August_EDFA_Digital

edfas.org 27 ELECTRONIC DEV ICE FA I LURE ANALYSIS | VOLUME 24 NO . 3 during the scan) is also represented (Fig. 4c) to highlight the morphology of the obtained cross-sectional sample. Fromtopography profiles, the size of theDTI is determined to be 1.5 µm wide and 8 µm deep. It is important to note that the RMS roughness was measured to be 2 nm for Rq and 1.6 nm for Ra, sufficient to guarantee the constant contact area during the scans. SCM FOR THE LOCATION OF THE DOPING AREAS Simultaneously to the topography, the SCM signals (amplitude and phase) are recorded with the multidi- mensional approach (also called DataCube (DCUBE)) that combines imaging and dynamic point spectroscopy at each pixel (Fig. 5a). This combined mode produces an integrated and complete 3D data set: in every pixel of the mapping, the local variation of the SCM signals as a function of the applied bias V DC are recorded. During the “hold” segment, the conductive AFM tip is main- tained in contact with the surface, while the sweep of V DC voltage is applied (Fig. 5b). In this case, analysis is performed for V DC values from -3 V to 3 Vwith a step of 0.1 V. Afterwards, the tip is withdrawn from the surface for mea- suring on the next pixel, and this for all 256 x 256 pixels. This big data approach results in high dimensional data that can be sliced along any axis or plane. In SCM measurements, the phase sign indicates the majority carrier type under the probe and allows the distinction between n-type and p-type semiconductor regions (Fig. 5c). Acquiredat V DC = 0.4 V, the SCMphasemap allows one to distinguish the DTI structure, the epitaxial layer, theburiedp-type layer (BP) and the channel stopper. SCM amplitude extracted mappings (DCUBE slices) re- corded from -3 V to 3 V with V AC of 1 V at 90 kHz are repre- sented in Fig. 5d. This data shows the evolution of the SCMamplitude as a function of the applied voltage. Inter- estingly, for V DC < -0.5 V, it is impossible to detect the p-type layers (BP and channel stopper) on the mapping. But the multidimensional approach shows that there is a V DC window to detect thewhole structure of the sample. From -0.5 V to 1 V, the different doping areas (in terms of level and type) are discriminated in the silicon substrate. It is also possible to determine the optimum V DC for the imag- ing of this structure at V DC = 0.4 V, considering the distinc- tion of the layers and the level of the signal amplitude as a criterion. In fact, with these applied voltages, the SCM amplitudemapping highlights the DTI structures with the p+ type channel stopper implanted with a thickness of 900 nm. As indicated, channel stoppers prevent the two layers on either side of the DTI from being connected by using an inversion channel along the deep trench sidewall ofhighlyp-dopedsilicon.Only fewvaluesof V DC allowitsdis- tinction fromthe substrate. This result clearly indicates the impact of the V DC on the SCM analysis, and also the neces- sity to use amultidimensional approach as the function of the V DC by determining the optimum value of the applied voltage. This point is even more important in the context of a failure analysiswhere the structuremay not be known or when the failed device is compared to a fresh one. Fig. 5 SCM analysis, (a) schematic of the probe-sample system in the SCM multidimensional approach; (b) AFM force-curve as a function of time; (c) SCM phase mapping at V DC = 0.4 V; and (d) SCM amplitude DCUBE slices extracted for various V DC values from – 3V to 3 V. (a) (c) (b) (d) (continued on page 30)