Many failure analysts say that no two projects are exactly alike. Every defect is subtly shaped by its surrounding circumstances – the type of process used to construct the device, the environment in which the device was used, and the application that the device is used in can all contribute to the nature of the malfunction. Though they may be relatively unique in their specifics, most IC defects can still be classified with fairly broad brushstrokes; indeed, these classifications are vital to the failure analysis customer, as they often determine the type of corrective action that must be taken.
Some of the most common IC defects are those relating to electrical overstress, or EOS. In most cases, electrical overstress implies that a device has been used in ways that the manufacturer never intended – current or voltage far beyond the device’s specifications was somehow applied to the IC, often resulting in spectacularly catastrophic damage to the circuit. In severe cases, electrical overstress can reduce parts of a die to a mess of molten silicon slag; however, more often, the damage will be so small and so localized that specialized techniques like photoemission are necessary to locate them. Even though the electrical overstress event often causes such severe damage that it is difficult to determine where the defect may have originated, finding EOS is still valuable as it can prompt a customer to re-evaluate their application and consider whether their application may be inadvertently causing damage to the IC.
More troublesome than defects relating to electrical overstress are processing defects: IC defects that result from the improper manufacture of a chip. These types of defects can take many forms: scratches on a die from a poor packaging procedure, improperly etched metal traces causing short circuits, and malformed bonds are all examples of the sorts of things that would be considered processing defects. These types of defects can be far more worrisome to a manufacturer since they imply that there is an inherent process weakness with some (or all) of their product on hand, which may represent a sizeable amount of lost sales. In these cases, many manufacturers will perform thorough qualification of any questionable product in order to determine whether the defects found through failure analysis were simply random, or if they are indicative of an endemic issue.
Of course, these two categories are extreme overgeneralizations – there are many other types of IC defects that don’t necessarily fit with the examples given here. The value of good failure analysis is the ability to determine the specific circumstances and characteristics of a given defect, as opposed to dealing in the aforementioned generalities.