Emission Microscopy - A Guiding Light for Failure Analysis
The ability to isolate a defect in a sea of circuitry, pinpointing a problem hiding amongst a plethora of transistors and metal lines, is one of the cornerstones of successful failure analysis. An analyst would be hard pressed to study an anomaly in depth without first knowing where the anomaly is. The resourceful analyst has many tools and techniques to aid in the detection of defects on an integrated circuit; some, like liquid crystal or thermal imaging, are best used to find short circuits that generate large quantities of heat, while others, like time domain reflectometry, are best suited to finding open circuits. Unfortunately, these techniques are often not sufficient, and an analyst must find a way to characterize a device, creating a baseline against which to contrast a failing unit in order to detect the defect at the root of a failure. In these cases, emission microscopy provides the perfect platform upon which to build an analysis.
Emission microscopy (often referred to as light emission microscopy, photoemission microscopy, or by the trade name EMMI - EMission MIcroscopy) utilizes a high-gain camera to detect the miniscule amounts of light emitted by some semiconductor devices and defects. A device is placed in view of the microscope, which is surrounded by an enclosure to block out all ambient light, which would ruin the sensitivity of the technique. The device is powered up, either in some functional condition or in a simple pin-to-pin biasing scheme; at this point, the camera system takes over, mathematically integrating data for as little as a few tenths of a second to as long as several hours.
The final output of the system is an image with splashes of color at sites where the camera detected photoemission. While photoemission can be indicative of defects like gate oxide pinholes or transistors that have been damaged by electrostatic discharge, there are some semiconductor devices that photoemit even when operating properly; the electron-hole recombination that takes place in a forward-biased bipolar junction transistor, for example, gives off light. At first glance, it would seem that this phenomenon would limit the utility of emission microscopy for fault detection; on the contrary, however, it is the ability to analyze devices that are working properly as well as those that are defective that make emission microscopy so valuable.
While emission microscopy can certainly be used successfully to locate things like damaged junctions on a diode or overstressed protection structures on an IC, the technique truly shines when it is used to compare malfunctioning devices to a properly operational unit. By first testing a good unit, an analyst can create a characteristic map of the device under a given set of test conditions, serving in the same way a star chart would serve a sailor lost at sea. By using the photoemission results from the properly functioning unit as a guiding light, it becomes easy to pinpoint anomalies on the failing unit simply by looking for emission sites that do not appear on the good unit. On units where there are more potential sites for failure than there are stars in the sky, using emission microscopy in this fashion is an efficient, straightforward way to detect a defect.
Emission microscopy is a powerful tool for finding any of a number of different semiconductor defects. In and of itself, though, the tool is only another way to gather data; just as an inexperienced sailor may find their ship run aground despite consulting the best star charts, a detail-oriented, technically proficient analysis team is necessary to capitalize on the data from an emission microscope, using the information to cast light on the problem and successfully identify the root cause of the failure.
Derek Snider is a failure analyst at Insight Analytical Labs, where he has worked since 2004. He is currently an undergraduate student at the University of Colorado, Colorado Springs, where he is pursuing a Bachelors of Science degree in Electrical Engineering.