Reliability statistics perspective on standard wafer-level electromigration accelerated test (SWEAT)

Electromigration is an important reliability issue for ULSI. However, conventional electromigration test employing NIST structure usually takes too long for practical purposes. As the electromigration of metal interconnection has improved drastically over the year, the practical usefulness of the conventional electromigration test is even less, and the trend for electromigration test now is toward fast wafer level test, and SWEAT is the most popular test in the industry today. This is because it provides the test result in a very short test time. Conventional electromigration test has become merely a standard test for comparison with other fast electromigration tests. Recently, Meeker [1998] listed out several potential pitfalls of accelerated life testing from the perspective of reliability statistics. The purpose of this work is to provide a critical review on SWEAT from the same perspectives with the aids of the progressive understanding of the electromigration physics. It is found that SWEAT has several pitfalls that render its test results questionable. These pitfalls are the results of the high current densities and temperature used in the test. These pitfalls made SWEAT results questionable even in comparing alternatives, which is what SWEAT is designed for. In this work, these pitfalls will be described, and recommendations to eliminate/reduce the effect of the pitfalls are suggested. From this work, one can also see that both the reliability statistics and reliability physics are crucial in designing reliability test.