February_EDFA_Digital

edfas.org 7 ELECTRONIC DEV ICE FA I LURE ANALYSIS | VOLUME 24 NO . 1 In this work, the burst-signal energy is used to filter the measured AE signals. Using this AE feature, the crack signals are clearly distinguishable fromother signals, like plastic deformation, in terms of higher amplitudes and signal lengths. The burst-signal energy E burst is calculated from the sum of all squared voltage samples U i of a hit (i = 1, 2 … n ) multiplied by the sampling time interval T s and has the unit eu (1 eu = 10 14 V 2 s). [1] To identify the correct filter criteria for the “first oxide cracks,” 100 contact cycles were performed sequentially on test structureW11. The test structurewas again stressed for each contact cycle up to amax. contact force of 650mN to achieve nearly 100% crack probability in combination with indenter FP10. Figure 8 shows a scatter plot of the burst-signal energy in logarithmic scale for all recorded AE hits as a function of the contact force cumulated for 100 contact cycles. The burst-signal energy and the number of measured AE hits are relatively high at low contact forces. With increasing contact force, the burst-signal energy is getting lower, while the imprint depth increases. This effect can be explained by weakening of the plastic deformation of the Al-Cu top layer at higher forces. After the imprint depth saturation region between a contact force of approxi- mately 200 and 280 mN, the burst-signal energy of the AE hits is rising again, while the contact force is continuously increased. As shown in Fig. 8, the oxide crack generation starts above a force value of approximately 280 mN. The occurrence of cracks within the range of critical contact force is also confirmed by optical inspections (see Fig. 7b). In this experiment, the lower filter limit of the burst- signal energy is set to 7 eu and for the contact force it is set to 280 mN. Only AE hits meeting these filter criteria are considered and utilized for further data evaluation. After filtering the data from Fig. 8, only those hits are selected that have occurred for the first time per indent, so-called “first oxide cracks.” The critical contact forces are statistically evaluated fromthe remaining data points. Figure 9 shows the result for all 100 contact cycles as a histogram plot over the contact force. The graph also shows the distribution function as a blue line, which has been fitted to the data points. The distribution function graph is not symmetrical, as themaximumcount of AEhits is shifted slightly to the right. (a) (b) Fig. 7 (a)Measuring the imprint depth after Al-Cu indentationusing a confocalmicroscope. (b) Differential interference contrast (DIC) microscopy images of oxide cracks after indentation on Al-Cu top layer and chemical crack preparation at different maximum contact forces. [1] Fig. 8 Burst-signal energyof all hits recorded for 100contact cycles on test structure W11 using indenter FP10 as function of contact force (all triangles) and clustering of AE signals regarding first oxide cracks (triangles marked in red).