edfas.org ELECTRONIC DEV ICE FA I LURE ANALYSIS | VOLUME 25 NO . 1 16 PROCESSES FOR THINNING AND POLISHING HIGHLY WARPED DIE TO A NEARLY CONSISTENT THICKNESS: PART II Kirk A. Martin RKD Systems, Aptos, California kirk@rkdsystems.org EDFAAO (2023) 1:16-19 1537-0755/$19.00 ©ASM International® INTRODUCTION Part I of this series, published in EDFA, November 2022, deals with the problems and considerations for thinning a large, warpeddie toauniformthickness of 50 +/- 2microns and suggests processes and procedures for obtaining this result. InPart I, it was suggested that insteadof attempting to thin an entire die to a thickness of less than 10microns, the entire die can instead be thinned to 50 microns for initial evaluation. The initial evaluation would result in defining general, rather small areas of interest. The small area of interest would be thinned to the desired, less than 10 microns, remaining silicon thickness. The reasons for thinning largedie to 50microns instead of the desired 1 to 5 microns remaining silicon thickness (RST) were discussed. A large die sample can be processed to 50 microns RST quickly. It will be robust enough to survive de-mounting and can be powered in a test socket. Analysis can be performed that identifies the area of the die that is of interest.[1] The area of interest is then thinned to the desired thickness, leaving the rest of the die thick enough to allow powering the sample. Thinning only a small area significantly reduces processing time and results inamuch more robust sample. In this article, the processes and considerations for locally thinning an area of interest to the desired RST are discussed. THE NEED FOR A ROBUST SAMPLE The forces that cause the die surface to curve are built in when the die is mounted on the substrate. As the die is thinned, the force is redistributed, causing the die surface to flatten when the sample is removed from the holder.[3] This causes increasing compressive strain in the remaining silicon. The force involved does not change but is applied to thinner silicon. At some thickness, the remaining silicon will no longer be able to support the force and the die will fracture. During the thinning process, the forces generated by the process add to the packaging forces. This can cause the die to fracture at greater thickness than the desired RST. Some of the packaging force is distributed to the wax that attaches the substrate to the sample holder. When the sample is de-mounted, most of the packaging forcewill be applied to the silicon. Thinning the entire die could easily cause the die to fracture during de-mounting and make any handling of the sample problematic. An RST of 50 microns should survive de-mounting, insertion in a test socket, or any other post-processing handling required. Locally thinning an area of interest will have little effect on themechanical strength of the sample as long as the area thinned is limited to a small fraction of the die surface. LIMITS ON THE THINNING PROCESSES Each processing step causes some level of damage to the remaining silicon. The depth of the damage is a function of the abrasive size and the down force of the tool. The general rule of thumb is that the bulk material damage extends 1 to 1.5 times the grit size of the abrasive. This indicates that a grinding tool with75-microndiamond should not be used to thin to less than 120 microns RST. The bulk of the silicon can be removed using large grit size if possible, but as the desired endpoint is approached, finer grits should be used. The endpoint for each process step shouldbe at a thickness equal to the final RSTplus the grit size plus the RST tolerance. If the final thickness is to
RkJQdWJsaXNoZXIy MTMyMzg5NA==