Physical Approach to Reliability

From the material in this book, more or less complete regularities concerning transistor failures can be derived. Based on the knowledge of these regularities, several questions might be asked. For example: What is the application value of this knowledge? Is it possible to develop and to make each particular transistor in the production lot more reliable? What are the methods for SOA evaluation of new devices? Is it possible to predict with desired accuracy catastrophic failures of the particular devices and systems? Some opportunities are summarized in this final chapter. They are based on the physical approach to reliability. One of the possible ways can be presented by means of fulfillment of a few complementary conditions. These conditions are expected to provide a considerable increase in the probability of reliable operation transistor. Without excessive detail, these conditions can be listed as follows:

• (i) Gain in-depth understanding of the basic physical failure mechanisms for given semiconductor device structure and process

• (ii) Maintain understanding of phenomena that are limiting safe operation regimes and long-term reliability parameters

• (iii) Develop recommendations for device design using knowledge of failure mechanisms toward possible reduction, compensation, or even elimination of particular failure mechanisms while preserving the basic operation parameters within desired spec limits

• (iv) Reveal potentially “weak” places in the device architecture, process technology, and manufacturing for identified failure mechanisms

• (v) Provide competent, physically reasonable choice of the test regimes and long-term reliability tests

• (vi) Understand the limitations of diagnostic methods according to the basic failure mechanisms

• (vii) Estimate real transistor reliability in physically reasonable regimes of safe operating area (SOA)

• (viii) Estimate the long-term reliability parameters with the help of accelerated tests focusing on the indications of potential reliability issues.