Suite à la demande constante de réduire la taille des transistors et celle des dispositifs électroniques, guidée par la loi de Moore, l'intégration 2D n'est plus adaptée à cette demande croissante. Cela a conduit à l'intégration 3D des dispositifs actifs à l'aide de piliers Cu/Ni recouverts d'alliages de brasage à base de Sn. Dans les années à venir, les applications qui demandent des interconnexions à haute densité (optoélectroniques, microdisplays, les détecteurs IR, MEMS) nécessiteront l’utilisation de pas d'interconnexions inférieurs à 10 µm. Cependant, les piliers Cu/Ni/alliage de brasure base Sn n'ont jamais été étudiés en profondeur pour un si petit pas d'interconnexion. Avec la réduction de la dimension d'interconnexion, le diamètre des piliers Cu/Ni/alliage de brasure est réduit également. De ce fait, la formation des intermétalliques, qui joue un rôle primordial dans la bonne tenue de la jonction, peut poser des problèmes majeurs en raison de la réduction des dimensions du pilier de Cu et de l’alliage de brasure.Le travail de cette thèse est consacré à l'étude métallurgique et à la caractérisation d’interconnections de très petites dimensions (diamètre de 5 µm et pas de 10 µm) avec comme objectif principal l’étude des mécanismes physicochimiques des interactions entre les alliages de soudure Sn-Ag et les couches de Ni ou Ni/Au. Les mécanismes des interactions à l'état solide entre Sn et Ni ainsi que l’évolution du joint vers la transformation totale en intermétallique Ni3Sn4 (Solid-Liquid-Intercondiffusion - SLID processus), ont été étudiés pour la première fois dans de tels systèmes de dimensions micrométriques. De plus, les propriétés mécaniques et électriques ainsi que la stabilité thermique de ces interconnexions ont été étudiées. L’observation pour la première fois de la formation de la phase Ni3Sn2 à l’interface Ni/Sn à 200°C lors des vieillissements thermiques présente un intérêt pratique de grande importance. / With the constant demand for reducing the feature size of transistors and that of the devices, which is guided by Moore’s law, 2D integration is no longer fit to adapt the growing demand. This has led to 3D integration of active devices with the help of Cu/Ni pillars capped with Sn based solder alloys. In the coming years, applications which demand high density interconnects (optoelectronic, microdisplays, IR-detectors, MEMS) will require an interconnect pitch of 10 µm and below. However, Cu/Ni/solder pillars have never been investigated in depth for such a small interconnection pitch. With reduction of interconnect dimension, the diameter of Cu/Ni pillar and solder alloy also reduces. Thus, it is feared that the intermetallic formation, which is the key phenomenon responsible for the bonding, may be problematic due to the reduction in size of Cu pillar capped with solder alloy.The thesis is dedicated to the metallurgical study and its characterization for very small interconnects (5 µm) at 10 µm pitch, where the main focus is given on the physio-chemical mechanisms of soldering between Sn-Ag solder alloy and Ni or Ni/Au layers. For the first time, the mechanism of solid-state interactions between Sn and Ni is studied in depth and also for the first time the Ni3Sn4 SLID (Solid-Liquid-lnterdiffusion) system as an interconnect has been investigated at these dimensions. Moreover, the mechanical and electrical properties as well as the thermal stability of these interconnects are studied. Interestingly, during the latter part of this study, Ni3Sn2 layer is observed during aging of the Ni/Ni3Sn4 system for low temperature (200°C).

The problem of inaccurate measurement techniques for quantifying isothermal solidification kinetics during transient liquid phase (TLP) bonding in binary and ternary systems; and resulting uncertainty in the accuracy of analytical and numerical models has been addressed by the development of a new technique using differential scanning calorimetry (DSC). This has enabled characterization of the process kinetics in binary and ternary solid/liquid diffusion couples resulting in advancement of the fundamental theoretical understanding of the mechanics of isothermal solidification. The progress of isothermal solidification was determined by measuring the fraction of liquid remaining after an isothermal hold period of varying length. A 'TLP half sample', or a solid/liquid diffusion couple was setup in the sample crucible of a DSC enabling measurement of the heat flow relative to a reference crucible containing a mass of base metal. A comparison of the endotherm from melting of an interlayer with the exotherm from solidification of the residual liquid gives the fraction of liquid remaining. The Ag-Cu and Ag-Au-Cu systems were employed in this study. Metallurgical techniques were used to compliment the DSC results. The effects of sample geometry on the DSC trace have been characterized. The initial interlayer composition, the heating rate, the reference crucible contents, and the base metal coating must be considered in development of the experimental parameters. Furthermore, the effects of heat conduction into the base metal, baseline shift across the initial melting endotherm, and the exclusion of primary solidification upon cooling combine to systematically reduce the measured fraction of liquid remaining. These effects have been quantified using a modified temperature program, and corrected using a universal factor. A comparison of the experimental results with the predictions of various analytical solutions for isothermal solidification reveals that the moving interface solution can accurately predict the interface kinetics given accurate diffusion data. The DSC method has been used to quantify the process kinetics of isothermal solidification in a ternary alloy system, with results compared to a finite difference model for interface motion. The DSC results show a linear relationship between the interface position and the square root of the isothermal hold time. While the numerical simulations do not agree well with the experimental interface kinetics due to a lack of accurate thermodynamic data, the model does help develop an understanding of the isothermal solidification mechanics. Compositional shift at the solid/liquid interface has been measured experimentally and compared with predictions. The results show that the direction of tie-line shift can be predicted using numerical techniques. Furthermore, tie-line shift has been observed in the DSC results. This study has shown that DSC is an accurate and valuable tool in the development of parameters for processes employing isothermal solidification, such as TLP bonding.

The problem of inaccurate measurement techniques for quantifying isothermal solidification kinetics during transient liquid phase (TLP) bonding in binary and ternary systems; and resulting uncertainty in the accuracy of analytical and numerical models has been addressed by the development of a new technique using differential scanning calorimetry (DSC). This has enabled characterization of the process kinetics in binary and ternary solid/liquid diffusion couples resulting in advancement of the fundamental theoretical understanding of the mechanics of isothermal solidification. The progress of isothermal solidification was determined by measuring the fraction of liquid remaining after an isothermal hold period of varying length. A 'TLP half sample', or a solid/liquid diffusion couple was setup in the sample crucible of a DSC enabling measurement of the heat flow relative to a reference crucible containing a mass of base metal. A comparison of the endotherm from melting of an interlayer with the exotherm from solidification of the residual liquid gives the fraction of liquid remaining. The Ag-Cu and Ag-Au-Cu systems were employed in this study. Metallurgical techniques were used to compliment the DSC results. The effects of sample geometry on the DSC trace have been characterized. The initial interlayer composition, the heating rate, the reference crucible contents, and the base metal coating must be considered in development of the experimental parameters. Furthermore, the effects of heat conduction into the base metal, baseline shift across the initial melting endotherm, and the exclusion of primary solidification upon cooling combine to systematically reduce the measured fraction of liquid remaining. These effects have been quantified using a modified temperature program, and corrected using a universal factor. A comparison of the experimental results with the predictions of various analytical solutions for isothermal solidification reveals that the moving interface solution can accurately predict the interface kinetics given accurate diffusion data. The DSC method has been used to quantify the process kinetics of isothermal solidification in a ternary alloy system, with results compared to a finite difference model for interface motion. The DSC results show a linear relationship between the interface position and the square root of the isothermal hold time. While the numerical simulations do not agree well with the experimental interface kinetics due to a lack of accurate thermodynamic data, the model does help develop an understanding of the isothermal solidification mechanics. Compositional shift at the solid/liquid interface has been measured experimentally and compared with predictions. The results show that the direction of tie-line shift can be predicted using numerical techniques. Furthermore, tie-line shift has been observed in the DSC results. This study has shown that DSC is an accurate and valuable tool in the development of parameters for processes employing isothermal solidification, such as TLP bonding.