May_EDFA_Digital

edfas.org ELECTRONIC DEVICE FAILURE ANALYSIS | VOLUME 22 NO. 2 4 EDFAAO (2020) 2:4-12 1537-0755/$19.00 ©ASM International ® INNOVATIVE PUCK DESIGN FOR THE MECHANICAL CROSS-SECTIONING AND SUBSEQUENT ANALYSIS OF SEMICONDUCTOR PACKAGED SAMPLES IN FAILURE ANALYSIS Fauzia Khatkhatay and Pradip Sairam Pichumani GlobalFoundries Inc., Malta, New York pradipsairam.pichumani@globalfoundries.com INTRODUCTION Mechanical cross-sectioning is the preferred technique in the failure analysis of semiconductor flip-chippackaged devices, especially to examine and analyze the interface between the die and package. An epoxy puck containing the impregnated sample is polished to expose the site- of-interest. The main drawback is the long polishing time owing to the large size of flip-chip packaged devices. The insulating nature of the epoxy also causes charging arti- facts when analyzing the cross-sections in the scanning electron microscope. It is also very difficult to retrieve the sample after it is impregnated in epoxy, making subsequent analysis more challenging. Leveraging the learning from studies of chemical mechanical polishing (CMP) in wafer fabrication, a redesigned epoxy puck was proposed to reduce the polishing time. Based on results fromblank epoxy pucks aswell as epoxy pucks containing impregnated samples, polishing times have successfully been reduced by at least 84%. Additional revisions to the modified puck reduce charging under the electron beam and allow removal of the impregnated sample, as needed. By significantly reducing the polishing time, and facilitating options for subsequent analysis on the polished cross-section, this work addresses the major drawbacks associated with epoxy pucks in mechanical cross-sectioning. CHEMICAL-MECHANICAL POLISHING The widespread use of the polishing process in the chemical-mechanical planarization (CMP) of silicon wafers in integrated circuit (IC) fabrication has led to detailed characterization and in-depth investigationof the material-removal rate (MRR). ThePrestonmodel describes the MRR to be directly proportional to the vertical pres- sure and the polishing velocity. [1] Several modifications have been proposed to the Preston model based on the materials systems involved, such as the different metal or dielectric CMP processes. [2-3] Regardless of specific CMP step in the process flow, several observations canbemade based on the circular geometry of the silicon wafer. The net polishing velocity, frictional force, and relative tem- perature at any given point on the wafer increase as the radial distance fromwafer center increases. [4] The frictional force is also greater at the leading edge of the wafer than at the trailing edge, relative to the polishing direction. [5] The von Mises stress and therefore the MRR are found to be greater at the wafer edge than at the wafer center [6-8] Moreover, theMRR is lower for highpatterndensity regions on the wafer compared to regions with low pattern den- sity. [9-10] Consequently, within-wafer-uniformity after CMP has become an important parameter, especially as wafer diameters in IC fabrication have increased over the years. [11] To overcome the wafer center-to-edge polish- ing thickness variation, the wafer is divided radially into several zones and vertical pressure is tailored according to the zone so that a uniform polish can be achieved across the wafer. [11-12] Another area where CMP is used extensively in the semiconductor industry is in failure analysis labs. Espec- ially at the package-level, after non-destructive analysis has been completed, either layer-by-layer deprocessing or mechanical cross-sectioning are performed to expose the site-of-interest. [13-15] Of the two, mechanical cross-section- ing ismore critical and time consuming; critical because it exposes the interface between the die connection to the laminate, and time consuming due to potentially greater distance that has to be polished to reach the site-of- interest. Mechanical cross-sectioning has long been the preferred analysis method in destructive failure analysis as it can provide tangible evidence of a defect or a failure mechanism across a larger area compared to ion-beam based cross-sectioning techniques.