Water-assisted sub-critical crack growth along an interface between polyimide passivation and epoxy underfill

Direct chip attach (DCA) microelectronic packaging technology is gaining prominence due to its numerous advantages. Delamination (debonding) of the underfill epoxy/ polyimide passivation interface of a DCA during hydro-thermal reliability testing has always been one of the salient problems. We have studied the water-assisted sub-critical crack growth along this interface and our measurement offers important clues as to the origins of the poor hydro-thermal testing results for these interfaces. A modified asymmetric double cantilever beam (ADCB) testing technique has been used to measure the sub-critical crack growth velocity v at various relative humidities and temperatures as a function of the crack driving force (strain energy release rate) G*. The presence of a significant partial pressure of water p(H2O) produces a marked decrease (by up to a factor of 12) in the threshold G* for crack growth at measurable velocities. Above the threshold log v rises linearly with rootG* but then enters a regime where the crack velocity (v = v*) is almost independent of rootG*. Finally, at the values of G* corresponding to rapid crack propagation in the absence of water, log v increases very rapidly with G*. By analogy to the classic work on water-assisted sub-critical crack growth in silica-based glasses, where very similar features are observed, we believe that the sub-critical crack growth along the polyimide-epoxy interface results from stress-assisted hydrolysis of primary covalent bonds, in our case ester bonds across the interface. The regime of rootG* just above the threshold corresponds to a physicochemical situation where the water activity (p(H2O)) at the crack tip is the same as that of the gaseous environment. In the regime where v = v* approximate to constant, the water activity at the crack tip is below that in the environment and the crack growth velocity is limited by the transport of water vapor to the bonds ahead of the crack tip. We develop a model of this crack growth following Wiederhorn 1967 that allows us to predict the sub-critical crack growth as a function of G* for arbitrary relative humidity and temperature conditions.