May_EDFA_Digital

edfas.org 7 ELECTRONIC DEVICE FAILURE ANALYSIS | VOLUME 22 NO. 2 The modified pucks show a significant increase in mate- rial removed compared to the conventional pucks for the same polishing times. The removal rate is calculated to be 10.514 μm/min for the modified pucks compared to 1.171 μm/min for the conventional pucks. The resultant improvement shown by themodified pucks over the con- ventional pucks translates to an approximately 9x faster polishing speed or 89% reduction in polishing time. It must be noted that though the degree of improvement depends on the relative dimensions of themodified puck —any changes to the diameter of the puck, the width of the outer ring, or the shape and dimensions of the central rib will most likely show dif- ferences in the MRR—it validates the initial premise of the modified puck design. POLISHING TIME COMPARISON FOR CONVENTIONAL VS. MODIFIED EPOXY PUCKS WITH IMPREGNATED SAMPLES The next step was to verify whether the results from the blank epoxy pucks could be replicated if real samples were introduced. Two identical flip-chip samples were diced to the same dimensions and impregnated in epoxy, one in a conventional puck and the other in a modified puck. Shown in Fig. 3a and b are the conventional puck and the modified puck, both with impregnated samples, referred to as Config0 and Config1, respectively. It may be noted that the width of the central rib was increased to provide a sufficient thickness of epoxy on either side of the sample. The sample was oriented verti- cally in the epoxy puck according to Fig. 1d, where the polishing would first expose the laminate and then approach the die. Each puckwas polishedoneat a timeusing600grit sandpaper to polish through the laminate, a polishing cloth with 1 μm alumina suspen- sion after the die was exposed, and then changing to a polishing cloth with 0.3 μm alumina suspension upon approaching the first row of C4 bumps. The samples were polished until reaching the center of the first row of C4 bumps, and the polishing times were carefully recorded. Figure 3c shows the time per step required to reach each of the checkpoints, from Time0, or polish start, to exposing the laminate, then exposing the die, and finally reaching the center of the first (a) (b) (c) (d) Fig. 3 Opticalimagesofthe(a)conventionalpuck,Config0,and(b)themodified puck, Config1, both with impregnated samples. (c) Polishing time per step required to reach each polishing checkpoint. (d) Cumulative polishing time to reach each checkpoint. row of C4 bumps. Figure 3d shows the cumulative time starting from Time0 to reach every checkpoint. Config1 shows a clear reduction in polish time to reach each of the checkpoints considered, resulting in 84% reduction in time per step as well as the cumulative polishing time. Shown in Fig. 4 are the SEM images of the Config0 cross-section after reaching the center of the first row of C4 bumps. Figures 4a and b are lowmagnification images of the sample at the left and right corners of the die, respectively. In both corners, the die, C4 bumps, laminate,