Structural Analysis of Semiconductor Devices by Using STEM/EDS Tomography

Introduction

Semiconductor devices are widely used in electronic products all over the world. Historically, high density-integrated semiconductor devices have been implemented by downsizing transistors on a chip. Recently, semiconductor devices are designed to be a three-dimensional (3D) stacking architecture for high integration and performance [1-2]. The 3D observation with nanometer scale resolution is essential for development and failure analysis of new devices [3-4]. Electron Tomography (ET) is one of the methods of acquiring 3D structure of the samples with nano-scale 3D resolution by using transmission electron microscopy (TEM). EDS tomography is performed by combining with Energy Dispersive X-ray Spectroscopy (EDS) and ET to realize 3D chemical characterization from sets of the 2D elemental tilt series maps taken by TEM equipped with EDS detectors [5-6]. The technique is applied to the new semiconductor devices and metal materials to observe these 3D structures [7-11]. When the first result of EDS tomography was reported in 2003, silicon-lithium type EDS detectors, which had low analytical counting rate, were commonly used to obtain EDS map of high electron dose and long acquisition time. Therefore, EDS tomography was not useful for 3D elemental analysis because of irradiation beam damage and contamination on samples. But recently, large-sized silicon drift detectors (SDD) and multiple EDS detection system for TEMs were developed [12]. By using the new SDD system, EDS maps can be obtained about 13 times faster than by using the previous silicon-lithium type detection system [13]. EDS tomograms from hard materials have become to be easily obtained by using the dual SDD system, though it is still difficult to obtain EDS tomograms from beam sensitive materials like bio samples.
EDS tomography is a powerful tool to analyze 3D elemental structure qualitatively, but EDS tomography has two kinds of limitation for 3D quantitative analysis caused by the relative positions of TEM samples and EDS detectors [14]. Figure 1(a) shows absorption effect, which is a limitation for quantitative analysis in EDS tomography. The black arrows indicate the paths of X-rays generated by incident electrons. When the structure of the sample is symmetric, the total amount of the generated X-rays from the near side and far side of the EDS detectors is the same. However, the generated X-rays from he far side can be absorbed by the sample itself. As a result, the detected X-rays from the far side are fewer than that from the near side. The absorption effect can cause artifacts in the resulting 3D elemental map. The recent study has attempted to compensate the self-absorption effect by calculation on the acquired EDS tomograms [15]. Figure 1(b) shows another limitation, that is shadowing effect. Normally, TEM samples were fixed on a grid and the gird is fixed on a sample holder. When two EDS detectors are located symmetrically with respect to the tilt axis of the sample holder in the previous EDS system, some of the generated X-rays are blocked by a grid or a sample holder in a specific tilting angle range. The shadowing in this configuration can also cause artifacts in the resulting 3D elemental map. In order to avoid this shadowing effect, it is necessary to correct the measured intensity of the EDS map after acquisition, according to the expected or pre-measured detection efficiency. For 3D quantitative analysis, both absorption effect and shadowing effect have to be removed from EDS tomograms.
The previous EDS-detector configuration is shown in Fig. 2(a). The EDS detectors are located on both sides of the tilting axis. In this configuration, the solid angles of the SDDs are varied on the tilting angle. We have improved to this variation by a new EDS-detector configuration, where SDD is placed on the tilting axis (Fig. 2(b)). It was expected that the SDD on the tilting axis has no shadowing effect in all tilting angle range for EDS tomography. The purpose of this study is to make clear that a new EDS-detector configuration have no shadowing effect in EDS tomography and to obtain 3D quantitative elemental maps from the semiconductor devices.

Fig.1 Schematic diagrams of the two limitations for 3D quantitative analysis in EDS tomography.

• Absorption effect by sample. The X-rays from the far side of the sample are more absorbed by the sample itself than that from the near side.
• Shadowing effect by sample holder. The X-rays from the sample are blocked by the sample holder or grid bars in a specific tilting angle range. The black arrows show the ray path toward the EDS detectors.

Fig.2

• Previous EDS detection system, which consists of two SDDs located symmetrically. One on the right side of the tilt axis and the other on the opposite side.
• New configuration of the EDS detection system. There are also two SDDs, and SDD2 is added to the tilt axis. It is expected that quantitative EDS tomogram without shadowing effect can be obtained by using only SDD2 for EDS tomography.

Experimental

The microscope used for our experiments was an aberration corrected 300 kV TEM (JEM-ARM300F, JEOL Ltd.) equipped with two SDDs (see Fig. 3). One detector is located on the tilting axis of a sample holder (SDD2), and the other is in the right side of the tilting axis (SDD1) (Fig. 2(b)). For obtaining the EDS tomogram quickly with no shadowing effect by using only SDD2, the 300 kV TEM equipped with a new pole piece, a new analytical high tilting holder and large-sized SDDs, was used. The new pole piece, named the wide gap pole piece (WGP), was designed thinner than the previous pole piece in order that EDS detectors can approach close to the sample holder. The spatial resolution of the 300 kV TEM with WGP is 0.062 nm due to the Cs aberration corrector. The new analytical high tilting holder was developed for EDS tomography. The tip of this holder is narrow and thin so as to make the generated X-ray unblocked for EDS tomography. By using these attachments and large-sized SDDs, whose detection area is 158 mm2, the solid angle of only SDD2 is reached to be more than 1.1 sr. Our EDS detection system can realize high spatial resolution and high analytical counting rate even with a single detector [13].
We prepared two kinds of samples for our experiments. One is a paint film sample to evaluate the shadowing effect in the new EDS detection system. The other is the fin-type field effect transistor (FinFET) which is one of cutting-edge semiconductor devices. The bulk of paint films, which were embedded by epoxy resin, were sliced by microtomy to be 200 nm thick. The thin sections were mounted on a supported membrane of the 3 mm-diameter thin-bar grid. 2D EDS maps were obtained using the 300 kV TEM with a probe current of 300 pA. A tilt series of EDS maps from -60 to +60 degrees with a 5 degree increment were automatically acquired by the tomography software (TEMography, SYSTEM IN FRONTIER Inc.) which is installed in the control PC for the TEM. The size of each EDS map was 256 by 256 pixels. The pixel size was 9.766 nm/pixel. The acquisition time was 190 minutes. All of EDS maps taken in our experiment were translated to the net counts maps from gloss counts maps with the same condition by using batch processing implemented in the EDS analyzer software (Analysis StationTM, JEOL Ltd.). At the beginning in the reconstruction procedure, tilt series of HAADF-STEM images were aligned by fiducial markers and the 3D structures without elemental information were reconstructed. The 3D reconstruction algorithm for this reconstruction was simultaneous iterative reconstruction technique (SIRT). Both the same alignment condition and the same reconstruction condition were applied to the EDS tilt series by using batch processing implemented in the TEMography software. Finally, we obtained the 3D elemental maps of paint film samples.
The FinFET sample was roughly cut by a low speed diamond wheel saw. The section of the sample was thinned by mechanical polishing. Finally the samples were milled by Argon ions milling machine (Ion SlicerTM, JEOL Ltd.) for TEM observation [16]. The gold colloidal particles with a diameter of 5 nm were dropped down on the sample and used as a fidicial marker for alignment of the tilt series. The TEM was operated at an accelerating voltage of 200 kV. The tilt series of EDS elemental maps were automatically acquired in a tilting angular range from +64 to -64 degrees. The degree step is 4 degree. The size of each EDS map was 256 by 256 pixels. The pixel size was 1.953 nm/pixel. Current density was 300 pA. The total acquisition time was about 120 minutes by using the single SDD (SDD2). The reconstruction procedure of the FinFET sample was the same as the paint film sample.

Fig.3

• 300 kV TEM (JEM-ARM300F, JEOL Ltd.) with the two large sized SDD shown in (b).
• The detection area is 158 mm².
• High tilt analytical holder developed for EDS tomography. The tip of the holder is narrower and thinner than the standard one so as not to block the generated X-rays.

Results and Discussion

EDS tomograms from the paint film sample were obtained in order to evaluate the shadowing effect in the new EDS detection system, which was installed in 300 kV TEM. The HAADF images and 3D elemental maps were shown in Fig. 4. The paint film sample consisting of titanium oxide particles, small silica particles, small iron oxide particles and carbon resin are marked in yellow, green, magenta and blue colors in Fig. 4(b), respectively. The measured total intensities of Ti Kαfrom the EDS maps of the paint film sample are plotted against the tilting angle of the sample stage in the TEM. The red, blue and green dots shown in Fig. 5 correspond to the total intensities detected by SDD1, SDD2 and SDD1+SDD2, respectively. Because the volumes of the titanium oxide particles in each EDS map are constant, the total intensities from Ti Kαmaps have to be constant in the result of quantitative EDS analysis. However, the intensities detected by SDD1 and SDD1+SDD2 decreased around -20 degrees by the shadowing effect. On the other hands, the intensities detected by SDD2 kept almost constant over the sample tilting range. This result indicates that the EDS detector located on the tilting axis had almost no blocking of X-rays from the holder and the grid of the sample. A nearly shadow-less EDS tomography system was realized by using this single EDS detector.
The shadow-less EDS tomography system was applied to the semiconductor devices. The HAADF image and the EDS tilt series maps of the FinFET were obtained by the 300 kV TEM shown in Fig. 6. The white dots in the HAADF image correspond to the gold nano-particles which were used as fiducial makers for the alignment in 3D reconstruction procedure. 3D elemental maps of the FinFET shown in Fig. 6 were reconstructed by using SIRT algorithm. In the resulted maps, germanium, titanium, tungsten, oxygen, nitrogen and silicon were detected. The 3D structures of the gate electrode on the silicon substrate were clearly observed by EDS tomography. The 3D nitrogen map corresponding to the insulating film was still noisy. Higher electron dose or longer acquisition time is necessary to observe 3D elemental distributions of light elements with high signal to noise ratio even by using the highly sensitive 158 mm2 SDD. Figure 7 shows the slices normal to the X, Y and Z directions extracted from the obtained 3D elemental volume map. The position of the Y-cut slice map is indicated by a yellow line in the corresponding Z-slice map shown below the Y-cut slice map. The germanium stressor (yellow), tungsten electrode (green) and the silicon channel (blue) are clearly seen in the Y-cut slice map (a). And the nitrogen (magenta) located between the channel and electrode was obtained in the 3D elemental maps. From these results, we can conclude that the EDS tomography is useful to analyze 3D elemental structures of modern semiconductor devices.
In this study, we tried to remove the shadowing effect in EDS tomography by improving the detector configuration in the TEM. However, the absorption effect also has to be removed for 3D quantitative analysis. When samples consist of light elements together with heavy elements, the effect of absorption was serious in quantitative 3D elemental analysis by EDS tomography. We have evaluated the effect of X-ray absorption by the paint film sample itself in EDS tomography. The EDS tomogram data set were taken by a 200 kV TEM (JEM-F200, JEOL Ltd.) equipped with two SDDs which were located on the tilting axis and in the right side of the sample holder, respectively. The 3D elemental maps of titanium, iron, aluminum, silicon, oxygen and carbon were reconstructed from the EDS tilt series maps of the paint film samples. The elemental composition ratio with voxel by voxel was calculated by using Cliff-Lorimer method. We have compared the 3D composition ratios from the three titanium dioxide particles which were indicated by the yellow arrows in Fig. 8. The 3D composition ratio from the particles with the diameter of 115 nm was consistent with the fact that the particles were titanium dioxide. But the atomic percent of oxygen in the particles with diameters of 190 nm and 315 nm was underestimated. It is considered that he X-rays from oxygen were absorbed by the sample itself. The corrections of the absorption effect are desired to analyze 3D quantitative elemental distribution with high accuracy in EDS tomography.

Fig.4

• The HAADF image of the paint film sample, taken by JEM-ARM300F equipped with the EDS detector, whose detector area is 158 mm2.
• The 3D image reconstructed from the sample shown in (a). Colors indicate atomic species: yellow (Ti), green (Si), magenta (Fe) and blue (c).

Fig.5

Total intensities of Ti Kα maps of the paint film sample using different detector signals plotted against the tilting angle. The blue dots, red dots and green dots correspond to the Ti Kα maps made by the signals of the SDD1, SDD2 and SDD1+SDD2, respectively. The intensities detected by SDD2 were almost constant over the sample tilting range for ET.

Fig.6

(a) HAADF image of the FinFET sample taken by a JEM-ARM300F equipped with single large-sized SDD whose detection area is 158 mm². White dots correspond to gold nano-particles used as fiducial markers. (b)-(h) 3D elemental maps of the same sample in (a), reconstructed from the EDS tilt series maps, showing the distribution of germanium, titanium, tungsten, oxygen, nitrogen and silicon atoms, respectively. The sample was thinned by ion milling. The size of the reconstructed volume is (381, 377, 121) nm.

Fig.7

Y-cut and Z-cut slice maps extracted from the reconstructed 3D elemental volume map of the FinFET sample using EDS tomography. (a) and (b) show elemental maps of the Y-cut slices at different positions indicated by a yellow line in the corresponding Z-cut slice maps below. The Z-cut slice is parallel to the wafer surface. The germanium stressor, tungsten electrode and silicon channel and substrate are colored yellow, green and blue, respectively.

Fig.8

3D elemental map of the paint film sample reconstructed by EDS tomogram obtained using JEM-F200 equipped with the SDDs with 100 mm² detection area (the left figure). 3D quantitative analysis of the three titanium dioxide particles indicated by the yellow arrows was shown in the right table. The composition ratio of oxygen in the particle 2 and 3 is underestimated due to the X-ray absorption effect. The absorption effect cannot be ignored for quantitative analysis in EDS tomography.

Summary

In order to obtain quantitative 3D elemental analysis, we developed the new EDS detection system for TEMs, and installed the system to the aberration corrected 300 kV TEM. The detection system consisted of the two differently configured EDS detectors, which are placed on the tilting axis of the sample stage (SDD2), and in the right side of the tilting axis (SDD1). The detection areas of these SDDs are 158 mm². In addition, we developed the high tilt analytical holder whose tip is narrower and thinner than the standard holder to make the generated X-ray from samples unblocked. By using this microscope with these attachments, the solid angles of SDD1 and SDD2 are 1.106 sr and 1.108 sr, respectively. EDS tomograms from the paint film samples were obtained by using the new EDS detection system. The total intensities from the Ti Kα maps of the paint film were almost constant over the sample tilting range for tomography. The result indicated that the new EDS detection system can obtain EDS tomograms with no shadowing effect. The 3D elemental maps from the FinFET were obtained with no shadowing effect by using the new EDS detection system. The shadowing effect can be removed by our EDS detection configuration. The correction of the absorption effect is desired to analyze 3D quantitative elemental distribution with high accuracy in EDS tomography.

References

• D. Hisamoto, W. C. Lee, J. Kedzierski, E. Anderson, H Takeuchi, K Asano, T. J. King, J. Bokor and C. Hu, "A folded-channel MOSFET for deep-sub-tenth micron era", IEDM Tech. Dig., (1998) pp. 1032-1034.
• H. Tanaka, M. Kido, K. Yahashi, M. Oomura, R. Katsumata, M. Kito, Y. Fukuzumi, M. Sato, Y. Nagata, Y. Matsuoka, "Bit Cost Scalable Technology with Punch and Plug Process for Ultra High Density Flash Memory" Proceedings of the Symposium on VLSI Technology, (2007) pp. 14-15.
• A. M. Cormack,"Representation of a function by its integrals, with some radiological applications", J. Applied Physics, 34 (1963) pp. 2722-2727.
• M. Hayashida, L. Gunawan, M. Malac, C. Pawlowicz and M. Couillard, "High-accuracy electron tomography of semiconductor devices", Microsc. Microanal . 21, 3 (2015) pp. 1609-1610.
• B. Fu, M. Gribelyuk, L. Dumas, C. Fang, N. LaManque, L. Hodgkins and E. Chen, "The Combination of STEM Tomography and STEM/EDS Analysis of NiSi Formation Related Defects in Semiconductor Wafer-foundries", Microsc. Microanal . 22, 3 (2016).
• K. Lepinay, F. Lorut, R. Pantel and T. Epicier, "Chemical 3D tomography of 28 nm high K metal gate transistor: STEM XEDS experimental method and results", Micron 47 (2013) pp. 43‒49.
• P. Burdet, J. Vannod, A. Hessler-Wyser, M. Rappaz and M. Cantoni, "Three-dimensional chemical analysis of laser welded NiTi-stainless steel wires using a dual-beam FIB", Acta Mater., 61, 8 (2013) pp. 3090–3098.
• A. Genc, L. Kovarik, M. Gu, H. Cheng, P. Plachinda, L. Pullan, B. Freitag and C. Wang, "XEDS STEM tomography for 3D chemical characterization of nanoscale particles", Ultramicroscopy 131 (2013) pp. 24‒32.
• B. Goris, L. Polavarapu, S. Bals, G. Van Tendeloo and L. Liz-Marzan, "Monitoring galvanic replacement through three-dimensional morphological and chemical mapping", Nano Lett . 14, 6 (2014) pp. 3220‒3226.
• G. Möbus, R. Doole, B. Inkson, "Spectroscopic electron tomography", Ultramicroscopy, 96, 3‒4 (2003) pp. 433‒451.
• Z. Saghi, X. Xu, Y. Peng, B. Inkson and G. Möbus, "Three-dimensional chemical analysis of tungsten probes by energy dispersive x-ray nanotomography", Appl. Phys. Lett. 91 (2007) p. 25.
• L. Strueder, P. Lechner and P. Leutenegger, "Silicon drift detector̶the key to new experiments", Naturwissenschaften 85, 11 (1998) pp. 539‒543.
• I. Ohnishi, K. Miyatake, Y. Jimbo, Y. Iwasawa, M. Morita, T. Sasaki, H. Sawada and E. Okunishi, "Ultrahighly Efficient X-Ray Detection System of Two Very Large Sized SDDs for Aberration Corrected 300 kV Microscope", Proceedings of Microscopy & Microanalysis 2016, 22, S3 (2016) pp. 318-319.
• C. S. M. Yeoh, D. Rossouw, Z. Saghi, P. Burdet, R. K. Leary and P. A. Midgley, "The dark side of EDX Tomography: Modeling Detector Shadowing to Aid 3D Elemental Signal Analysis", Microsc. Microanal. 21, 3 (2015) pp. 759‒764.
• Pierre Burdet, Z. Saghi, A. N. Filippin, A. Borrás and P. A. Midgley, "A novel 3D absorption correction method for quantitative EDX-STEM tomography", Ultramicroscopy 160 (2016) pp. 118‒129.
• A. Yasuhara, "Development of Ion Slicer (Thin-Film Specimen Preparation Equipment)", JEOL News 40 (2005) pp. 46-49.