Micro Electro-Mechanicals Systems (MEMS) include moving elements in some cases and they require a controlled atmosphere to operate. Hence they are protected by a cap assembled to the sensor by wafer-level packaging. For specific MEMS packaging, the process should provide strong and hermetic sealing walls of reduced size, less than 100 μm which also ensure electrical conductivity. If a high-temperature-resistant sealing (>200 °C) is also required, Al-Ge eutectic bonding is suitable. In eutectic bonding, the constituents of the alloy are deposited on at least one of the two wafers which are brought into contact. The melting and then the solidification of the solder close the interface to form an assembly. The Al-29.5at%Ge eutectic alloy, with a melting temperature of 424 °C [1], has been shown allowing strong and void-free bonds [2-4]. Nevertheless, during bonding, the eutectic liquid tends to squeeze out of the seal ring which can lead to void formation at the sealing interface and areas with unwanted metal. Moreover, influences of the underlayer on the bonding mechanisms have not been investigated whereas it is known to have an impact on Al/Ge bilayers [5]. In this paper we study the formation and behaviour of liquid Al-Ge alloy, its solidification, the resulting microstructures and the appearance of voids. The main investigated parameters are: - the underneath layer which is related to the wettability of liquid Al-Ge eutectic - the annealing temperature and cooling rate and their influence on solidified microstructure - the seal rings geometry The 200 mm silicon substrates used in this study are either thermally oxidized or covered by a 50nm-thick titanium nitride layer to prevent metal diffusion into silicon substrate. Blanket and patterned Al/Ge deposited wafers with 0.59 μm Ge/1 μm Al are used as single wafers and in bonded structures. The seal rings are 5 mm square with widths varying from 50 to 200 μm. Thermal treatment and bonding are performed under either vacuum (10-3 mbar) or inert gas (N2) at atmospheric pressure. Experimental temperature is varied from 400 °C to 500 °C with cooling rates varying from 1 °C/min to 20 °C/min. In order to evaluate the influence of the underneath layer on the behaviour of liquid Al-Ge eutectic alloy during bonding, the wetting of different substrates is investigated by dispensed drop method, under a vacuum of 5x10-7mbar. High purity aluminum and germanium (99.999 %) are used to prepare the alloy in an alumina crucible ending by a capillary to deposit a droplet on the studied surface. Observations are made in situ by camera and analyzed using Drop Shape Analysis software (Fig 1a). In addition, the melting and solidification process of many 1x1 cm2samples from blanket Al/Ge deposited wafers are studied by Differential Scanning Calorimetry (DSC) by using different cooling rates. The microstructures are then analyzed by optical microscopy, Secondary Electron Microscopy SEM and Energy Dispersive X-ray spectrometry (EDX). In some cases large dendrites are observed, which could be critical for a bonded structure (Fig 1b). Finally blanket and patterned Al/Ge deposited wafers are bonded to silicon wafers either thermally oxidized or covered by a titanium nitride layer. The influence of underneath layer, geometry (for patterned wafers) and cooling rates on squeeze-out and voiding phenomenon are characterized using Scanning Acoustic Microscopy SAM and FIB SEM (Fig 2). Image analysis is used to compare voiding densities between bonded samples (Fig 3). The results appear to be dependent on the process conditions. Underneath layer and cooling rate seem of great importance to control the squeeze-out and solidified microstructures in Al-Ge eutectic sealing and to determine the quality of the final assembly. Thanks to these characterizations and especially to the wettability measurements, mechanisms will be proposed to explain the squeeze-out result. REFERENCES [1] Rodney P. Elliott et al., Bulletin of Alloy Phase Diagrams, vol 1, n° 1, pp 65-68, 1980 [2] B.Vu et al., J. Vac. Sci. Technol.B, vol. 14, n° 4, pp. 2588-2594, 1996. [3] F.Crnogorac et al., J. Vac. Sci. Technol.B, vol. 30, n° 6, 06FK01, 2012. [4] S.Sood et al., Advancing Microelectronics, vol 41, n° 5, pp 30-37, 2014. [5] K. Nakazawa et al., Jpn. J. Appl. Phys., vol 53, n°4, 04EH01, 2014. Figure 1
Read full abstract