Cette thèse porte sur l’étude des transistors de puissance à haute mobilité électronique en Nitrure de Gallium à enrichissement récemment apparus sur le marché. Ces travaux se focalisent sur l’analyse physique de la défaillance lors du vieillissement de ces composants sous des profils de mission réels de commutation. Un banc de stress a été conçu et construit, permettant d’imposer des conditions de fonctionnement réel sur le transistor sans utiliser des charges et donc en minimisant la consommation d’énergie. Une autre originalité de ce banc est qu’il contient des transistors témoin qui subissent une partie du stress imposé au transistor sous test, aidant ainsi à la compréhension des dégradations. Les composants vieillis sont caractérisés électriquement pour établir les modes de défaillance. Des méthodes de décapsulation de composants et de préparation d’échantillons mulit-échelle (mm au nm) ont été également développées afin d’accéder à la zone active des transistors. Par l’analyse de la microstructure des dispositifs, avec des techniques de photoémission spectrale, microscopie électronique à balayage et en transmission, les mécanismes de défaillance sont déterminés et la corrélation avec les modes dedéfaillance correspondants est réalisée. / This PhD Thesis is about the reliability of enhancement Gallium Nitride high electron mobility power transistors, recently on the market. This work is focused on physics of failure analysis on aged devices under real mission switching profiles. A novel stress bench able of stablish real operating conditions on a transistor without using a load, so minimizing energy consumption. Also, the circuit has two additional devices that undergo separate current and voltage stresses and are used for better understanding of degradations on main device under test. Electric characterization of aged devices allows determination of failure modes. Decapsulation and sample preparation procedures at different scales were developed to access the active zone of the devices and perform micro-structural analyses by the means of spectral photoemission, scanning and transmission electronic microscopies. Then, failure mechanisms are correlated with the corresponding modes.

The paradigm change in automotive power electronics towards wide bandgap semiconductor devices poses new challenges and requirements for the die-related packaging technologies as well as the assessment of reliability and lifetime. Here, the use of sintered silver for the die-related packaging in particular has proven promising. However, the empirical lifetime models for power modules developed over many years are not suitable any more. A holistic Physics-of-Failure approach can provide remedy as it allows for a significant reduction of testing time via finite element simulations. This approach requires a detailed understanding of the relevant failure mechanisms as well as an electrical, thermal and mechanical characterisation of the involved materials. A failure analysis of the complete power module revealed that the top-side sinter layer connecting the copper foil to the semiconductor die is prone to degradation. Therefore, the core of this work is the mechanical characterisation of porous sintered silver and, in particular, the primary and secondary creep behaviour. A newly developed creep model which - for the first time - takes load reversal for primary creep into account is implemented with a subroutine. This allows for lifetime simulations within a Physics-of-Failure framework resulting in a first lifetime model on module level for a complex automotive power module employing sintered silver.