Interfacial degradation of epoxy coated silicon nitride

Silicon (Si) wafers passivated with a nitride film, fabricated by low-pressure chemical vapor deposition (LPCVD) and coated with epoxy were used as test specimens to characterize the interfacial degradation. For the sample preparation, the surface was examined by x-ray photoelectron spectrometry (XPS) and contact angle measurement. The stress test conditions were 121°C and 100% relative humidity. Optical microscopy was used to monitor the growth of damage as a function of time. Results of electron microscopy and XPS indicate that two different failure modes exist at the interface. One type of degradation is etching of the silicon nitride, Si₃N₄ (SiN) film, which appears in the early stages of aging. In more severely etched areas, the concentration of oxygen (O) is quite high and surface cracks develop. The other degradation mechanism is the nucleation and growth of blisters. Phenomenologically, their common characteristics were observed to be rapid growth initially followed by gradual maturing. When a blister became sufficiently large that it could no longer sustain its shape, it collapsed. Often additional blistering propagated subsequently. Blister growth can be accelerated by condensed water that oxidizes the SiN film. Adhesion failure, in the presence of oxygen, is propagated by the degradation of the epoxy silane and SiN film. Precipitate, consisting primarily of sodium (Na), carbon (C) and oxygen (O) with sodium being the main constituent, formed in many of the damage locations. XPS confirmed that a significant source of sodium was the epoxy. Depending on the environmental conditions, the morphology of the precipitate changed. The growth kinetics of the interfacial damage in terms of geometrical properties was observed and statistically analyzed. It was shown that a two-parameter Weibull cumulative distribution function (cdf) could be used to represent the blister growth at each time interval. Naturally, the total blister area increased as the total number of blisters increased. Furthermore, the mean blister area as a function of time was estimated quite well by a power function; however, the scatter about the mean was rather significant. In fact, the statistical variability was sufficiently high throughout that further inves- tigations are warranted. Although progress has been made toward an understanding of the growth kinetics of interfacial damage, the complex processes that compromise the mechanical and electrical integrity require further study.