edfas.org 35 ELECTRONIC DEV ICE FA I LURE ANALYSIS | VOLUME 24 NO . 4 had at ambient. The wax interferes with the return to a free-state profile, flattening the profile and building forces into the mounted sample system. It has been shown that the profile variation is not uniform across the die surface resulting in time variant profile changes greater than +/- 4 microns for some observed profile measurement points.[2] These time variant profile changes occur in times that are comparable to the gross thinning processes. With a thinning system that reproduces the initial measured die surface profile on the thinned die, the dynamic changes in surface profile will have the original profilemachinedonto them. Figure 1 graphically demonstrates the RST variation resulting from the thinning process and the substrate-die interface profile changes created by force redistribution resulting from the thinning. DYNAMICS OF A MOUNTED SAMPLE The short-term changes in sample surface profile have been characterized for flip chip devices.[2] The same variables also apply to plastic encapsulated devices although the magnitude of the changes will vary from those observed with flip chip samples. The basic rule is that the larger the die, the larger the deviation of the die surface from a plane and the greater the short-term changes from mounting forces and force redistribution resulting from thinning the die. SAMPLE MOUNTING PROCEDURES AND REQUIREMENTS Samples are usually mounted to some support structure with mounting wax such as Crystal Bond 509. All efforts need to be made to reduce the stresses in the package/die systemas describedpreviously.[2]. The sample should be covered with an insulating material to reduce thermal gradients and the stresses they create. Enough time needs to be allowed for full thermal equilibration both in heating, mounting, and cooling to ambient. If enough time is allowed, theonly additional stresses added to the die/substrate system are related to the difference between the solidus temperature of thewax and ambient. These stresses have been described before[3] and can be significant, but the reduction or elimination of thermal gradient stresses will greatly reduce profile changes encountered during thinning. It should be noted than neither the die nor the substrate are truly flat except at a single, elevated temperature. The metallization density across the die surface is not constant and multiple metal layers are often coupled vertically to reduce the impedance of power connections. Themetallization and silicon have different coefficients of thermal expansion and the metallization may be applied and alloyed at elevated temperatures. This will produce both local small scale aswell as total die profile variations. The same effects from metal density and the vertical distribution of density variations will result in substrate warpage with temperature change. As the die and substrate are attached at elevated temperature, cooling to ambient can create very complex curvatures in the C4 ball “plane.” The die and substrate may each come close to planar, but at very different temperatures and the union of the two at a completely different temperature. This ensures that there will never be a “relaxed” state where the die surface is flat and stable. MEASURING THE INITIAL PROFILE It is assumed that the starting silicon thickness variation is zero. Therefore, themeasured die surface profile is the same as the die active surface profile. This is generally a good assumption as even thinneddie processed inwafer form results in nearly uniform thickness. The initial profile is the best reference to the first steps in die thinning and therefore need to be used to establish the tool path. If the part moves during the thinning processes, it will result in thickness variations that may be compensated for later in the process. PROCESS VARIATIONS AND SAMPLE PROFILE DRIFT At the beginning of the thinning process, the sample should be as close as possible to its unmounted or “free state” curvature. In the free state, all forces are in equilibriumand the profile of the die surface is static at any given temperature. If the sample couldbemounted in free state, therewould be no forces on themountingmediumresulting in the least changes to the die profile during processing. This is not possible, as there needs to be something that secures the sample to the holding fixture. Mechanical clampingwill not adequately support the sample andmay introduce additional distortions. Other adhesives, other thanmounting wax, may get closer to a free-statemounting but generate problems with reusability of fixtures and demounting the sample for further processing. Only a percentage of the change in the die surface profile will occur during the grinding process as it only takes about 60 minutes to thin a 16 by 20 mm die from 770 microns to 155 microns. During this time, the die center may rise as much as 15 microns resulting in the center being thinner by that amount. The sample movement will continue for many hours. Letting the sample
RkJQdWJsaXNoZXIy MTMyMzg5NA==