Abstract
It has been shown that a normal temperature distribution within a sample set can produce a log‐normal failure rate in electromigration experiments if it is assumed that the equation tf=Ajn exp(Ef/kT) holds for individual stripe lifetimes. It has also been pointed out that if σ’s determined under accelerated‐stress conditions are used to predict failure under normal‐use conditions, incorrect failure rates may result. It has also been demonstrated that most of the σ’s observed in accelerated tests can be attributed to variations in temperature.