Leak Rates and Residual Gas Pressure in Cavities Sealed by Metal Thermo-Compression Bonding and Silicon Direct Bonding

Tuesday, 7 October 2014: 16:20
Expo Center, 1st Floor, Universal 9 (Moon Palace Resort)
K. Schjølberg-Henriksen (SINTEF), N. Malik (University of Oslo, SINTEF), A. Sandvand (Memscap, HBV), G. Kittilsland (SensoNor), and S. T. Moe (SINTEF)
Low, controlled, and stable gas pressures in sealed cavities are key features of wafer-level bonding processes used for hermetic sealing. The gas pressure is determined by e.g. the permeability of the seal and capping materials, gas produced during the bonding process, and device and material out-gassing. Both thermo-compression bonding with Au [1] and high-temperature silicon direct (fusion) bonding [2, 3] are used for hermetic sealing of commercial micro electro-mechanical (MEMS) devices today. Low-temperature silicon direct (plasma) bonding and thermo-compression bonding with Al are technologies with unverified hermeticity. This paper presents a comparison of the hermeticity of plasma and Al thermo-compression bonding and the hermeticity of fusion and Au thermo-compression bonding. 

Test cavities were realized by TMAH etching of 280 µm thick 6-inch silicon wafers with SiO2 mask. The cavities had a volume of 1.6 mm3and formed membranes with side edge 2.5 mm and 36 µm thickness. The cavities were bonded to silicon wafers by fusion bonding, plasma bonding, or thermo-compression bonding applying Al or Au. An overview of the bonded laminates and bonding parameters is found in Table 1. Wafers sealed by fusion and plasma bonding were laminated to blanket silicon wafers. Wafers sealed by Au thermo-compression bonding were laminated to silicon wafers with patterned Au bond frames of width 20 – 200 µm. Wafers sealed by Al thermo-compression bonding were laminated to silicon wafers with defined protruding silicon frames of width 20 – 200 µm that defined the bonding area. Figure 1 shows cross-sectional sketches of the four laminate types and the geometry of the cavities.

After bonding, the wafers were stored for a time ranging from 2 – 5 months. The membrane deflection was measured during this period by white light interferometry on 3 – 481 membranes. A maximum estimate for the leak rate was calculated. Individual dies underwent a Helium (He) fine leak test. The dies were exposed to 3 bar absolute pressure of He for 4 hours. Then the He leak rate was measured on individual dies within 30 minutes using an industry standard He leak detector. The detection limit of the measurement system was found to be 7 – 8×10-10 mbar×l×s-1, measured by leak testing of the empty test chamber.

The He leak rates measured on individual dies were not significantly different from the leak rate of the empty test chamber. This indicates that the measured He leak rate of all bonds were of the order of 10-10 mbar×l×s-1 or lower. The calculated maximum estimates for the leak rates based on deflection measurements are listed in Table 1. The average deflections are also listed. No significant change in membrane deflection was observed, giving a maximum leak rate for all die types of 2×10-11 mbar×l×s-1. The differences in estimated leak rates reflect the variations in laminate storage time only, and longer storage times may reveal that the maximum leak rates are even lower.

The presented results indicate that the hermeticity of plasma bonded seals is similar to the hermeticity of fusion bonded seals, and that the hermeticity of thermo-compression bonded seals applying Au and Al is similar. However, for many devices, true leakage rates below 10-15 mbar×l×s-1are required [3], and knowledge of residual gas composition may also be desired. Such measurements are underway and are expected to be presented in the full paper.


[1] G-S. Park et al. Electrochem Solid-State Letters 8(12) p.G330, 2005.

[2] S. Mack et al.J. Electrochem Soc 144(3) p.1106, 1997.

[3] K. Birkelund et al. Sensors and Actuators A 92 p. 16, 2001.