The Failure Analysis Process

Introduction. How does an experienced Failure Analysis engineer crack a tough analysis? What makes the difference between a new engineer's Failure Analysis approach and the seasoned, effective analysis method of a veteran? Can our industry capture a formulation of correct Failure Analysis methodology to accelerate the development of new engineers and aid the strong skills of experienced analysts? More crucial, perhaps, is the underlying question, Are we able as analysts to think abstractly about the nature of our work? Or are we instead bound to a tool-focused, concrete way of thinking that has all the adaptability of the construction material of the same name? While our historical inability in the FA industry to abstract a general process from our daily work will probably not doom our discipline, this shortcoming will certainly rob us of a great deal of efficiency, flexibility and basic understanding of what we do in electronic device Failure Analysis. This author has the conviction that we are a better group of people than to allow this shortcoming to continue undetected and uncorrected for much longer. While many or most Failure Analysis departments make the definition and application of specific analysis process flows a significant priority, the subject of general FA methodology remains a minor one in the literature as of this writing. By way of a summary remediation, this tutorial paper presents a fully-general analysis formalization intended to make the failure analysis design efficient, rigorous and consistent from analysis to analysis and from failure to failure. The formalization amounts to a practical form of scientific method, customized for the failure analysis discipline, and enabling faster analysis to satisfy business constraints. The analysis design methodology, or decision engine, can be formulated as six steps arranged in a cycle or loop, shown in Fig. 1. This may not look like a Failure Analysis process flow, which is appropriate, since it is not, strictly speaking, such a flow. Instead, these steps in this order form a metaprocess, that is, a process whose function is to create another process. Applied to semiconductor failure analysis, this general decision engine metaprocess creates a specific analysis flow for a specific failure based on the facts about the failure, the results of analysis operations, the body of knowledge about physical causes and their effects and, finally, inferential logic. This metaprocess is intended to work properly—that is, generate a correct unique failure analysis flow—for any failure analysis of a manufactured product coming from any of the semiconductor wafer and packaging processes predominant today. Using the approach described in this tutorial, the analyst of any experience level should become able to make analysis choices in ways that reflect mature failure analysis skills. This approach can thus accelerate the maturity