diaphragm separating both volumes becomes flat, i.e. zero-balance method. The zero-balance method has advantages to other techniques in terms of fabrication process simplicity and pressure evaluation range. Two approaches are introduced to reduce the impact of the residual stress, which typically exists in the conventionally used silicon-on-insulator (SOI) diaphragm. The introduction of a stress compensation layer was able to reduce the impact of the residual stress of SOI-based diaphragms. However, a significant hysteresis is developed on the diaphragm, which limits the precise measurement of a sealed cavity. On the other hand the residual stress can be avoided by the usage of intentionally tensile stressed boron doped Si diaphragm, which is produced by alkaline wet etching. Using such a diaphragm, a sealed cavity pressure of several Pa can be measured with a single digit Pa accuracy. [2020-0014]

This paper describes an event-driven tactile sensing system including 100 CMOS-MEMS integrated 3-axis force sensors of 2.7 mm square footprint for providing robot tactile sensation. The system has the following features: (1) the event-driven function of the CMOS LSI integrated in the sensor can reduce the number of stimulated sensors, (2) the stimulated sensors send data packets on a shared differential bus line of 90 cm by asynchronous serial communication, and (3) the data collision avoidance and adaptation functions of the sensor enable multiple sensors to be stimulated simultaneously. According to the event-driven function, the sensor sends data packets only when the digital sensing value is over a set threshold. The collision avoidance function detects data collisions, stops sending data, and avoids further collisions. The adaptation function varies the time interval between data packets and reduces the bus occupation by a specific sensor. These functions make the sensor autonomous to some extent that can judge whether and when to send data packets. The sensor in the system can operate at a data sending rate of 1 Mbps, 2 Mbps, 4 Mbps, or 8 Mbps by configurations. When 100 sensors with the same function settings are stimulated, almost the same amount of data packets are received from each sensor. As the data rate changing from 1 Mbps to 8 Mbps, the average sampling frequency of the 100 sensors is about 36.87 Hz, 73.00 Hz, 142.99 Hz, and 274.28 Hz with the data packet format of 244 bits.