Make Use of Rbd to Distinguish Different Failure Modes

摘要

Gate oxide integrity (GOI) plays an important role in device reliability especially for modern products where high speed and memory capacities are in demand. There are many test methods (e.g. Qbd, Vbd, and TDDB) to ensure gate oxide reliability. After the tests, we usually analyze data by a Weibull or lognormal plot to identify failure mechanisms and to ensure the failures are within the acceptable range. However, in recent years, more and more S-shaped curves are observed especially from a Weibull plot. The S-shaped curve in the Weibull plot may be derived due to more than one failure mode. It is crucial in practice to estimate the lifetime, the percentage, and the failure behavior (i.e., to see if we are dealing with a DFR, Decreasing Failure Rate, or an IFR, Increasing Failure Rate, failure) of Mode A and B, the early and random failures, respectively. In this paper, a methodology to distinguish different failure modes is introduced using the post-breakdown resistance, Rbd, and the extreme-value distributions. The Rbd (in Ω) used in this analysis is defined as 0.5V/Ibd where Ibd is the gate current measured under Vg = 0.5V after oxide breakdown. By doing so, the percentage of each group can be estimated and failure analysis on samples from different groups can be expedited as the samples with low Rbd are hard breakdowns (the normal failures) and the ones with high Rbd are defect-related (soft breakdowns). By the curve fitting using Weibull and the extreme-value distributions, simple models are proposed for lifetime estimation.