Temperature and current are two major parameters that impact electromigration reliability. Due to the large current used in the accelerated electromigration test, the Joule selfheating associated with the stress current can be significant. The paper presents a study of electromigration fails in Pb-free interconnect from the point of view of localized Joule heating. The Joule heating effect in two types of packages, a fully assembled flip chip module with standard C4s and a silicon to silicon assembly with microbumps, is considered. A thermal FEM model is used as a guide to interpret the experimental observations.
IntroductionAs the EU and RoHS deadline on Pb-free solders draws close, the reported reliability of Sn based Pb-free solders has improved significantly. Advances in interconnect structure, materials and process have not only improved mechanical integrity of the large chip organic package, but also increased current carrying capability of Pb-free interconnects by more than an order of magnitude over the past decade [1-2]. Electromigration (EM) and thermal migration mechanisms of Pb-free solder have been proposed and studied [3-7], providing insight for further improvements. Since the EM driving force follows a power of the current density, reducing current density has one of the focal points for research. The introduction of thick UBM (under bump metallurgy) significantly reduced or eliminated current crowding [8][9], thus extending the current-carrying capability of Pb-free interconnect.Temperature is another important factor in the EM process, represented by the exponential factor in Black's Law. Due to the low melting temperature of Pb-free solders, the effect of temperature is particularly pronounced. Joule heating due to current crowding has been modeled and observed by IR microscopy [10][11]. Thermal migration due to temperature gradients within the solder joint has also been studied [12][13]. Although it is commonly accepted that the voids caused by mass transport, driven by electron wind, are the source for EM failure, the mechanism that leads to accumulation of voids which in turn leads to catastrophic failure has not been established. If the resistance change is mainly due to void formation, the voiding area needs to be over 90% of the conduction area before the change in resistance becomes observable. In this paper, an example of the correlation between measured resistance changes and bump temperature is shown. The EM test results from two types of test vehicles are compared, and the relationship between Joule heating, heat removal rate, and ultimate EM-induced failure is discussed. A thermal finite element model (FEM) model is used as a guide to interpret the experiments observations. An electro-thermal model will be onsidered in a separate study.