Time-Zero Dielectric Reliability Test by a Ramp Method

Using a long-established feature of time dependent dielectric breakdown (TDDB) it is demonstrated that a ramped voltage breakdown histogram of a sample population can be used to accurately forecast the rate of breakdown failures in the field. It is shown that such a histogram can be interpreted as the field dependence of failure at constant time. The tamp-TDDB relationship involves no fitting parameters and only a single material-related parameter. The temperature dependence of this parameter is established for SiO2 Extensive ramp-life test measurements have verified the relationship experimentally. It is argued that the usual models used to relate laboratory life tests to reliability failures are inherently faulty. The faults stem from the temperature dependence and the distributions of failure times, both of which must be assumed in order to extrapolate accelerated life tests to use conditions. On the other hand the actual distribution is measured in a ramp test and the temperature acceleration is not needed. This finding has far-reaching implications for reliability assessment. Dielectric life tests can be replaced by the relatively simple and rapid ramp test with increased confidence in projection. From the analysis it is shown that the effect on reliability of a high field screen can be quantitatively determined in an absolute manner.