Soudure par interdiffusion dans les systèmes Au-In et In-Ni

Hermiticity is an essential requirement of a MEMS package for a device to work properly. Hermeticity also must be preserved during the operations following packaging, during which the package bond should not remelt. Bonding temperature is limited however to avoid damaging the device. Solid-Liquid InterDiffusion (SLID) bonding is interesting in this context, because this process allows one to achieve bonds that are able to subsequently withstand higher temperatures than the temperature at which the bonding is done. In this work, the Au-In and In-Ni binary systems have been studied because they are good candidates to realize such bonds. Indium melts at a relatively low temperature (157 °C), and all of the Au-In and In-Ni intermetallic compounds susceptible to form by interdiffusion have melting points above 400 °C. The optimum parameters for the bonding process depend strongly on the intermetallic compounds that are formed, in particular on their growth kinetics. To study the Au-In and In-Ni reactions from a more fundamental point of view, diffusion couples were fabricated. The intermetallic compounds formed in the temperature range of the bonding process have been identified, and their growth kinetics have been determined. In the Au-In system, the intermetallic compounds AuIn2, AuIn and Au7In3 were observed after heating to either 150 or 250 °C, independent of the reaction time. The majority of the reaction zone consisted of AuIn2, growing irregularly and mainly by diffusion of In. The AuIn phase was observed only in a thin layer. The third intermetallic compound, Au7In3, grows in a columnar morphology and more regularly, its interface with Au being almost planar. Judging by the pores visible in the Au7In3 phase near the Au interface, probably due to the Kirkendall effect, this layer grows mainly by diffusion of Au. In the In-Ni system, only the intermetallic compound In7Ni3 was observed after the sample was treated at 200 or 300 °C. This intermetallic compound grows in the form of fine columns, mainly by In diffusion. A latency period is observed before this growth starts, probably due to the presence of an oxide layer on the initial surface of the nickel. A front-tracking 1D computational model considering only the diffusion of In was developed in order to reproduce the growth of the intermetallic compounds observed. Diffusion in the Au and Ni phases is computed by discretization of Fick's second law. Because the intermetallic compounds are stoechiometric, and thus have fixed compositions, atomic fluxes were determined from the gradients of chemical potential. The velocities of the interfaces were then obtained by applying mass and solute conservation equations, in which the density differences between phases were taken into account. The physical parameters needed as input for the model are the diffusion coefficients and atomic mobilities of In in the various phases. These parameters are not known precisely, and thus they were treated as fitting parameters to reproduce the parabolic growth kinetics determined experimentally. Finally, the temperature dependence of the diffusion coefficients in the various phases was deduced. Bonding tests were also performed by joining glass substrates on which bonding frames had been deposited. The best Au-In bonds were obtained by treating at 200 °C for 15 min with an applied pressure of 2.3 MPa. When the nominal composition was near 60 at. % In, about 17% of the sealings were hermetic. The numerical model, developed for semi-infinite domains, was adapted to simulate the bonding process, i.e. in a system of finite size. The computations reproduce the experimental observations well. Thus, the model can be used with confidence to predict the optimal bonding temperature and time as a function of the initial Au and In thicknesses. The In-Ni bonding tests were not conclusive. Ternary Au-In-Ni bonds were also fabricated. Two ternary phases, each having different compositions and structures, were formed in this case. The hermeticity results obtained were promising. However, this ternary system is relatively unknown, and therefore it is difficult to understand the microstructures formed and their evolution.