Reliability physics underlying thermal stress failures in electronic, optoelectronic and photonic structures is discussed, with an emphasis on the roles of predictive modeling and failure-oriented-accelerated testing (FOAT). The following major topics are addressed: 1) Bi-material and die-substrate assemblies; 2) Analytical modeling vs. finite-element analysis (FEA); 3) Thermal stress failures in solder joints; 4) Global and local thermal mismatch stresses; 4) Thermal stress in assemblies bonded at the ends; 5) Thermal stress relief in assemblies with low modulus adhesives at the ends; 6) Design recommendations for reducing thermal stresses; 7) Thermally matched assemblies; 8) Thermal stress in thin films; 9) Thermal stress in plastic packages of IC devices; 10) Bow-free assemblies; 11) Probabilistic approach; 12) Thermal stress in optical fibers; 13) Thermal stress in nano-systems. It is concluded that physically meaningful and easy-to-use modeling, and especially analytical modeling, is an important tool for the prediction and prevention of thermal stress falues and electronics, opto-electronics and photonics systems.

Drop tests are often substituted in qualification or life testing of microelectronic and optoelectronic products by shock tests. The existing (e.g., Telcordia) qualification specifications require that a short term load of the given magnitude and duration (say, an “external” acceleration with the maximum value of 500 g, acting for 0.001 s) is applied to the support structure of the product under test. The natural frequencies of vibration are not taken into account. The objective of our study is to develop simple analytical (“mathematical”) predictive models for the evaluation of the dynamic response of a structural element in a microelectronic or an optoelectronic product/package to an impact load occurring as a result of drop or shock tests. We use the developed models to find out if a shock tester could be “tuned” in such a way that the shock tests adequately mimic drop test conditions. We suggest that the maximum induced curvature and the maximum induced acceleration be used as suitable characteristics of the dynamic response of a structural element to an impact load. Indeed, the maximum curvatures determine the level of the bending stresses, and the maximum accelerations are supposedly responsible for the functional (electronic or photonic) performance of the product. We use the case of an elongated rectangular simply supported plate as an illustration of the suggested concept. We show that in order to adequately mimic drop test conditions, the shock test loading should be as close as possible to an instantaneous impulse, and that the duration of the shock load should be established based on the lowest (fundamental) natural frequency of vibrations. We show also that, for practical purposes, it is sufficient to consider the fundamental mode of vibrations only, and that the shock load does not have to be shorter than, say, half the quarter of the fundamental period. We demonstrate that, if the loading is not short enough, the induced curvatures and accelerations can exceed significantly the curvatures and accelerations in drop test conditions. Certainly, the results of such shock tests will be misleading. After the appropriate duration of the shock impulse is established, the time dependence and the maximum value of the imposed (“external”) acceleration in shock tests should be determined, depending on the most likely drop height, in order to adequately mimic drop test conditions. We demonstrate that the application of a probabilistic approach can be helpful in understanding the mechanical behavior and to ensure high short- and long-term reliability of an electronic or photonic device that might be or will be subjected to an accidental or expected impact loading. We conclude that although it is possible to “tune” the shock tester, so that the drop test conditions are adequately reproduced, actual drop tests should be conducted, whenever possible. The results of the analysis can be helpful in physical design and qualification testing of microelectronic and photonic products, experiencing dynamic loads of short duration.

In this paper it is shown that the bathtub-curve (BTC) based time-derivative of the failure rate at the initial moment of time can be considered as a suitable criterion of whether burn-in testing (BIT) should or does not have to be conducted. It is also shown that the above criterion is, in effect, the variance of the random statistical failure rate (SFR) of the mass-produced components that the product manufacturer received from numerous vendors, whose commitments to reliability were unknown, and their random SFR might vary therefore in a very wide range, from zero to infinity. A formula for the non-random SFR of a product comprised of mass-produced components with random SFRs was derived, and a solution for the case of the normally distributed random SFR was obtained.