What is failure analysis?
Failure analysis can be defined as a method or a series of actions undertaken to find the reasons a particular failure exists and correcting the causes.  It is vitally important to understand that these two distinct elements are both essential if one is to embark on a study of failure analysis.   There is no particular relevance to knowing the cause or causes of a certain failure is if the problem cannot be resolved.  In this definition, failure analysis can obviously apply to numerous fields of study.  We are studying failure analysis as it applies to the semiconductor industry. 

How is failure analysis relevant to the semiconductor industry?
There are many reasons failure analysis is practiced in the production of semiconductor chips. The production of silicon wafers is costly in both time and money and the yield is nowhere near 100%.  This means there is a certain amount of failure and much work is being done to increase yield.  What makes certain dyes on a wafer pass and others fail?  One reason is that the yield decreases significantly as the distance of the dye from the center of the wafer increases.  This is because during the lithography process, everything is focused on the center.  Another reason is much more transistors are being fit into a smaller area to increase the performance of the chip which means that a small microscopic particle that lands in the wrong place can cause a short or an open.  Actually, shorts and opens are less common and partial shorts or opens are more common.  The resistance of a wire increases as the perpendicular area decreases, so as the wires on a chip become smaller, minute defects in the manufacturing process are magnified.  This usually occurs in the form of small dust particles that land on the wafer in the manufacturing process.  It is easy to get a high concentration of aluminum in one place and a small concentration in others thereby altering the resistance.  We are interested in finding the effects of these abnormalities as they apply to the real time functioning circuit. 

What are we doing with this atomic force microscope thing anyway?
The questions remains, how is the performance of a chip affected by these manufacturing anomalies?  And how can the anomalies be tested if there is no real way of simulating these chance events? It may seem intuitive to find one dye with a certain inconsistency and compare its performance with one without the inconsistency.  The flaw with this method is that you don?t know it there are other inconsistencies because there are literally up to millions of transistors and wires on a dye and they can not all be looked at.  So it makes more sense to find two or more dyes that have the same performance and inject a fault into one of them and then study the effects.  The atomic force microscope (AFM) is capable of scanning to a resolution down to the atomic level but can also be used to manipulate particles.  Using the instrument in this way, we are trying to develop a technique to partially or completely cut the outer wires in a dye with surgical precision.  Another technique that is currently used in industry today is to use a focused ion beam to blast away part of a wire.  The limitation with this method is that there is residual damage incurred on the dye in places not desired.  Also, ions are charged and induce an undesired current.  The AFM is of particular interest because of its surgical precision, and does not introduce a current.  As components become smaller and smaller, the ion beam method will become obsolete and perhaps the AFM will pick up where it leaves off. 

Copyright © 1999 
University of Rhode Island