edfas.org ELECTRONIC DEV ICE FA I LURE ANALYSIS | VOLUME 25 NO . 1 18 Although the area of interest may appear planar, it will usually have a slope in relation to the axis of the spindle on the polisher. The flat face of the tool is normal to the spindle axis and, therefore, the tool face plane is not parallel to the plane of the cavity surface. This requires the tool position relative to the cavity surface to be determined by the actual area of contact between the two planes. Fig- ure 2 shows the effect of this and the cavity edge fillet. If the cavity size is limited to the area of interest plus two times the tool diameter, there canbe edge distortion at the lowest part of the cavity floor that extends into the area of interest. If the slope is minimal, or the cavity extends off the die edge, this may not be a concern. Otherwise, the cavity edge may need to be more than one tool diameter from the area of interest to prevent edge distortion from extending into the desired area. As the sample needs to be removed from the polisher for cleaning at the end of each step and replaced on the polisher, there is the possibility of some misalignment. Even if thedie corners are realigned, therewill still be some difference in position after a sample has been removed, cleaned, and replaced. The alignment difference needs to be added to the cavity size to prevent the tool from contacting the cavitywall. This implies that the total cavity size needs tobe decreasedwith eachprocess step,making the cavity similar to an upside-down wedding cake. If the sample can be repositioned towithin 20microns, then the first step of a five-step process would have to be 200 microns larger than the final. The cavity size would then be reduced by 40 microns for each of the following steps. This approach works well for preventing the tool from banging against the cavitywall and reduces the buildup of fillet height as a series of smaller fillets are formed. THICKNESS AND PROFILE MEASUREMENT At the beginning of the area of interest thinning process, amechanical profile needs to be taken as well as thickness measurements. The mechanical profile can be used as the physical reference for the entire process if it is corrected for the thicknessmeasurements taken before eachprocessing step. The thicknessmeasurements should be in a 9-point pattern with all but the center point at the edge of the area of interest. An area of interest larger than 4 mm square should be avoided as surface curvature can introduce variations in the resulting thickness, depending on the actual curvature and mean slope. A 5 mm square area may require 25 thickness measurements in a repeatable pattern. With an RST target of 5microns, this may not be necessary due to the larger tolerance of the target thickness. The thicknessmeasurements need to be taken before each process step to ensure that the material removed comes from the right place. Since littlematerial is removed in final polish, themeasurements are not requiredbefore the final polish. THICKNESS CONTROL At 5-micron RST absolute thickness control is not so critical. Asmost polishing systems have 1micron position repeatability, a final average thickness variation of less than +/- 1.5 microns can be expected. This is generally in the acceptable range for a 5-micron target. Getting to a 1-micron target is problematic. It can be done, but not every time with every sample. When one is working with field failures, many of which are one of a kind, a “hit or miss” process is not acceptable. No data should ever be lost due to sample preparation. The +/- 1micron tool face position accuracy is the problem. In normal polishing, the tool face is in contact with the die surface. This makes a 1-micron RST essentially not possible, or, at best, iffy. There are only threemechanisms for abrasivematerial removal, scratching, rolling-scratching, and rolling-indenting.[2] Fixed abrasives, such as a grinding tool or lapping film, operate in the scratching mode. A lapping tool and abrasive slurry will operate in the “rolling-scratching” mode with some abrasive particles held by the tool face scratching the work surface while other particles are rolling across the surface. The use of a polishing cloth is close to the “rolling-indenting”mode but still holds some abrasive particles to allow scratching. If a hard-faced tool is operated at some position above the work surface, the mode of removal canbe exclusively rolling-indenting. This mode removes the least material per unit time but ends up being themost controllable. The rolling of the abrasive particles is due to the velocity gradient of the slurry. The tool is rotating at high speed and the die surface is stationary. This produces a relatively linear gradient in fluid velocity between the die surface and the tool face. If the gradient is small in relationship to the average diameter of the abrasive, the particles will not be very effective. As the gradient increases, by the gap between the tool face anddie surfacedecreasing, the abrasiveparticles domore. If one assumes that in normal operation, the tool face is separated from the work surface by only the abrasive average diameter, and that the mode of material is the same, doubling the separation should reduce the removal rate by 50%. Increasing the separation by a factor of 10will reduce the material removal rate by a factor of 10. With a 1-micron slurry, operating the tool 10 microns from the die surface will reduce removal rates by a factor of 10,
RkJQdWJsaXNoZXIy MTMyMzg5NA==