Archive TOULOUSE Archive Ouverte ( OATAO )

Electronic power modules devices are paramount components in the aeronautical, automotive and military applications. The solder layers are the most critical parts of the module and are usually subjected in their whole life to complex loading conditions. To improve the design task, realistic thermoelastoviscoplastic and lifetime prediction models which can describe efficiently the deformation-damage of the electrical device must be chosen carefully. Some of the most common behavior models are based on the separation between creep and plasticity deformations such as power law, Garofalo, Darveaux... So, to take into account the creep-plasticity interaction, the thermal cycling as well as the hardening-softening effects, unified viscoplastic models are increasingly being used to describe more efficiently the physical state of the material. We propose in this framework a survey of some unified viscoplastic models used in the electronic applications for the viscoplastic modeling of the solder as well as creep-fatigue life prediction rules. The models are used for the characterization of a SnAgCu solder and are briefly compared within tensile, creep data and stabilized responses. Introduction Nowadays, electronic power modules give promising perspectives for entirely electric engines with less fuel-dependent systems in aeronautic industry. Typical power module is composed of three major components: a semi-conductor chip which may be an IGBT and/or a diode, a metalized ceramic substrate and a base plate. These components are soldered to each other’s and positioned over a heat sink for thermal management (Fig .1). The dielectric ceramic substrate is double bonded with thick copper or aluminum metallization for improving heat spreading. The base plate is made also of high thermally conductive material such as AlSiC, copper or aluminum. The brazing is achieved using solder alloys considered as one of the principal failure causes in the module due to its weaker thermomechanical properties as compared to the other constitutive materials. So, regardless of their advantages, the design of the electronic power module devices represents a great challenge due to the extreme environmental and operational conditions which involve creep, fatigue as well as creep-fatigue interaction and high thermal variation. The high reliability of such components is then related to the optimized choice of the constitutive materials of the devices which consists basically on the microstructure properties, material compatibility and thermomechanical characteristics. Once the components materials are chosen, the experimental data becomes necessary for the modeling task in order to reproduce the material behavior in its real environmental and assembled state. As indicated earlier, the solder alloy is subjected to a viscoplastic deformation due to high temperature levels; the solder behavior must be described by appropriate viscoplastic models and lifetime prediction rules. ha l-0 07 38 76 4, v er si on 1 5 O ct 2 01 2 Fig. 1: Internal architecture of a typical electronic power module. There are in the literature a wide range of models which can be classified within the formulation nature of the constitutive equations i.e. coupled or uncoupled deformations. Uncoupled models are based on the summation between the plastic and creep deformations considering that there is no interaction between them [1]. The plastic deformation follows generally an isotropic-plasticity as Ramberg-Osgood model [2] or also a Drucker-Prager plasticity model [3]. The creep deformation evolves according to a creep model such as Norton [4], Garofalo [5], Darveaux [6],...The uncoupled deformation models are characterized by a simple formulation so they can be easily integrated and identified. In the other side, work-hardened solders show creep-plasticity interaction especially in the case of cyclic loading and thermal variations i.e. high-to-low or low-to-high temperatures. In this case, some kinds of unified thermoelastoviscoplastic models which consider that the plasticity and creep effects may be represented by only one time-dependent deformation are suitable for the modeling under complex loading conditions. The concept of state variables is also introduced in these models [7] in order to reproduce some other physical phenomena such as isotropic or kinematic hardening, dynamic and thermal static recovery, ratcheting, aging,... In despite of their complex formulation based on stiff, highly coupled first order differential equations, unified models are frequently used and are more integrated in the finite element codes. Anand’s unified viscoplastic model is usually employed for the deformation modeling of the solder alloys [810]. It’s integrated in a uniaxial form and has a unique scalar state variable which represents the resistance to plastic deformation. However, Bauschinger effects and cyclic hardening couldn’t be well described [11]. Other authors such as Krempl, tried to incorporate on its own overstress-based viscoplastic model a kinematic state variable. Busso et al, introduced a stress-dependent Arrhenius term to describe more efficiently the thermal variation effects [12]. Basaran and Chandaroy formulate a dependency of the material viscoplastic behavior with its microstructure changes by introducing a grain size variable [13]. McDowell et al, included in their model temperature rate terms in both isotropic and kinematic hardening variables for the thermomechanical cycling [14]. Thermodynamical restrictions are also demonstrated for thermoplasticity and thermoelastoviscoplasticity cases [15]. Chaboche’s used a power law term to describe the plastic flow and may contains several kinematic and isotropic hardening variables as well as a plastic strain memory for the saturation state [16]. The model proved to be thermodynamically consistent. McDowell and Chaboche’s viscoplastic models appear to include the most features that the solder joints material can exhibit and to match well with the thermomechanical responses of several materials references [17, 18]. We propose in this framework, (i) a description of the formulation of these unified models (ii) an integration of the first order differential constitutive equations and its implementation in a FE code ABAQUS (iii) a comparison of the viscoplastic models compared to the experimental data and the numerical simulations. Formulation of the viscoplastic constitutive equations The unification of the inelastic deformations and the description of several physical material states are inducing in nature a huge number of material parameters and a coupled set of stiff and first order Chip (Diode or IGBT) Metallization s Solder alloys Base plate Heat sink Wire bonding Insulant ceramic substrate ha l-0 07 38 76 4, v er si on 1 5 O ct 2 01 2 differential equations which couldn’t be integrated without accurate and stable numerical methods. It’s the case for the McDowell and Chaboche’s viscoplastic models. McDowell viscoplastic model. Contrary to Anand’s model based on Garofalo hyperbolic sine function, McDowell’s model contains a Zener-Hollowman exponential parameter with respect to the power law term [19]. The temperature dependency of the model is written as a function of the Arrhenius term in the flow law. Flow Law The inelastic strain rate is determined by the following rate equation:     2 3 , , 2 ij ij in ij v ij ij s A F S D J          (1) With,   2 v ij ij S J R      (2) Where       ij ij ij ij ij ij s s J         : 3 2 2 denotes the second invariant of the reduced stress   ij ij s   . ij s is the deviatoric stress tensor, ij  the backstress tensor, and  is a thermal diffusivity parameter expressed as an Arrhenius function. A is a temperature independent material parameter.