J Mater Sci: Mater Electron (2008) 19:S324–S327 DOI 10.1007/s10854-008-9653-x
EBIC imaging using scanning transmission electron microscopy: experiment and analysis Shigeyasu Tanaka Æ Hiroki Tanaka Æ Tadahiro Kawasaki Æ Mikio Ichihashi Æ Takayoshi Tanji Æ Koji Arafune Æ Yoshio Ohshita Æ Masafumi Yamaguchi
Received: 28 September 2007 / Accepted: 13 February 2008 / Published online: 5 March 2008 Ó Springer Science+Business Media, LLC 2008
Abstract The electron-beam-induced-current (EBIC) technique using scanning transmission electron microscopy (STEM) has been applied to the observation of grain boundaries in polycrystalline Si. It was shown that defects in thin regions can disappear or give bright contrast in EBIC images, but are distinct in STEM images. For the sample with a high carrier concentration, the surface effect was shown to dominate the EBIC current for a thin region. The bright contrast of the defects observed for the sample with a low carrier concentration can be attributed to the combination of the diffraction effect and the built-in electric field induced by the depletion of the entire thickness.
1 Introduction The electron-beam-induced-current (EBIC) technique using scanning transmission electron microscopy (STEM) is expected to provide information on both electrical activities and structural defects simultaneously. In the STEM–EBIC technique, an electron beam scans across a
S. Tanaka (&) M. Ichihashi T. Tanji EcoTopia Science Institute, Nagoya University, Furo-cho, Chikusa-ku, Nagoya 464-8603, Japan e-mail:
[email protected] H. Tanaka T. Kawasaki Department of Electronics, Nagoya University, Furo-cho, Chikusa-ku, Nagoya 464-8603, Japan K. Arafune Y. Ohshita M. Yamaguchi Toyota Technological Institute, Hisakata, Tempaku-ku, Nagoya 488-8511, Japan
123
thin sample with a pn junction or a Schottky contact. Minority carriers generated by the electron beam diffuse and some recombine or become trapped in defects. Minority carriers that reach the depletion region are driven by the built-in electric field and become a short-circuit current. Therefore, by measuring short-circuit current and synchronizing images with the STEM scanning signal, the local electrical activity of defects can be mapped. The STEM–EBIC technique was first demonstrated in the late 1970s by Sparrow and Valdre [1] in their correlation of crystal defects and the electrical properties of Si transistors. Following their demonstration, several researchers have used this technique in analyzing various electrical properties: dislocation and nonradiative recombination properties of GaAlAsP [2], the electrical activity of microstructural defects in polycrystalline Si [3], and inhomogeneities in Si pn junctions [4]. The STEM–EBIC technique should be a powerful tool for the study of the electrical activity of defects, however, its usage seems to be limited. We carried out STEM–EBIC using polycrystalline Si as a sample, and calculated EBIC using a cylindrical model to better understand defect contrast in STEM–EBIC images. It was found that defects in thin regions can disappear or give bright contrast in EBIC images, but are clear in STEM images.
2 Method Poly-Si crystals (p-type) about four inches in diameter were grown by the cast technique [5]. They were borondoped at a concentration of either 1016 or 1018 cm-3. As-grown ingots were sliced into wafers. Then, the wafers were polished, and an ohmic contact was formed on one side of the wafer. For convenience in STEM–EBIC, the
J Mater Sci: Mater Electron (2008) 19:S324–S327
wafers were Secco-etched to form etch pits on the other side [5]. From the wafers, small rectangular pieces were cut out. The cutting out was carried out in such a way that each sample piece contains at least one etch pit line. Next, the sample was thinned by standard plan-view thinning through dimpling followed by ion milling. Dimpling was performed from part of the ohmic contact side. Ion milling was performed on both sides. Finally, the Schottky electrode was formed on the other side by Ti deposition. The thinned sample was attached to an EBIC holder designed to allow electrical contact with the sample. The STEM–EBIC technique was carried out using a Hitachi H-8000 accelerated at 200 kV. For the determination of the types of
S325
structural defect, the transmission electron microscopy mode of the STEM system was used. The EBIC signal was converted to a voltage signal with a Keithley current amplifier, and the voltage signal was acquired with a computer for the EBIC imaging. EBIC was calculated using the cylindrical model shown in Fig. 1. An electron beam accelerated at 200 kV is assumed to enter the model along the axes, generating electron–hole pairs. Schottky contact is assumed to be made on the top surface. The width of the depletion region is not considered in this model. But, a finite width of the depletion region can affect the experimentally obtained EBIC as described later. A collection of excess minority carriers (electrons) by this junction field gives rise to EBIC. The transport of excess minority carriers is described by a steady-state-diffusion equation [6], and is solved numerically using the finite element method. Minority carriers are assumed to be rapidly extracted from the Schottky contact side, and to recombine quickly on the other side. An empirical expression was used for the generation of the carriers by the electron beam [6]. The presence of line defects was treated as a columnar region where recombination rate is enhanced, i.e., the diffusion length L is reduced. Such a defect region of 2L width was embedded along the axes of the cylindrical model. For a perfect region without defects, the L was taken as infinity. EBIC was calculated for various L values.
3 Results and discussion
Fig. 1 Schematic diagram for modeling calculation of EBIC currents
Figure 2a and b shows a STEM dark-field (DF) image of a low-angle-tilt grain boundary and a simultaneously observed EBIC image, respectively. The boron concentration of the sample is 1018 cm-3. The sample’s thickness increases from the lower left to the right upper corner in Fig. 2. The sample’s edge seems to be shifted to the thicker
Fig. 2 (a) STEM DF image and (b) corresponding EBIC image of sample having boron concentration of 1018 cm-3
123
S326
region in the EBIC image, but this is simply due to a change in the sample’s thickness. Note that although the grain boundary continues up to the near-edge region as seen in the STEM image, the EBIC contrast of the grain boundary almost disappears in the thin region, as indicated by the arrow in the EBIC image. The thickness of this region was estimated to be about 1 lm as deduced from a cross-sectional observation of a similar sample by scanning electron microscopy (SEM). Because of the high concentration of holes, the width of the depletion layer should be much smaller than the thickness of the region. Thus, the
Fig 3 Calculated EBIC as function of sample thickness in the cases with and without defects. EBIC is normalized by the current in the case without defects at a thickness of 5 lm Fig. 4 (a) STEM DF image and (b) corresponding EBIC image of sample having boron concentration of 1016 cm-3
123
J Mater Sci: Mater Electron (2008) 19:S324–S327
disappearance of defect contrast must be due to the surface effect. To confirm this, we calculated EBIC using the cylindrical model shown in Fig. 1. Figure 3 shows the calculated EBIC as a function of the thickness for the cases with and without defects. The diffusion length in the case without defects was taken as infinity. The EBIC in the case without defects increases linearly with the thickness. With increasing thickness, the EBIC in the case with defects deviate from the linear relationship depending on the diffusion length of the defect region. This is due to recombination in the defect region. However, the deviations are very small for thicknesses less than 1 lm. In this thickness range, the effect of the surface is stronger than that of recombination in the defect region. Thus, in practice, defect contrast is expected to disappear in such thin regions, consistent with our observations. We also observed a sample having a boron concentration of 1016 cm-3. Figure 4a and b shows a STEM DF image and a simultaneously observed EBIC image, respectively. A low-angle-tilt grain boundary is indicated by the arrow in the STEM DF image. This grain boundary is also seen in the EBIC image, but the contrast changes with sample thickness in this case. In the thick region, the grain boundary shows an expected dark contrast indicating enhanced recombination, but near the sample edge, the grain boundary shows an unexpected bright contrast. The sample thickness where the grain boundary shows bright contrast was estimated to be about 1 lm, which roughly corresponds to the expected depletion layer width for this sample. Because of the built-in electric field, minority carriers generated within the depletion region are expected to be less affected by recombination at defects. Thus, defect contrast is expected to disappear in depletion region. To examine the cause of the bright contrast, an STEM– EBIC observation was carried out for various sample tilt angles. The result showed that the contrast of the grain boundary in the thin region changed with sample tilt angle.
J Mater Sci: Mater Electron (2008) 19:S324–S327
Thus, we consider that the bright contrast is due to the diffraction effects [7]. Near the grain boundary, there must be a region where a strongly diffracted beam is excited. If this beam travels through a longer path in a thicker region, more electrons will be generated and thus stronger EBIC signals will be obtained. However, this requires a precipitous sample edge, because the Bragg angles are very small. This may explain the reason why such a bright contrast was not obtained in the sample in Fig. 2. Also note that the grain boundary in the EBIC image consists of dots rather than a line, although the grain boundary is a straight line in the STEM image. Thus, the electrically active sites should align as dots along the grain boundary. These active sites may be due to the accumulation of impurities or strain. But, the reason why such a dot-like contrast was not observed in the sample in Fig. 2 is not clear. Further research is required to clarify these points.
4 Summary The STEM–EBIC technique was carried out using polycrystalline Si as a sample. It was shown that defects in thin regions can disappear or give bright contrast in EBIC
S327
images, but are clear in STEM images. For the sample with a high carrier concentration, the surface effect was shown to dominate the EBIC current for a thin region. For the sample with a low carrier concentration, the bright contrast of the defects can be attributed to the combination of the diffraction effects and the built-in electric field induced by the depletion of the entire thickness. Acknowledgments This work was partly supported by the Incorporated Administrative Agency New Energy and Industrial Technology Development Organization (NEDO) under Ministry of Economy, Trade and Industry (METI), Japan.
References 1. T.G. Sparrow, U. Valdre, Philos. Mag. 36, 1517 (1977) 2. P.M. Petroff, R.A. Logan, A. Savage, Phys. Rev. Lett. 44, 287 (1980) 3. C. Cabanel, J.Y. Laval, J. Appl. Phys. 67, 1425 (1990) 4. C. Cabanel, H. Maya, J.Y. Laval, Philos. Mag. Lett. 79, 55 (1999) 5. K. Arafune, T. Sasaki, F. Wakabayashi, Y. Terada, Y. Ohshita, Physica B 376–377, 236 (2006) 6. C. Donolato, Phys. Stat. Sol. (a) 65, 649 (1981) 7. G.C. Perreault, S.L. Hyland, D.G. Ast, Sol. Energy Mater. Sol. Cells 30, 309 (1993)
123