edfas.org 19 ELECTRONIC DEV ICE FA I LURE ANALYSIS | VOLUME 25 NO . 1 but will also reduce the effects of positional uncertainty. A 1-micron variation in tool position only results in a 10% change in material rate. Operating with a defined separation between the tool face and the die surface, therefore, makes material removal rate less dependent on absolute tool position. Although this is a gross simplification of the physics ofmaterial removal, the concept is valid. Operating witha significant gapbetween the tool face anddie surface will reduce the material removal rate and its dependence on absolute tool position. Using this method may require stopping the thinning process tomeasure average thickness and continuing the process to the desired endpoint, but it ismuch better than having 0.0 microns RST. CONCLUSIONS With careful processing, cleaning, RSTmeasurements, and a sample processing machine that moves the grinding, lapping, and polishing tools to a thickness corrected surface profile, samples can be reliably processed to a 50 micron thickness with an RST variation of +/- 2.5 microns across the majority of the die. There are variations in RST near the edges of the die that are created by the lapping and polishing tools not moving off the die surface anddistortions relating to the slope of the surface near the die edge. All thickness measurements need to be made inside of the die edge distortions. Within these limits, +/- 2.5micron, or better, RST variation is achievable without operator intervention. All the operator needs to do is clean everything, measure the RST at 9 points on the die surface, install tools, and apply the correct slurries. Adjustment of nominal material removal in each stepmay be required to get the final desired thickness, but no operator involvement should be required during processing. Push the run button and go to lunch. The resulting sample is robust enough to go into a test socket and powered up. It is thin enough for evaluation and identification of areas of interest that can be thinned further for detailed analysis. Local thinning to less than 10microns can be done quickly and does not compromise the robust natureof the sample. This two-stepprocess gets samples completed in hours instead of days required for whole sample thinning to less than 10 microns. Thinning to 1micronRST is a bitmore difficult andmay require material removal with a significant gap between the tool face and die surface. This mode of operation reduces absolute tool position effects onmaterial removal rates butmay require interrupting the process tomeasure average thickness. The described processes allow rapid processing of a samplewhile keeping it robust enough to be handled and powered up and allowing analysis of specific areas at the thinnest possible location. REFERENCES 1. K. Martin: “Processes for Thinning and Polishing Highly Warped Die to a Nearly Consistent Thickness: Part I,” Electronic Device Failure Analysis, 24(4), p. 34-38, 2022. 2. S. Bhagavat, et. al.: “Effects of Mixed Abrasive Slurries on Free Abrasive Machining Processes,” SUNY at Stony Brook. 3. S. Timoshenko: “Analysis of Bi-metal Thermostats, J. Opt. Soc. Am., 11(3), p. 233, 1925. ABOUT THE AUTHOR KirkMartin has almost 50 years of experience in designing andbuilding specialized equipment for all aspects of the semiconductor industry, fromcrystal growth through final test and failure analysis. In 2017, he became a founder of RKDSystems, whichdesigns andbuilds equipment for semiconductor failure analysis sample preparation. Martin has patents in the fields of sample preparation, chemical vapor generation, fluid handling, and electrostatic discharge detection and mitigation. His previous positions include vice president at Nisene Technology Group, director of Advanced Products at Texas Materials Labs, amanufacturer of specialty semiconductormaterials, and vice president at Automated Technology Inc., a manufacturer of front-end test and measurement systems.
RkJQdWJsaXNoZXIy MTMyMzg5NA==