TY - JOUR
AU - McCartney, Martha, R.
AB - Abstract Quantitative analysis of electrostatic potential in semiconductor device samples using off-axis electron holography in the electron microscope is complicated by the presence of charged insulating layers. Preliminary results indicate that the behavior of p-type material near the Si–insulator interface may differ from that of n-type if the insulator is charging. Coating one side of the sample surface with carbon usually eliminates charging effects. Holographic phase measurements on thin silicon oxide film at liquid nitrogen temperature indicates that the maximum electric field near the edge of an charged region is 18 MV cm−1, on the order of the breakdown voltage. electron holography, electrostatic potential, charging, semiconductor device, breakdown voltage Introduction The semiconductor industry is very interested in a quantitative, high resolution technique that can provide 2-D measurements of the p–n junction potential distribution at source–drain regions in deep-sub-micron devices. Electron holography has this capability, precisely because it is sensitive to the presence of ionized dopant atoms in the depletion regions of the junctions. For most ‘clean’ samples, however, exposure to the electron beam results in drastic charging effects. The presence of charged dielectrics at the Si–insulator and Si–metal–insulator interfaces can result in phase shifts due to fringing electric fields and pronounced band-bending in the silicon. While standard transmission electron microscope (TEM) techniques can reveal sample charging, particularly by defocusing the image, electron holography is capable of directly measuring phase shifts of the incident electron wave caused by uncompensated electrostatic charge in the sample and the associated electric fringing fields. Charging effects in the TEM are of intrinsic interest, but these effects are detrimental to many types of analysis. In the study of semiconductor devices, the presence of layers of dielectric material are unavoidable as they form integral parts of the device structure. When insulating layers accumulate charge during holographic imaging, the resulting electron phase shifts severely complicate analysis of the electrostatic potential distribution of the device. Reduction of the electron dose may be thought of as a way of ameliorating the problem. However, the minimum dose needed to fully charge SiO2 at liquid nitrogen temperature is well below the dose needed for holographic analysis. A more successful approach has been to coat the sample surface with carbon. In this paper, examples of effects of charging on holographic phase images will be described for a number of different cases, including thin oxide films and semiconductor samples. Methods Off-axis electron holography in the TEM is achieved by using a positively charged biprism to overlap the part of the electron wave that has passed only through vacuum with the part of the electron wave that has passed through the sample, as indicated schematically in Fig. 1 [1]. Fig. 1 Open in new tabDownload slide Schematic of TEM off-axis electron holography. Electrostatic biprism overlaps wave from sample with wave from vacuum to create interference fringe pattern. Fig. 1 Open in new tabDownload slide Schematic of TEM off-axis electron holography. Electrostatic biprism overlaps wave from sample with wave from vacuum to create interference fringe pattern. The off-axis electron holography described here was performed using a Philips CM200 FEG (field-emission electron gun) TEM operated primarily at 200 keV. This instrument was equipped with an electrostatic biprism in the selected-area aperture plane and a Lorentz minilens, located just below the bore of the lower objective-lens pole-piece [2]. This additional lens enabled large fields of view (∼0.3–0.5 μm) to be obtained for holographic viewing of semiconductor samples. Biprism voltages were in the range of 120–160 V with magnifications of 20–70 kX. A Gatan 794 multiscan CCD camera was used for digital recording. Electron holograms of the thin SiO film [3] were acquired in normal low magnification mode with the Lorentz lens switched off. Reference holograms with the sample removed were routinely recorded so that corrections could be made for non-linearities in the imaging and recording system [4]. The holograms were reconstructed using Digital Micrograph scripts. During observation, the doped silicon samples were usually rotated by ∼6° away from the [110] zone axis about the substrate surface normal and then slightly off the Kikuchi band to minimize diffraction effects. However, local thickness estimation via convergent beam electron diffraction (CBED) required zone-axis observation [5]. The semiconductor samples were taken from various implanted wafers, and they were prepared for TEM by low-angle wedge polishing using the MultiPrep apparatus made by Allied, followed by low-angle, low-voltage ion milling to remove any residual surface debris and to smooth the surface. After initial evaluation, carbon was deposited on the cross-sectional samples by thermal sputtering up to a thickness of ∼30–40 nm. The SiO-coated carbon sample was prepared, as described by Downing and Glaeser [3]. Results and discussion Figure 2a is a holographic phase image showing the corner of a wedge-polished cross-sectioned Si sample that had 100 nm of thermal oxide deposited on the wafer surface. The grayscale range is 2π for this phase image and the presence of multiple grayscale contours indicates a fringing electric field with a total phase shift of >12 rad. The phase of the electron wave is shifted by the presence of electrostatic potential changes, which appear to emanate from the Si–SiOx interface. Figure 2b shows a phase image of the same area after coating the sample with ∼30 nm of carbon. Here, no large fringing field is visible. Figure 2c compares line scans from a region of the sample away from the corner before and after carbon coating. Before carbon coating, the phase profile shows a downward ramp away from the sample surface. In principle, the phase shifts should be proportional to the electrostatic potential present in the sample; however, in this case the downward slope is due to the fringing electric field outside the sample. The position of a p–n junction, ∼75 nm from the Si surface, is indicated by arrows for the two cases, with n-type dopant near the surface while the bulk is p-type. If the slope of the uncoated profile is removed by subtracting an appropriate phase ramp, the size of the phase shift at the position of the junction is the same in both cases. In addition, the behavior of the phase profiles in the n-doped Si near the SiOx interface is very similar both before and after carbon coating, with no observable band bending in the Si associated with this interface, even when significant charge is present. The positive charge in the insulating layer attracts electrons in the Si layer adjacent to the interface. For n-type material, this represents an accumulation whose lateral extent is very small for highly doped material and is not noticeable at the resolution used here. Fig. 2 Open in new tabDownload slide (a) Holographic phase image showing cross-section of silicon p–n junction sample with 100 nm oxide layer on wafer surface. Contours indicate fringing electric field. (b) After carbon coating no fringing field is seen. (c) Line profiles from the phase images before and after carbon coating. Downward slope indicates fringing field. Position of junction is indicated. Fig. 2 Open in new tabDownload slide (a) Holographic phase image showing cross-section of silicon p–n junction sample with 100 nm oxide layer on wafer surface. Contours indicate fringing electric field. (b) After carbon coating no fringing field is seen. (c) Line profiles from the phase images before and after carbon coating. Downward slope indicates fringing field. Position of junction is indicated. This behavior is in contrast to the case where the dopant in the Si adjacent to the charging interface is p-type. Figure 3a shows a silicon sample that had layers of CoSi2 and TEOS (tetraethylorthosilicate) oxide deposited on the wafer surface prior to TEM sample preparation. Line profiles from the phase image before carbon coating (Fig. 3b) show that the phase rises at the oxide–vacuum interface and then begins to decline. Because of severe signal noise due to lack of coherence in this heavily scattering layer, the phase in the CoSi2 layer was undetermined for most sample thicknesses. However, the phase shift in the Si is apparently nearly flat. These phase shifts can be compared with the phase shift after carbon coating, also shown in Fig. 3b, where the phase mirrors the increasing thickness of the wedge-polished sample as expected. Fig. 3 Open in new tabDownload slide (a) Phase image of silicon p–n junction sample with silicide and oxide surface layers before carbon coating. (b) Line profiles of phase image in (a) (circles, before coating; squares, after coating). Fig. 3 Open in new tabDownload slide (a) Phase image of silicon p–n junction sample with silicide and oxide surface layers before carbon coating. (b) Line profiles of phase image in (a) (circles, before coating; squares, after coating). The different behaviors of the potential in the p-type Si layer near the silicide interface are shown in Fig. 4. In the absence of diffraction effects, the potential is derived from the phase shift by dividing by the thickness and an interaction constant, 0.00728 rad V−1 nm−1[6]. Figure 4 also includes a calculated potential shift based on SIMS measurements of the sample. It can be seen that the experimental potential accurately follows the predicted one when the sample is carbon coated. However, before carbon-coating the potential near the silicide interface exhibits band-bending consistent with inversion of the p-type Si, presumably due to charge transfer to the interface with the insulating TEOS oxide. Fig. 4 Open in new tabDownload slide Comparison of potential profiles derived from data of Fig. 3, with simulation based on SIMS. solid line, simulation; circles, before coating; squares, after coating. Junction position is arrowed. The rise in potential at the silicide interface for p-type silicon in uncoated sample indicates inversion. Fig. 4 Open in new tabDownload slide Comparison of potential profiles derived from data of Fig. 3, with simulation based on SIMS. solid line, simulation; circles, before coating; squares, after coating. Junction position is arrowed. The rise in potential at the silicide interface for p-type silicon in uncoated sample indicates inversion. Further experiments were performed to ascertain the degree and sign of the charging phenomenon in insulating silicon oxides. As described by Downing and Glaeser [3], holographic experiments were performed on an SiO thin film deposited on a carbon substrate and examined at liquid nitrogen temperature. Holograms were acquired using minimum dose techniques from an area that had been previously illuminated. Figure 5a shows a low-dose hologram (average counts = 0.95 electrons per pixel) where the previously illuminated area appears bright. Holographic interference fringes are not visible due to the very high noise of the image. Nevertheless, the hologram could still be reconstructed, and the unwrapped phase image is shown in Fig. 5b. The phase profile of the first hologram across the boundary between the illuminated and un-illuminated area is shown in Fig. 5c, and indicates a phase difference of 40 rad. The phase shift inside the illuminated region is higher than that outside, indicating that the sign of the charge is positive. This accumulated positive charge is most likely due to emission of secondary electrons from the insulating oxide. The maximum gradient of this phase shift corresponds to an electric field of 18 MV cm−1. This value is on the order of the breakdown voltage of SiO2. Two additional exposures were made, taking care to limit the electron dose. The profile of the same area as shown in Fig. 5b is shown in Fig. 5c after a total dose of 0.045 electrons nm−2 and does not show any phase difference across the boundary, indicating that the ‘outside’ region has charged up to the same degree as the ‘inside’. The implication of this result is that it will be impractical to avoid charging of dielectric layers by reducing the electron dose since the dose required to avoid charging will also result in holograms that are too noisy to be useful. Fig. 5. Open in new tabDownload slide (a) Low-dose hologram from edge of previously irradiated area (bright) arrow indicates direction of holographic fringes. (b) Unwrapped phase image from hologram in (a) with line profile along indicated direction. (c) Phase image and line profile from same area as (b) after total dose of 0.045 electron nm−2. Fig. 5. Open in new tabDownload slide (a) Low-dose hologram from edge of previously irradiated area (bright) arrow indicates direction of holographic fringes. (b) Unwrapped phase image from hologram in (a) with line profile along indicated direction. (c) Phase image and line profile from same area as (b) after total dose of 0.045 electron nm−2. Concluding remarks Electron beam irradiation in the TEM causes charging of dielectric layers in semiconductor device structures. These layers may charge up to the extent the insulator sustains maximum internal electric fields. Fringing electric fields near these layers complicate holographic phase analysis as the electron wave is phase-shifted outside the sample. The effect of charging layers on the analysis of adjacent semiconductor junction profiles may be severe, especially in the case of p-doped material where significant band bending and inversion have been observed. These charging effects are apparently eliminated by coating the sample with ∼30 nm of carbon. Presumably, this coating provides a path to ground and replenishes electrons lost to secondary emission. The electron holography was performed at the Center for High Resolution Electron Microscopy at Arizona State University. The author gratefully acknowledges the collaboration of Dr Ken Downing. Portions of this work were funded by the Semiconductor Research Corporation No. 942.001. References 1 Lichte H ( 1991 ) Electron image plane off-axis holography of atomic structures. Adv. Opt. El. Micr. 12 : 25 –91. Google Scholar 2 McCartney M R, Smith D J, Farrow R F C, and Marks R F ( 1997 ) Electron holography of epitaxial FePt films. J. App. Phys. 82 : 2461 –2465. Google Scholar 3 Downing K H and Glaeser R M ( 2004 ) Experimental characterization and mitigation of specimen charging on thin films with one conducting layer. Microsc. Microanal. , 10 : 783 –789. Google Scholar 4 de Ruijter W J and Weiss J K ( 1993 ) Detection limits in quantitative off-axis electron holography. Ultramicroscopy 50 : 269 –283. Google Scholar 5 Li J, McCartney M R, Dunin-Borkowski R E, and Smith D J ( 1999 ) Measurement of mean inner potential of germanium using off-axis electron holography. Acta Crystallogr. A 55 : 652 –658. Google Scholar 6 Reimer L ( 1991 ) Transmission Electron Microscopy. (Springer, Berlin). Google Scholar © The Author 2005. Published by Oxford University Press on behalf of Japanese Society of Microscopy. All rights reserved. For permissions, please email: journals.permissions@oxfordjournals.org
TI - Characterization of charging in semiconductor device materials by electron holography
JF - Journal of Electron Microscopy
DO - 10.1093/jmicro/dfi035
DA - 2005-06-01
UR - https://www.deepdyve.com/lp/oxford-university-press/characterization-of-charging-in-semiconductor-device-materials-by-PfOEwK6ueC
SP - 239
EP - 242
VL - 54
IS - 3
DP - DeepDyve
ER -