Decode Semiconductor Dislocation Traces Unveil Chip Leakage Causes with TEM

Functions of current TEM/STEM are powerful, a TEM/STEM system has more than ten micro-nano material analysis techniques available. Among them, there are three techniques can analyze the 3D trace of dislocation in the crystal. They are TEM tomography ^{[5, 6]}, 3D TEM, and TEM stereo ^{[1, 3]}.

(1) TEM tomography

TEM tomography is the most advanced one, but its operation is complicated and time consuming, so it is adapted in TEM laboratories of academic and research institutes, but not in industrial ones, and the range covered by the sample is usually too narrow.

(2) 3D TEM：

Plane-view type TEM specimens were usually used for TEM observation firstly, then adequate protection layer was coated on the region containing dislocations, finally one more FIB cutting was conducted. This takes some risk of destroying the TEM specimen and getting rid of parts of dislocations, and gives improper data.

(3) TEM stereo：

On the other way, TEM stereo needs one TEM specimen only, photo dislocations from two orientations of same diffraction condition, then do some basic geometrical measurement and calculations, the trace of dislocations can be descripted. The method is much simple for TEM sample preparation and has higher yield.

It is hard to judge which one is farther when two sailboats are far away from the coast, as shown in Figure 1. The distances of them from the coast can be calculated by using basic geometry if a large flat region is available. Figure 2 is a top view schematic diagram of Figure 1. The observer travels a distance L parallel to coast, the L_A and L_B are distances between sailboat A and sailboat B to the observer respectively, and θ_A and θ_B are corresponding angles of sailboats to the two ends of the L line. The values of L_A and L_B can be calculated to be a certain accuracy when a protractor and a ruler of certain accuracy are available.

The ray diagram of a simple thin convex lens is shown in Figure 3. The yellow arrow at the left side stands for the object, and the blue arrow at right side stands for the magnified upside-down real image on the imaging plane. For a fixed image plane, the depth of field (DOF), Dfd, is the total distance the object moves forward and backward while the image is still in acceptably sharp focus. For a fixed object, the depth of focus (DOS), Dfs, is the total distance over which the image plane can be displaced while the image is still in acceptably sharp focus. α is the angle between the optical axis and the line connecting the center point of the object (or image) and periphery of the lens. Both DOF and DOS are functions of α and the resolution of the lens, d. Generally, DOF and DOS are 400 nm and 1000 m respectively for a TEM operated at 200 KeV. The large value of DOF means that features in the specimen, from top to bottom, are all in focus. So, the relative depth of two features is not able to be identified from a single TEM image, as blue line and yellow line described in Figure 4(a).

The corresponding structure of the schematic projected image of Figure 4(a) can be either Figure 4(b) or Figure 4(c). It means that the 3D structure is unable to be determined from a single image such as Figure 4(a). If we tilt the specimen counterclockwise, as Figure 5(a), and the spacing between yellow and blue lines increases, as shown in Figure 5(b), then we can deduce that the 3D structure is that shown in Figure 4(b), not that shown in Figure 4(c).

Dislocations do not terminate in the crystal unless they form loops. So, dislocations induced in silicon substrates due to improper annealing heat treatment after ion implantation will run through the silicon crystal. They will become leakage paths if they penetrate the p-n junctions.

Traces of dislocations in a TEM specimen can be characterized by specimen tilting similar to that shown in Figure 5(a). Because atoms of dislocations are all silicon, same to matrix around, diffraction contrast of TEM images with dislocation is the dominating image contrast mechanism. Thus, morphology of dislocations changes with diffraction conditions. Dislocations can be invisible at some diffraction conditions. Two beam diffraction conditions with optimum deviation parameters should be used to analyze dislocations. Two images taken at two angles separating 15 ~ 20 degree should be kept at same two beam condition, so the change in morphology of dislocations can be minimized.

A set of typical TEM stereo images are shown in Figure 6. Two dislocations labelled A and B in a TEM BF image taken at [100] exact zone diffraction condition, as illustrated by the CBED pattern inserted at the lower right corner, is shown in Figure 6(a). Figure 6(b) and Figure 6(c) are two TEM BF images of same two beam condition, but about 8 degrees opposite orientation related to Figure 6(a). The distance between dislocation A and dislocation B in Figure 6(b) is distinctly smaller than that in Figure 6(c), i.e. d1 < d2. The distance, L, between dislocation A and the edge of the active region (marked by a red dash line) decreases from Figure 6(b) to Figure 6(c). Changes in distances in Figure 6(b) and Figure 6(c) are similar to that described in Figure 5. So, we can deduce that dislocation A and dislocation B locate at different depth at first glance.

By using the edge of the active region to be a reference line, data of distances of some inflection points of dislocation A and dislocation B are measured. The traces of dislocations, as shown in Figure 7, of these two dislocations can then be depicted after some basic geometry calculation. Dislocation B runs in surface layer only, but dislocation A extends down to 260 nm. This means that dislocation A penetrated p-n junction and formed a leakage path which caused a leakage fail in this device.