Mechanical cross sectioning in the context of failure analysis is the grinding of die or die and package, usually orthogonal to the surface of the die, to examine defects or structure. Mechanical cross sectioning has been used in conjunction with failure analysis since the 1950s and 1960s. Mechanical cross sectioning is more of an art than science, as it may take several years for the analyst to become proficient at cross sectioning. A proficient analyst should be able to bisect a 3-micron-wide metal stripe in a plane parallel to its long axis.
Mechanical cross sectioning, although not as accurate or easy to perform as focused ion beam cross sectioning, is relatively inexpensive to perform. A good cross sectioning wheel and the necessary supplies cost less than $10k.
A bad cross section is worse than no cross section at all, since the process is destructive. Therefore, we recommend that cross sectioning be performed on a comparison IC before starting the failed IC. The analyst should hone his or her skills on non-critical failures or failures involving many samples, such as failed wafer lots or package lots.
Mechanical cross sectioning should be performed to verify electrical test data and other analytical data acquired during the failure analysis process. Many defects, including shorts, ESD, EOS, and numerous processing defects, can only be physically verified in cross section.
For most laboratories, the cost of focused ion beam equipment is prohibitive, so mechanical cross sectioning must be performed to view defects from a second axis.
Analysts use two major methods for mechanical cross sectioning: potting the sample and grinding with abrasive compounds, and glass wheel cross sectioning. Potting and grinding with abrasive compounds can be performed on most any sample. Glass wheel cross sectioning requires that the die be removed from its package.
For additional procedures refer to Failure Analysis Techniques, by Doyle and Morris, section III-N and The Microelectronics Desk Reference, by Lee and Pabbisetty, pp. 97-110.
First, coat the specimen potting cup with mold release compound. Place the sample in a specimen potting cup with the specimen oriented so that the cross section plane is face down and parallel to the bottom surface of the potting cup. Mix together the appropriate portions of epoxy and resin in a paper cup. Next, pour the contents of the paper cup into the potting cup until it is almost completely full. At this point, the mixture will probably bubble. You may want to place the potting cup in a vacuum for a short period (1 minute) to draw the bubbles out of the mixture and force the mixture into the package well of the IC. If all of the bubbles are not gone, this can be repeated once or twice. Do not pull a vacuum for too long a period of time on the sample, as this will cause the mixture to pull away from the specimen. Once you have removed the bubbles, set the specimen out to cure. At room temperature, a small sample will cure in approximately 2-4 hours. To accelerate the curing of the epoxy, the sample can be placed in an oven at 55 degrees centigrade for one hour. One should take caution not to over accelerate the curing process as this can also cause the epoxy to pull away from the sample. The curing of epoxy and resin is an exothermic reaction, meaning that heat is released. The larger the volume of epoxy and resin, the more internal heat will be generated. A large amount of heat released can cause the epoxy to pull away from the sample, or in severe cases cause the mixture to boil or melt solder connections. If you need to pot a large package like a multi-chip module, you need to provide a pan of water or another form of cooling to avoid these pitfalls.
Once the sample is potted and cured, the sample can be ground and polished. The approach here is to start with coarse grit, move to finer grit, and finish with polishing as you approach the defect or structure of interest (see Figure 1). Sandpaper for grinding ranges from 60 grit (very coarse) to 600 grit (fine). Be sure to only use the very coarse grits to grind through packaging material or dice that are not to be examined. Sandpaper grit coarser than 320 grit will cause the silicon to chip.
In order to see the surface of the specimen as you approach the location of interest, cut a window in the puck parallel to and above the surface of the die. After making the cut, grind and polish the window such that you can see the surface of the die through the window.
When you are within 8-10 microns of the defect, switch to a 6um diamond polish. To polish samples effectively, it is useful to have a 6um diamond paste, 1um diamond paste, and a 0.05 micron alumina slurry. An individual wheel covered with cloth for each compound is most effective. Place a small amount of polish compound on the cloth and work the compound into the cloth with a finger. Use generous amounts of polishing oil with the diamond pastes. For the final polish using 0.05 micron alumina, use only a small amount of water and polish for 30 seconds or less, as rounding of the specimen will occur.
To remove the polishing oils and diamond paste from the samples for viewing and changing to finer grits, agitate the sample in a ultrasonic bath with a small amount of liquid detergent. This frees the sample of oils that are trapped in cracks and prevents scratching at the finer polishing steps.
First, remove the die from the package. Under a microscope, remove the bond wires with a pair of fine point tweezers. If you hand is not steady, consider performing this step on a microprobe station with a mechanical probe tip. To remove the die from the package, place the package on a hot plate capable of reaching at least 300 degrees centigrade. If the sample has leads, bend the leads so the sample can lay flat against the hot plate for better heat conduction. Let the package heat on the hot plate for at least 10 minutes after the hot plate is up to temperature. Using a pair of tweezers, remove the die from the package.
Once the die is free, be sure to handle the die carefully because it is very easy to lose. Place the die on a glass wheel mounting fixture and glue the die to the fixture such that the defect overhangs the edge of the mounting fixture and the edge of the die is parallel to the surface of the glass wheel when placed down for grinding. While running a stream of water over the plate, turn the wheel on and gently place the mounting fixture down for polishing. Time the rate of grinding to determine how long the process will take if you plan on letting the grinding occur while you are not present. As you approach the location of interest, be sure to check more often to make sure you don't grind too far.
Mechanical cross sectioning should be one of the last steps performed on the IC. Never cross section the IC if you don't know what you are looking for. If the defect can be observed correctly by layer deprocessing, use standard layer removal techniques, which are more controlled.