Predicted Device-Degradation Failure-Rate 2015-01-2555

There is a concern that the continuing trend on miniaturization (Moore's law) in IC design and fabrication might have a negative impact on the device reliability. To understand and to possibly quantify the physics underlying this concern and phenomenon, it is natural to proceed from the experimental bathtub curve (BTC) - reliability “passport” of the device. This curve reflects the combined effect of two major irreversible governing processes: statistics-related mass-production process that results in a decreasing failure rate with time, and reliability-physics-related degradation (aging) process that leads to an increasing failure rate. It is the latter process that is of major concern of a device designer and manufacturer.

The statistical process can be evaluated theoretically, using a rather simple predictive model. Owing to that and assuming that the two processes of interest are statistically independent one can assess the failure rates associated with the aging process from the BTC data by simply subtracting the predicted ordinates of the statistical failure rates (SFR) from the BTC ordinates. The objective of this analysis is to show how this could be done.

The suggested methodology proceeds from the concepts that the actual (“instantaneous”) SFR is a random variable with a known (assumed, established) probability distribution, that the experimental BTC can be represented by its infant mortality and the wear-out portions only (the steady-state portion in this case is simply the boundary between the infant mortality and wear-out portions) and that the two BTC portions considered can be approximated analytically. The cases, when the “instantaneous” SFR is distributed normally and in accordance with the Rayleigh law are used as suitable illustrations of the general concept.

The developed methodology can be employed when there is a need to better understand the relative roles of the statistics-related and physics-of-failure-related processes in reliability evaluations of electronic products. The methodology can be used also beyond the field of IC engineering, when there is a need to understand and, hence, to separate the roles of the two irreversible processes in question.

One of the major challenges of the future work is to determine the probability distributions of the actual (“instantaneous”) SFRs for particular products and applications.