edfas.org ELECTRONIC DEVICE FAILURE ANALYSIS | VOLUME 26 NO. 1 26 Some samples, those that have very high stress from packaging as in Fig. 1, or those with some types of low-K dielectric, may not be able to survive the high lateral force from a diamond-grinding tool. These samples can be thinned with 74 or 45-micron lapping film, as these will produce very little lateral force. The lapping film takes more time but the resulting sample has a chance at being functional. The constant push for reducing processing time is somewhat misguided as faster always means more force and a greater chance of destroying the sample. Faster is often not better. SMOOTHING THE GROUND SURFACE The use of diamond lapping film to remove the grinding damage works very well. The lapping film, at relatively low pressures, only damages the silicon to about half the grit diameter in depth. Using a 45-micron film to remove about 50 microns of silicon works well and is reasonably fast. This removes a large portion of the damaged silicon and leaves a surface that may allow thickness measurement. If thickness measurement is difficult, a 30-micron film can be used with minor increase in processing time. The next lapping step, using 20 or 15-micron diamond film, should bring the RST down to about 63 microns. At this point, almost all of the damage done by grinding is removed. The remaining steps: fine lap, coarse polish, and fine polish, continue to remove damage done by previous steps and get the remaining thickness down to about 50 microns. Each step has built in uncertainty, and it is usually biased toward a thicker result. All touchdown problems, from contamination or other things, will place the tool face higher than is desired. Sometimes, the errors produce a sample that is, realistically, too thick for the next process step. This may require running a process step a second time to get the remaining silicon thin enough for the next step. It is not wise to ask any process step to remove more than its grit diameter. In the finer grit steps, less is definitely better. Any step should not leave any more than the sum of the following step’s grits to remove. If 50 microns is the target, there should be no more than 63 microns RST at the beginning of fine lap. That leaves 9 microns removed by fine lap, 3 microns by coarse polish, and 1 micron removed by fine polish to get to the target. These values are the same as the grit used for the step. The next step, which removes little silicon, is a final polish with colloidal silica. This produces an excellent, specular, surface that is suitable for any optical evaluation to isolate defects in the die. In all of this is the uncertainty of material removal. The truth is in the thickness measurements. If enough material has not been removed, the step must be rerun to get to the required starting point of the next step. It needs to be remembered that all errors tend toward removing less material and each following step is limited in the material it can remove. The earlier in the process that the deficiency in removal is identified the better. Rerunning second coarse lap to remove 10 or 15 microns takes only 10 minutes or so. Trying to remove a 10-micron excess thickness in coarse polish could take more than an additional hour. Spend your time wisely by correcting removal at the coarsest step possible. THE PROCESS TO 50 MICRONS For a sample with a starting thickness of 770 microns, the process is as follows: 1. Grind to remove 640 microns of silicon. 2. First coarse lap using 45-micron film to remove 50 microns. 3. Measure and input thickness for profile correction. 4. Second coarse lap with 20-micron film to a thickness of 60 microns. 5. Measure thickness. If more than 60 microns average, then rerun second coarse lap entering the measured thicknesses. 6. Measure and input thicknesses for fine lap with a target of 53 microns. 7. Measure and input thicknesses for a coarse polish target of 51 microns. 8. Measure and input thicknesses for a fine polish target of 50 microns. 9. Run final polish with a programmed depth of 1 micron removing about 0.25 microns. Fig. 1 The result of grinding a die with large internal forces.
RkJQdWJsaXNoZXIy MTYyMzk3NQ==