3 D elemental analysis of the rod‐shape alloy by using EDS tomography

Abstract

Energy dispersive X‐ray spectroscopy (EDS) is widely used to obtain an elemental map of a sample. EDS tomography reconstructs a three‐dimensional (3D) elemental map from a tilt series of two‐dimensional (2D) EDS elemental maps by using the back projection theorem. The back projection theorem is applied to tilt series of transmission images. An EDS map is not transmission image but X‐ray emission image. Therefore, the tilt series of EDS maps should not be applied to normal tomography in principle. The basic signal types for reconstruction by tomography are categorized into two, that is, absorbance and emission. In the case of normal transmission image (BF‐TEM or BF‐STEM), we calculate the mass thickness from the absorbance (−log[I/I 0 ]), where I: detected electron intensity after transmission, I 0 : incident intensity of electron, according to Lambert‐Beer's law. While, in case of the emission type, which includes ADF‐STEM, X‐ray fluorescence and etc., the emitted signal is proportional to mass thickness or number of atoms in a irradiated probe diameter. Thus, for X‐ray elemental maps, we are able to reconstruct a 3D map of each element by measuring X‐ray intensity of a certain element. The X‐ray intensity does not represent the number of atoms for the certain element, when the generated X‐ray is absorbed in the sample itself. Therefore, only if the X‐ray absorption is small enough to be ignored, we can apply the standard calculation procedure to EDS tomography. In practical calculation, the detection efficiency depending on the sample tiling angle is also considered. The efficiencies on sample tilt angles were measured using a known sample beforehand. This study reports how to obtain the 3D elemental maps in high magnification condition and to improve the accuracy of the EDS tomography using a sample of an alloy. Figures 1(a) shows a DF‐STEM image of an alloy composed of Mn, Ga and Ni. The instrumentation we use for this experiment was a field emission microscope (JEOL, JEM‐2800) equipped with two SDD detectors whose sensor area is 100 mm 2 each. A tilt series of EDS elemental maps for the region of sample shown in Fig. 1(a) was collected at tilt angles ranged ±80 degree. And the tilt steps were 4 degree, resulting in 41 collected maps for an element. The number of pixels for each map was 256 × 256. Figures 1(b)‐1(e) show a 3D DF‐STEM and elemental maps of the alloy, reconstructed by simultaneous iterative reconstruction technique (SIRT), which can reconstruct a 3D tomogram faster from fewer number of images than one in conventional method. As a result, we found that the Ni and Ga made solid solution, but Mn was segregated into small particles. The 58 Mn particles were visible in the Mn map, and the average diameter of these particles was estimated to be 10.7 nm. Next, we measured the X‐ray self‐absorption effect for improvement of the accuracy of 3D elemental maps. Figure 2(a) shows the DF‐STEM image from the rod‐shape NAND flash memory made by FIB. The tilt series of EDS maps was obtained from the field of view shown in Fig. 2(a) by an electron microscope (JEOL, JEM‐2100) with the single EDS detector. Then, we measured the X‐ray counts from the small gold colloidal particle indicated by the arrow in Fig. 2(a). The X‐ray counts are plotted on the stage tile angles (Fig. 2(e)). Since the X‐ray from the particle was absorbed by the sample itself, X‐ray counts were not constant. In order to analyze the 3D structure quantitatively, it is necessary to correct this effect. The effect of the absorption was estimated to be about 0.8 at maximum. In conclusion, to make an accurate elemental map by EDS tomography, it is necessary to consider the effect of X‐ray absorption, sample shape and sample thickness. It is complicated but we can correct these in principle.