- Email: [email protected]
SiO2 interface using laser-assisted atom probe tomography and transmission electron microscopy
Micron 58 (2014) 32–37
Contents lists available at ScienceDirect
Micron journal homepage: www.elsevier.com/locate/micron
Study of vertical Si/SiO2 interface using laser-assisted atom probe tomography and transmission electron microscopy J.H. Lee a,b , B.H. Lee c , Y.T. Kim a , J.J. Kim b , S.Y. Lee b , K.P. Lee b , C.G. Park a,c,∗ a b c
Department of Materials Science and Engineering (MSE), Pohang University of Science and Technology (POSTECH), Pohang 790-784, Republic of Korea Semiconductor Business, Samsung Electronics, Hwasung 445-701, Republic of Korea National Institute for Nanomaterials Technology (NINT), Pohang University of Science and Technology (POSTECH), Pohang 790-784, Republic of Korea
a r t i c l e
i n f o
Article history: Received 5 August 2013 Received in revised form 13 November 2013 Accepted 16 November 2013 Keywords: Atom probe Shallow trench isolation Preferential evaporation Interface reaction
a b s t r a c t Laser-assisted atom probe tomography has opened the way to three-dimensional visualization of nanostructures. However, many questions related to the laser–matter interaction remain unresolved. We demonstrate that the interface reaction can be activated by laser-assisted field evaporation and affects the quantification of the interfacial composition. At a vertical interface between Si and SiO2 , a SiO2 molecule tends to combine with a Si atom and evaporate as a SiO molecule, reducing the evaporation field. The features of the reaction depend on the direction of the laser illumination and the inner structure of tip. A high concentration of SiO is observed at a vertical interface between Si and SiO2 when the Si column is positioned at the center of the tip, whereas no significant SiO is detected when the SiO2 layer is at the center. The difference in the interfacial compositions of two samples was due to preferential evaporation of the Si layer. This was explained using transmission electron microscopy observations before and after atom probe experiments. © 2013 Elsevier Ltd. All rights reserved.
1. Introduction The quality of the interface in semiconductor devices critically influences the device performance (Cavin et al., 2000; Thompson and Parthasarathy, 2006), so there is great interest in interpreting the interfacial composition at the atomic scale. Atom probe tomography (APT) has received a great deal of attention from semiconductor researchers for its three-dimensional visualization of nanostructures and its ppm-level sensitivity (Kelly et al., 2007); it is easily adaptable to the analysis of semiconductor devices, which are entering the 10-nm-level design rule regime and are beginning to employ 3D structures. However, despite the recent application of femtosecond pulsed lasers (Gilbert et al., 2011), care must be taken to ensure that the interpretation of the interfacial composition is reliable. Heterogeneous structures reportedly raise unavoidable artifacts in the reconstruction of the AP data, and various reasons for these artifacts have been suggested (Oberdorfer and Schmitz, 2011; Vurpillot et al., 2000; Larson et al., 2010). First, exposing the structure to a laser pulse to assist field evaporation causes asymmetric evaporation of atoms with respect to the laser direction (Marquis
∗ Corresponding author at: Department of Materials Science and Engineering (MSE), Pohang University of Science and Technology (POSTECH), Pohang 790-784, Republic of Korea. Tel.: +82 10 5310 2139. E-mail addresses: [email protected] (J.H. Lee), [email protected] (C.G. Park). 0968-4328/$ – see front matter © 2013 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.micron.2013.11.003
et al., 2011a; Sha et al., 2008). When the thermal diffusion length is thinner than the tip diameter, the thermal distribution of the tip becomes anisotropic, which induces an alteration in the shape of the tip (Bunton et al., 2007). In oxides, which have lower thermal diffusivity than metals, severe preferential evaporation occurs on the side illuminated by the laser (Müller et al., 2011). Moreover, the laser exposure changes the nature of multiple events, so the compositional information varies with the position with respect to the laser (Müller et al., 2012). Second, the differences in the evaporation field among the layers also cause the surface curvatures of layers to change from the initial statuses (Marquis et al., 2011a). This continuous evolution of the tip shape results in local magnification and disordering of the field evaporation sequence (Oberdorfer and Schmitz, 2011). As a result, the atomic concentration across the interface cannot be quantitatively measured when the evaporation field differs greatly. Third, the dependence of the ion trajectory on the crystallographic orientation has been widely recognized since field ion microscopy was initially used (Gault et al., 2010a; Miller et al., 1989). The difference in the evaporation field depending on the crystallographic orientation leads to local changes in the radius of curvature of the specimen and to local magnification (Gault et al., 2010a). These phenomena can be observed by transmission electron microscopy (TEM). When a SiO2 layer on a Si substrate is evaporating, the lower Si substrate evaporates before the upper SiO2 layer owing to the low evaporation field of Si. Fig. 1(a)–(c) are TEM images taken after laser-assisted field evaporation. In Fig. 1(a), the red dotted areas are preferentially evaporated regions of the
J.H. Lee et al. / Micron 58 (2014) 32–37
33
Fig. 1. (a) TEM image of Si/SiO2 (100 nm) multilayered sample after AP experiment. TEM image was recorded during the (temporarily paused) AP experiment. Pulsed laser illumination came from the right side. The red dotted area represents the preferentially evaporated Si area, which formed a flat surface in the (1 1 1) plane (the close-packed plane). High-resolution images of (b) left and (c) right sides. The native oxide was grown on the Si surface because of the strong electron beam and vacuum breaking. The preferentially evaporated Si surface formed a (1 1 1) facet. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)
Si substrate. The laser illumination came from the right side of the tip in the TEM image, and this is evident from a comparison of the radii of curvature on the left and right sides. Because the right side was exposed directly to the pulsed laser, the radius of curvature there is larger than that on the left side. Additionally, the facet of the (1 1 1) surface exposed at the Si layer means that the field evaporation is strongly related to the crystallographic orientation. Fourth, surface migration of the atoms on the tip surface degrades the lateral resolution (Kellogg et al., 1978; Gault et al., 2010b). Early research demonstrated the wandering of surface atoms under nanosecond laser pulses (Gault et al., 2006). If surface migration is possible in laser-assisted APT (Gault et al., 2012; Cerezo et al., 2007), we can assume that chemical reactions requiring low activation energy could also be activated. Because these reactions would also disturb the compositional quantification, this aspect of the treatment should be considered. In this paper, we will demonstrate the interface reaction at the Si/SiO2 interface and show that it depends on the inner structure of the tip by using TEM observations.
2. Experimental Shallow trench isolation (STI) was chosen for the interface analysis [Fig. 2(a)]. The STI structure is commonly used to isolate neighboring cells. The STI region is filled with SiO2 by chemical vapor deposition after a dry etch process. The width of the Si column and the depth of the STI are 25 nm and 250 nm, respectively. Two types of sample from the same structure were prepared using a focused ion beam. The Si column is centered in sample A, and the SiO2 STI is centered in sample B, as shown in Fig. 2(a). JEM-2100F was used for TEM observation. The APT analysis was performed by LAWATAP (laser-assisted wide-angle tomographic atom probe) developed by CAMECA. A 343-nm, 100-kHz UV laser was set to 0.09 J/pulse at 30 K (spot size ∼30 m). The UV laser was illuminated perpendicular to the Si column [from the left or right direction in Fig. 2(a)]. The standing voltage was controlled automatically with respect to the detection flux in order to acquire AP data at the AP image. It provides three reconstruction algorithms: the voltage-based approximation, cone-shape approximation, and tip-shape based approximation. Among them, the tip-shape-based
Fig. 2. (a) TEM image of shallow trench isolation (STI). The width of the SiO2 layer is around 30 nm, and the depth of the STI is around 250 nm. (b) Side view of AP image from area A in (a). (c) Side view of AP image from area B in (a). Blue indicates Si; red indicates SiO2 . (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)
34
J.H. Lee et al. / Micron 58 (2014) 32–37
Fig. 3. (a) Top view of AP image from sample A. (b) Top view of AP image from sample B. Blue represents Si; red indicates SiO2 . (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)
approximation was adopted for the reconstruction of our samples. When a stiff voltage change occurs during field evaporation, the voltage-based approximation does not reflect the exact tip image. The cone-shape approximation is also not effective for a multilayered structure because it assumes a homogeneous structure. The most reasonable APT image was manually reconstructed by the tip-shape-based approximation method. The width of the Si column is calibrated at multiple points along the longitudinal direction based on the TEM image until the reconstructed image is identical to the TEM image. Note that the dimension of the outer layer (the SiO2 layer in sample A and the Si layer in sample B) could not be calibrated, because it is impossible to identify the edge of the tip when the outer edge of the tip has evaporated outside of the detector area. Calibrated AP images from samples A and B are shown in Fig. 2(b) and (c), respectively. Areas of Si and SiO2 are shown in blue and red, respectively. We admit that this method cannot perfectly reconstruct the real structure of the tip. However, artifacts due to the laser are unavoidable, so we tried to match the reconstructed images to the TEM images taken just before the AP measurement as closely as possible. 3. Results and discussion To examine the Si/SiO2 interface, views of samples A and B from above are shown in Fig. 3(a) and (b), respectively. Again, blue and red represents the Si column and SiO2 region, respectively. The outermost region is cut according to Muller’s suggestion (Müller et al., 2012); the image distortion of the outermost region increases
because the local magnification effect disrupts the reconstruction of the three-dimensional structure. Nevertheless, the image distortion due to the trajectory aberration remains because the evaporation angle changes according to the environment surrounding the evaporating atoms. Therefore, we chose the central area for interface interpretation. Fig. 4 shows the concentration profile across the Si/SiO2 interface in both samples. The rough Si:O ratio at the bulk-like SiO2 layer was 1:2. The concentration in the bulk area between the two samples did not differ significantly, whereas the interface concentration shows considerable disagreement between samples A and B and between the left and right interfaces. Note that the oxygen concentration in the Si area does not seem to be negligible. Because it is unlikely that O atoms would have existed in the pure Si column, we assumed that O atoms in the Si layer are attributable to an overlap in the evaporation trajectory between the two layers. That is, O atoms detected in the Si area were detached from the SiO2 layer because the evaporation trajectory of the SiO2 layer is wider than the adjacent Si area. According to the study of Bachhav et al. (2012), laser energy is absorbed only on the illuminated side of the tip apex, and the energy is transferred to the opposite side by thermal diffusion. A recent study of the absorption of UV and green laser light by Koelling et al. (2011) also showed that the absorption of UV laser light is highly asymmetric compared with that of green laser light. Moreover, the UV laser ( = 343 nm) was absorbed almost entirely at the top surface of the illuminated side, resulting in asymmetric heating of the tip. The asymmetric absorption of laser energy will cause the interfacial quantification to vary with the geometric placement of the interface and laser illumination. Regarding this assumption, we compared two interfaces located on the laserilluminated side and the opposite side. Fig. 5 demonstrates the silicon monoxide concentration profile across the Si/SiO2 interface in the two samples. SiO was chosen for the interface indicator owing to its dominance. We identified the peaks at 22 atomic mass units (amu) and 44 amu as SiO2+ and SiO+ , respectively, because the system consists only of silicon and oxygen. Note that around 3.3–3.5 at.% of SiO is measured in the SiO2 region. We focus on two important facts in Fig. 5. First, sample A, the Si-centered tip, has a high SiO concentration at the left and right interfaces, whereas sample B, the SiO2 -centered tip, has a relatively small peak at each interface. The small peak at the interface can be understood to come from the inherent suboxide at the interface. Owing to imperfect formation of the interface, a transition
Fig. 4. Lateral profiles of the atomic concentration across the interface between Si and SiO2 in (a) sample A and (b) sample B. The lateral dimension is not accurate because of the local magnification effect. At the SiO2 layer in both samples, the Si:O ratio is roughly 1:2. Note that oxygen atoms are found in the Si area because of their wide evaporation angle. Other minor constituents are omitted for clarity.
J.H. Lee et al. / Micron 58 (2014) 32–37
Fig. 5. Compositional profiles of silicon monoxide across the vertical interface of Si and SiO2 in sample A (silicon-centered sample, blue squares) and sample B (SiO2 centered, red circles). Pulsed laser illumination originates at the right side of the graph. A significant amount of SiO exists at the interface in sample A, whereas weak peaks were detected at the interface in sample B. Additionally, a larger SiO peak is detected on the right (the laser illumination side) in both samples. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)
layer consisting of substoichiometric Si oxide inherently exists at the interface between the Si and SiO2 layers, and its existence has been proven by previous electron energy loss spectroscopy analysis (Muller et al., 1999). However, it is unlikely that the interface in sample A contains twice the SiO of that in sample B. By iterating the same experiment several times, we were able to conclude that the discrepancy is not because of an intrinsic difference, but rather it is due to the dependence of laser-assisted field evaporation on the experimental geometry. When no supply of materials from outside is expected, it is worth considering possible reactions between internal constituents that could occur in the transition layer. That is, the interfacial reaction could be activated by laser-assisted field evaporation, and this feature depends on the inner structure of the tip as well as the laser illumination direction. In contrast, dissociation should also occur at the interface; there is no reason to assume that decomposition of SiO2 in the external field would not happen near the interface. Second, the amount of detected interfacial SiO depends on the direction of laser illumination, and the interface on the exposed side contains more SiO in both samples A and B. It could also be proven that the interfacial composition depends on the direction of the laser illumination because the difference in the peak height between the left and right interfaces follows the tip when it is rotated 180◦ (not shown here). Therefore, it is logical to assume that the higher concentration of SiO is due to synthesis at the interface, and the amount of synthesis is related to the laser–matter interaction. This assumption needs to satisfy the prerequisite that the activation energy for combination is smaller than that for SiO2 field evaporation. Otherwise, SiO2 will not combine with Si but instead will evaporate as usual, even if the laser impact is sufficient to combine the two components. In our specimen, the reaction between silicon and silicon dioxide, Si + SiO2 → 2SiO, is the most likely candidate, as it was reported in the 1990s by several groups (Tromp et al., 1985; Streit and Allen, 1987; Harp et al., 1990). Streit and Allen proved that Si atoms that diffused from the substrate are involved in the rapid decomposition of the thin SiO2 layer (Streit and Allen, 1987). A thick SiO2 layer on a Si substrate barely dissociates by itself at 900 ◦ C, but an ultrathin SiO2 layer less than 1 nm thick could be easily removed at a lower temperature. The activation energies (Ea ) for SiO2 desorption both with and without the assistance of Si were identified in an experiment in which the SiO2 layer
35
Fig. 6. TEM images of (a) sample A and (b) sample B before AP experiment. Tip surface became amorphous because of ion-beam milling. Images were taken as quickly as possible to prevent various problems due to electron beam with 200-kV acceleration.
was removed during exposure to a silicon ion beam. The preferred reaction is SiO synthesis rather than SiO2 desorption because the activation energies for desorption are Ea = 3.3–4.0 eV and 0.84 eV for thermal SiO2 removal and Si-assisted removal, respectively (Streit and Allen, 1987). Therefore, the prerequisite mentioned above is satisfied. We can conclude that SiO2 tends to combine and evaporate with a Si atom rather than dissociate to Si and O atoms, so it could decrease the energy required for evaporation. To explain why different amounts of SiO were detected at the interface in the two samples, we prepared the same set of samples and recorded TEM images of them before and after the AP experiment. Fig. 6(a) and (b) are TEM images taken from samples A and B, respectively, before the AP experiment. The electron beam intensity used for TEM image acquisition was minimized to avoid beam damage and hydrocarbon contamination before AP operation. High-resolution images were not taken for the same reason. The summit of sample A, single-crystalline Si, became amorphous despite the low energy of the ion-beam milling (2 kV). Samples A and B initially had similar radii of curvature: 17 nm and 22 nm, respectively. However, they evolved to take significantly different shapes, as shown in Fig. 7(a) and (b). After the initial period of field evaporation, a groove in the center of the tip apex developed in the Si-centered sample, A; in contrast, a sharp tip apex formed in the SiO2 -centered sample, B. The Si layer, which has a low evaporation field, was field-evaporated before the SiO2 layer, which has a high field. This produced a groove at the summit of sample A. Because the curvature of the tip edge is inversely proportional to the field required to evaporate the material, the preferential evaporation of Si atoms increases the radius of curvature of the Si layer until the field applied to each material is equivalent. This will subsequently alter the trajectory of each element. The groove constrains the evaporation angle of silicon atoms, narrowing the trajectory of the evaporated Si atoms. In the same manner, the evaporation angle of the SiO2 layer is widened because of the groove. Preferential evaporation of Si also occurs in sample B. However, the opposite arrangement of the layers induced the distinctive evolution of the tip apex. The pulsed laser illumination was at the right side of both samples in the TEM images. The direction of the laser illumination could easily be confirmed by observing the tip shape. The red-dotted envelops in Fig. 8(a) and (b) represents
36
J.H. Lee et al. / Micron 58 (2014) 32–37
Fig. 7. TEM images of (a) sample A and (b) sample B after AP experiment. AP operation was stopped during evaporation to observe the shape evolution with respect to the inner structure. When the Si is centered, the preferentially evaporated Si layer forms a groove at the center of the tip. When the SiO2 is centered, the sample evolves into a sharp tip.
the radii of curvature on the left and right sides of the tip apex. In sample A, the radii of curvature were 25 and 35 nm at the left and right corners, respectively. In sample B, the radii of curvature were 35 and 45 nm at the left and right corners, respectively. The laser side developed a larger radius of curvature in both samples owing to the non-uniform field distribution (Marquis et al., 2011b). We believe the difference in how the shapes evolve explains the difference in interfacial composition between the two samples. The silicon layer confined by the SiO2 layer cannot evaporate easily as a pure silicon layer but develops a flat surface, as shown in Figs. 7(a) and 8(a). For this reason, considering the Si-assisted SiO2 evaporation of sample A, many Si atoms can combine with SiO2 molecules at the atomic distance of the SiO2 surface because preferential evaporation of Si atoms was retarded, as we described above. In contrast, SiO2 molecules in sample B had to evaporate
unassisted because adjacent Si atoms had already evaporated, as shown in Fig. 8(b). Because of these contrasting circumstances, each interface shows a different interfacial composition. In addition, the thermal behavior of the two materials could enhance the compositional difference. The thermal conductivity of Si is known to be two orders greater than that of SiO2 (Kim et al., 1999). Because of the geometry of the tip and laser, the top surface exposed to the UV laser became SiO2 in sample A and Si in sample B. Note that the diameter of the pulsed laser is large enough to cover the entire area of the tip. In sample B, the heat generated by the pulsed laser rapidly dissipated through the shank or evaporated the surface material. However, the absorbed heat was used to combine Si and SiO2 in sample A. The difference between the left and right interfaces could be explained by the absorption of the UV laser. As we mentioned above, the absorption depends strongly on the illumination direction and occurs only at the top surface (Bachhav et al., 2012; Koelling et al., 2011). This non-uniformity can explain why SiO synthesis occurred mainly at the interface of the laserilluminated side. In order to reinforce this claim, we conducted an additional experiment on the thermal effect of the pulsed UV laser on the tip surface. Fig. 9(a) and (b) are TEM images taken before and after laser illumination on the same STI structure. The power of the laser was the same as in the previous experiment. The sample was milled by a 30-kV Ga ion beam to intentionally damage the tip surface. Indeed, the surface of the tip had been about 25 nm deep so that the single crystalline surface became amorphous, as shown in Fig. 9(a). After laser exposure for 1 h (standing voltage was sustained at 5 kV), epitaxial recrystallization had occurred at the amorphous surface. It should be noted that the epitaxial recrystallization occurred at over 480 ◦ C by thermal annealing (not shown here). Although it is hard to explain the contribution of the standing voltage to the thermal reaction, it is reasonable to claim that the illumination of the pulsed laser possibly caused the thermal reaction. In this work, we observed the STI structure using APT and TEM. The interfacial Si and SiO2 composition was not consistent, but depended on the laser, inner structure of the tip, reaction kinetics, and geometric relationship between the laser and the sample. Because current reconstruction algorithms do not fully reflect the effect of the pulsed laser yet, these conditions should be considered when evaluating the interfacial composition of multilayered structures.
Fig. 8. High-resolution TEM images of (a) sample A and (b) sample B after AP experiment. Tip surface was slightly oxidized in the TEM chamber because of electron beam exposure. Pulsed laser illuminated the right sides of both tips. Red-dotted lines represent the radii of curvature of tip edges. In both samples, the right edges developed a larger radius of curvature due to asymmetric absorption of laser. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)
J.H. Lee et al. / Micron 58 (2014) 32–37
37
Fig. 9. TEM images of the STI structure (a) before pulsed laser exposure and (b) after laser exposure. Since the sample was intentionally damaged by the Ga ion beam, the surface of the Si layer became amorphous. Owing to laser exposure, the laser illuminated area had recrystallized with an epitaxial relationship. The epitaxial recrystallization was observed after laser illumination for 1 h. It is worth mentioning that the recrystallized length from the apex is approximately 350 nm, which is similar to the wavelength of the UV laser (343 nm).
4. Conclusions In summary, an STI structure consisting of Si and SiO2 was studied by APT and TEM. The interface composition was found to change depending on the arrangement of the AP tip. At the interface, SiO2 tends to combine and evaporate with Si when Si atoms are atomically close to the evaporation site. This Si-assisted SiO2 evaporation is attributed to a greater reduction in the activation energy for SiO combination than in that for dissociative evaporation, leading to considerable modification of the interfacial composition during AP analysis. TEM observations proved that the amount of interfacial SiO depends on the shape evolution of the tip. The use of a pulsed laser is essential for AP analysis of insulating materials. However, its role should not be dominant but rather assistive because its influence is not fully understood. Therefore, in compositional analysis of heterogeneous structures using an AP, the laser–matter interaction must be prudently considered to reduce erroneous analyses. Acknowledgements This research was supported by Samsung Electronics Co. The authors thank NINT for supplying the analysis equipment. References Bachhav, M.N., Danoix, R., Vurpillot, F., Hannoyer, B., Ogale, S.B., Danoix, F., 2012. J. Appl. Phys. 111, 064908. Bunton, J.H., Olson, J.D., Lenz, D.R., Kelly, T.F., 2007. Microsc. Microanal. 13, 418. Cavin III, R.K., Herr, D.J.C., Zhirnov, V.V., 2000. J. Nanopart. Res. 1, 213.
Cerezo, A., Clifton, P.H., Gomberg, A., Smith, G.D.W., 2007. Ultramicroscopy 107, 720. Gault, B., Menand, A., de Geuser, F., Deconihout, B., Danoix, R., 2006. Appl. Phys. Lett. 88, 114101. Gault, B., La Fontaine, A., Moody, M.P., Ringer, S.P., Marquis, E.A., 2010a. Ultramicroscopy 110 (9), 1215. Gault, B., Müller, M., La Fontaine, A., Moody, M.P., Shariq, A., Cerezo, A., Ringer, S.P., Smith, G.D.W., 2010b. J. Appl. Phys. 108, 044094. Gault, B., Danoix, F., Hoummada, K., Mangelinck, D., Leitner, H., 2012. Ultramicroscopy 113, 182. Gilbert, M., Vandervorst, W., Koelling, S., Kambham, A.K., 2011. Ultramicroscopy 111, 530. Harp, G.R., Han, Z.L., Tonner, B.P., 1990. Phys. Scripta T31, 23. Kellogg, G.L., Tsong, T.T., Cowan, P., 1978. Surf. Sci. 70, 485. Kelly, T.F., Larson, D.J., Thompson, K., Alvis, R.L., Bunton, J.H., Olson, J.D., Gorman, B.P., 2007. Annu. Rev. Mater. Res. 37, 681. Kim, J.H., Feldman, A., Novotny, D., 1999. J. Appl. Phys. 86, 3959. Koelling, S., Innocenti, N., Schulze, A., Gilbert, M., Kambham, A.K., Vandervorst, W., 2011. J. Appl. Phys. 109, 104909. Larson, D.J., Geiser, B.P., Prosa, T.J., Gerstl, S.S.A., Reinhard, D.A., Kelly, T.F., 2010. J. Microsc. 243, 15. Marquis, E.A., Geiser, B.P., Prosa, T.J., Larson, D.J., 2011a. J. Microsc. 241, 225. Marquis, E.A., Geiser, B.P., Prosa, T.J., Larson, D.J., 2011b. J. Microsc. 241, 225–233. Miller, M.K., Hetherington, M.G., Burke, M.G., 1989. Metall. Trans. A 20 (12), 2651. Muller, D.A., Sorsch, T., Moccio, S., Baumann, F.H., Evans-Lutterodt, K., Timp, G., 1999. Nature 399 (24), 758. Müller, M., Saxey, D.W., Smith, G.D.W., Gault, B., 2011. Ultramicroscopy 111, 487. Müller, M., Smith, G.D.W., Gault, B., Grovenor, C.R.M., 2012. J. Appl. Phys. 111, 064908. Oberdorfer, C., Schmitz, G., 2011. Microsc. Microanal. 17, 15. Sha, G., Cerezo, A., Smith, G.D.W., 2008. Appl. Phys. Lett. 92, 043503. Streit, D.C., Allen, F.G., 1987. J. Appl. Phys. 61, 2894. Thompson, S.E., Parthasarathy, S., 2006. Mater. Today 9 (6), 20. Tromp, R., Rubloff, G.W., Balk, P., LeGoues, F.K., van Loenen, E.J., 1985. Phys. Rev. Lett. 55 (21), 2332. Vurpillot, F., Bostel, A., Blavetter, D., 2000. Appl. Phys. Lett. 76, 3127.
Contents lists available at ScienceDirect
Micron journal homepage: www.elsevier.com/locate/micron
Study of vertical Si/SiO2 interface using laser-assisted atom probe tomography and transmission electron microscopy J.H. Lee a,b , B.H. Lee c , Y.T. Kim a , J.J. Kim b , S.Y. Lee b , K.P. Lee b , C.G. Park a,c,∗ a b c
Department of Materials Science and Engineering (MSE), Pohang University of Science and Technology (POSTECH), Pohang 790-784, Republic of Korea Semiconductor Business, Samsung Electronics, Hwasung 445-701, Republic of Korea National Institute for Nanomaterials Technology (NINT), Pohang University of Science and Technology (POSTECH), Pohang 790-784, Republic of Korea
a r t i c l e
i n f o
Article history: Received 5 August 2013 Received in revised form 13 November 2013 Accepted 16 November 2013 Keywords: Atom probe Shallow trench isolation Preferential evaporation Interface reaction
a b s t r a c t Laser-assisted atom probe tomography has opened the way to three-dimensional visualization of nanostructures. However, many questions related to the laser–matter interaction remain unresolved. We demonstrate that the interface reaction can be activated by laser-assisted field evaporation and affects the quantification of the interfacial composition. At a vertical interface between Si and SiO2 , a SiO2 molecule tends to combine with a Si atom and evaporate as a SiO molecule, reducing the evaporation field. The features of the reaction depend on the direction of the laser illumination and the inner structure of tip. A high concentration of SiO is observed at a vertical interface between Si and SiO2 when the Si column is positioned at the center of the tip, whereas no significant SiO is detected when the SiO2 layer is at the center. The difference in the interfacial compositions of two samples was due to preferential evaporation of the Si layer. This was explained using transmission electron microscopy observations before and after atom probe experiments. © 2013 Elsevier Ltd. All rights reserved.
1. Introduction The quality of the interface in semiconductor devices critically influences the device performance (Cavin et al., 2000; Thompson and Parthasarathy, 2006), so there is great interest in interpreting the interfacial composition at the atomic scale. Atom probe tomography (APT) has received a great deal of attention from semiconductor researchers for its three-dimensional visualization of nanostructures and its ppm-level sensitivity (Kelly et al., 2007); it is easily adaptable to the analysis of semiconductor devices, which are entering the 10-nm-level design rule regime and are beginning to employ 3D structures. However, despite the recent application of femtosecond pulsed lasers (Gilbert et al., 2011), care must be taken to ensure that the interpretation of the interfacial composition is reliable. Heterogeneous structures reportedly raise unavoidable artifacts in the reconstruction of the AP data, and various reasons for these artifacts have been suggested (Oberdorfer and Schmitz, 2011; Vurpillot et al., 2000; Larson et al., 2010). First, exposing the structure to a laser pulse to assist field evaporation causes asymmetric evaporation of atoms with respect to the laser direction (Marquis
∗ Corresponding author at: Department of Materials Science and Engineering (MSE), Pohang University of Science and Technology (POSTECH), Pohang 790-784, Republic of Korea. Tel.: +82 10 5310 2139. E-mail addresses: [email protected] (J.H. Lee), [email protected] (C.G. Park). 0968-4328/$ – see front matter © 2013 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.micron.2013.11.003
et al., 2011a; Sha et al., 2008). When the thermal diffusion length is thinner than the tip diameter, the thermal distribution of the tip becomes anisotropic, which induces an alteration in the shape of the tip (Bunton et al., 2007). In oxides, which have lower thermal diffusivity than metals, severe preferential evaporation occurs on the side illuminated by the laser (Müller et al., 2011). Moreover, the laser exposure changes the nature of multiple events, so the compositional information varies with the position with respect to the laser (Müller et al., 2012). Second, the differences in the evaporation field among the layers also cause the surface curvatures of layers to change from the initial statuses (Marquis et al., 2011a). This continuous evolution of the tip shape results in local magnification and disordering of the field evaporation sequence (Oberdorfer and Schmitz, 2011). As a result, the atomic concentration across the interface cannot be quantitatively measured when the evaporation field differs greatly. Third, the dependence of the ion trajectory on the crystallographic orientation has been widely recognized since field ion microscopy was initially used (Gault et al., 2010a; Miller et al., 1989). The difference in the evaporation field depending on the crystallographic orientation leads to local changes in the radius of curvature of the specimen and to local magnification (Gault et al., 2010a). These phenomena can be observed by transmission electron microscopy (TEM). When a SiO2 layer on a Si substrate is evaporating, the lower Si substrate evaporates before the upper SiO2 layer owing to the low evaporation field of Si. Fig. 1(a)–(c) are TEM images taken after laser-assisted field evaporation. In Fig. 1(a), the red dotted areas are preferentially evaporated regions of the
J.H. Lee et al. / Micron 58 (2014) 32–37
33
Fig. 1. (a) TEM image of Si/SiO2 (100 nm) multilayered sample after AP experiment. TEM image was recorded during the (temporarily paused) AP experiment. Pulsed laser illumination came from the right side. The red dotted area represents the preferentially evaporated Si area, which formed a flat surface in the (1 1 1) plane (the close-packed plane). High-resolution images of (b) left and (c) right sides. The native oxide was grown on the Si surface because of the strong electron beam and vacuum breaking. The preferentially evaporated Si surface formed a (1 1 1) facet. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)
Si substrate. The laser illumination came from the right side of the tip in the TEM image, and this is evident from a comparison of the radii of curvature on the left and right sides. Because the right side was exposed directly to the pulsed laser, the radius of curvature there is larger than that on the left side. Additionally, the facet of the (1 1 1) surface exposed at the Si layer means that the field evaporation is strongly related to the crystallographic orientation. Fourth, surface migration of the atoms on the tip surface degrades the lateral resolution (Kellogg et al., 1978; Gault et al., 2010b). Early research demonstrated the wandering of surface atoms under nanosecond laser pulses (Gault et al., 2006). If surface migration is possible in laser-assisted APT (Gault et al., 2012; Cerezo et al., 2007), we can assume that chemical reactions requiring low activation energy could also be activated. Because these reactions would also disturb the compositional quantification, this aspect of the treatment should be considered. In this paper, we will demonstrate the interface reaction at the Si/SiO2 interface and show that it depends on the inner structure of the tip by using TEM observations.
2. Experimental Shallow trench isolation (STI) was chosen for the interface analysis [Fig. 2(a)]. The STI structure is commonly used to isolate neighboring cells. The STI region is filled with SiO2 by chemical vapor deposition after a dry etch process. The width of the Si column and the depth of the STI are 25 nm and 250 nm, respectively. Two types of sample from the same structure were prepared using a focused ion beam. The Si column is centered in sample A, and the SiO2 STI is centered in sample B, as shown in Fig. 2(a). JEM-2100F was used for TEM observation. The APT analysis was performed by LAWATAP (laser-assisted wide-angle tomographic atom probe) developed by CAMECA. A 343-nm, 100-kHz UV laser was set to 0.09 J/pulse at 30 K (spot size ∼30 m). The UV laser was illuminated perpendicular to the Si column [from the left or right direction in Fig. 2(a)]. The standing voltage was controlled automatically with respect to the detection flux in order to acquire AP data at the AP image. It provides three reconstruction algorithms: the voltage-based approximation, cone-shape approximation, and tip-shape based approximation. Among them, the tip-shape-based
Fig. 2. (a) TEM image of shallow trench isolation (STI). The width of the SiO2 layer is around 30 nm, and the depth of the STI is around 250 nm. (b) Side view of AP image from area A in (a). (c) Side view of AP image from area B in (a). Blue indicates Si; red indicates SiO2 . (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)
34
J.H. Lee et al. / Micron 58 (2014) 32–37
Fig. 3. (a) Top view of AP image from sample A. (b) Top view of AP image from sample B. Blue represents Si; red indicates SiO2 . (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)
approximation was adopted for the reconstruction of our samples. When a stiff voltage change occurs during field evaporation, the voltage-based approximation does not reflect the exact tip image. The cone-shape approximation is also not effective for a multilayered structure because it assumes a homogeneous structure. The most reasonable APT image was manually reconstructed by the tip-shape-based approximation method. The width of the Si column is calibrated at multiple points along the longitudinal direction based on the TEM image until the reconstructed image is identical to the TEM image. Note that the dimension of the outer layer (the SiO2 layer in sample A and the Si layer in sample B) could not be calibrated, because it is impossible to identify the edge of the tip when the outer edge of the tip has evaporated outside of the detector area. Calibrated AP images from samples A and B are shown in Fig. 2(b) and (c), respectively. Areas of Si and SiO2 are shown in blue and red, respectively. We admit that this method cannot perfectly reconstruct the real structure of the tip. However, artifacts due to the laser are unavoidable, so we tried to match the reconstructed images to the TEM images taken just before the AP measurement as closely as possible. 3. Results and discussion To examine the Si/SiO2 interface, views of samples A and B from above are shown in Fig. 3(a) and (b), respectively. Again, blue and red represents the Si column and SiO2 region, respectively. The outermost region is cut according to Muller’s suggestion (Müller et al., 2012); the image distortion of the outermost region increases
because the local magnification effect disrupts the reconstruction of the three-dimensional structure. Nevertheless, the image distortion due to the trajectory aberration remains because the evaporation angle changes according to the environment surrounding the evaporating atoms. Therefore, we chose the central area for interface interpretation. Fig. 4 shows the concentration profile across the Si/SiO2 interface in both samples. The rough Si:O ratio at the bulk-like SiO2 layer was 1:2. The concentration in the bulk area between the two samples did not differ significantly, whereas the interface concentration shows considerable disagreement between samples A and B and between the left and right interfaces. Note that the oxygen concentration in the Si area does not seem to be negligible. Because it is unlikely that O atoms would have existed in the pure Si column, we assumed that O atoms in the Si layer are attributable to an overlap in the evaporation trajectory between the two layers. That is, O atoms detected in the Si area were detached from the SiO2 layer because the evaporation trajectory of the SiO2 layer is wider than the adjacent Si area. According to the study of Bachhav et al. (2012), laser energy is absorbed only on the illuminated side of the tip apex, and the energy is transferred to the opposite side by thermal diffusion. A recent study of the absorption of UV and green laser light by Koelling et al. (2011) also showed that the absorption of UV laser light is highly asymmetric compared with that of green laser light. Moreover, the UV laser ( = 343 nm) was absorbed almost entirely at the top surface of the illuminated side, resulting in asymmetric heating of the tip. The asymmetric absorption of laser energy will cause the interfacial quantification to vary with the geometric placement of the interface and laser illumination. Regarding this assumption, we compared two interfaces located on the laserilluminated side and the opposite side. Fig. 5 demonstrates the silicon monoxide concentration profile across the Si/SiO2 interface in the two samples. SiO was chosen for the interface indicator owing to its dominance. We identified the peaks at 22 atomic mass units (amu) and 44 amu as SiO2+ and SiO+ , respectively, because the system consists only of silicon and oxygen. Note that around 3.3–3.5 at.% of SiO is measured in the SiO2 region. We focus on two important facts in Fig. 5. First, sample A, the Si-centered tip, has a high SiO concentration at the left and right interfaces, whereas sample B, the SiO2 -centered tip, has a relatively small peak at each interface. The small peak at the interface can be understood to come from the inherent suboxide at the interface. Owing to imperfect formation of the interface, a transition
Fig. 4. Lateral profiles of the atomic concentration across the interface between Si and SiO2 in (a) sample A and (b) sample B. The lateral dimension is not accurate because of the local magnification effect. At the SiO2 layer in both samples, the Si:O ratio is roughly 1:2. Note that oxygen atoms are found in the Si area because of their wide evaporation angle. Other minor constituents are omitted for clarity.
J.H. Lee et al. / Micron 58 (2014) 32–37
Fig. 5. Compositional profiles of silicon monoxide across the vertical interface of Si and SiO2 in sample A (silicon-centered sample, blue squares) and sample B (SiO2 centered, red circles). Pulsed laser illumination originates at the right side of the graph. A significant amount of SiO exists at the interface in sample A, whereas weak peaks were detected at the interface in sample B. Additionally, a larger SiO peak is detected on the right (the laser illumination side) in both samples. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)
layer consisting of substoichiometric Si oxide inherently exists at the interface between the Si and SiO2 layers, and its existence has been proven by previous electron energy loss spectroscopy analysis (Muller et al., 1999). However, it is unlikely that the interface in sample A contains twice the SiO of that in sample B. By iterating the same experiment several times, we were able to conclude that the discrepancy is not because of an intrinsic difference, but rather it is due to the dependence of laser-assisted field evaporation on the experimental geometry. When no supply of materials from outside is expected, it is worth considering possible reactions between internal constituents that could occur in the transition layer. That is, the interfacial reaction could be activated by laser-assisted field evaporation, and this feature depends on the inner structure of the tip as well as the laser illumination direction. In contrast, dissociation should also occur at the interface; there is no reason to assume that decomposition of SiO2 in the external field would not happen near the interface. Second, the amount of detected interfacial SiO depends on the direction of laser illumination, and the interface on the exposed side contains more SiO in both samples A and B. It could also be proven that the interfacial composition depends on the direction of the laser illumination because the difference in the peak height between the left and right interfaces follows the tip when it is rotated 180◦ (not shown here). Therefore, it is logical to assume that the higher concentration of SiO is due to synthesis at the interface, and the amount of synthesis is related to the laser–matter interaction. This assumption needs to satisfy the prerequisite that the activation energy for combination is smaller than that for SiO2 field evaporation. Otherwise, SiO2 will not combine with Si but instead will evaporate as usual, even if the laser impact is sufficient to combine the two components. In our specimen, the reaction between silicon and silicon dioxide, Si + SiO2 → 2SiO, is the most likely candidate, as it was reported in the 1990s by several groups (Tromp et al., 1985; Streit and Allen, 1987; Harp et al., 1990). Streit and Allen proved that Si atoms that diffused from the substrate are involved in the rapid decomposition of the thin SiO2 layer (Streit and Allen, 1987). A thick SiO2 layer on a Si substrate barely dissociates by itself at 900 ◦ C, but an ultrathin SiO2 layer less than 1 nm thick could be easily removed at a lower temperature. The activation energies (Ea ) for SiO2 desorption both with and without the assistance of Si were identified in an experiment in which the SiO2 layer
35
Fig. 6. TEM images of (a) sample A and (b) sample B before AP experiment. Tip surface became amorphous because of ion-beam milling. Images were taken as quickly as possible to prevent various problems due to electron beam with 200-kV acceleration.
was removed during exposure to a silicon ion beam. The preferred reaction is SiO synthesis rather than SiO2 desorption because the activation energies for desorption are Ea = 3.3–4.0 eV and 0.84 eV for thermal SiO2 removal and Si-assisted removal, respectively (Streit and Allen, 1987). Therefore, the prerequisite mentioned above is satisfied. We can conclude that SiO2 tends to combine and evaporate with a Si atom rather than dissociate to Si and O atoms, so it could decrease the energy required for evaporation. To explain why different amounts of SiO were detected at the interface in the two samples, we prepared the same set of samples and recorded TEM images of them before and after the AP experiment. Fig. 6(a) and (b) are TEM images taken from samples A and B, respectively, before the AP experiment. The electron beam intensity used for TEM image acquisition was minimized to avoid beam damage and hydrocarbon contamination before AP operation. High-resolution images were not taken for the same reason. The summit of sample A, single-crystalline Si, became amorphous despite the low energy of the ion-beam milling (2 kV). Samples A and B initially had similar radii of curvature: 17 nm and 22 nm, respectively. However, they evolved to take significantly different shapes, as shown in Fig. 7(a) and (b). After the initial period of field evaporation, a groove in the center of the tip apex developed in the Si-centered sample, A; in contrast, a sharp tip apex formed in the SiO2 -centered sample, B. The Si layer, which has a low evaporation field, was field-evaporated before the SiO2 layer, which has a high field. This produced a groove at the summit of sample A. Because the curvature of the tip edge is inversely proportional to the field required to evaporate the material, the preferential evaporation of Si atoms increases the radius of curvature of the Si layer until the field applied to each material is equivalent. This will subsequently alter the trajectory of each element. The groove constrains the evaporation angle of silicon atoms, narrowing the trajectory of the evaporated Si atoms. In the same manner, the evaporation angle of the SiO2 layer is widened because of the groove. Preferential evaporation of Si also occurs in sample B. However, the opposite arrangement of the layers induced the distinctive evolution of the tip apex. The pulsed laser illumination was at the right side of both samples in the TEM images. The direction of the laser illumination could easily be confirmed by observing the tip shape. The red-dotted envelops in Fig. 8(a) and (b) represents
36
J.H. Lee et al. / Micron 58 (2014) 32–37
Fig. 7. TEM images of (a) sample A and (b) sample B after AP experiment. AP operation was stopped during evaporation to observe the shape evolution with respect to the inner structure. When the Si is centered, the preferentially evaporated Si layer forms a groove at the center of the tip. When the SiO2 is centered, the sample evolves into a sharp tip.
the radii of curvature on the left and right sides of the tip apex. In sample A, the radii of curvature were 25 and 35 nm at the left and right corners, respectively. In sample B, the radii of curvature were 35 and 45 nm at the left and right corners, respectively. The laser side developed a larger radius of curvature in both samples owing to the non-uniform field distribution (Marquis et al., 2011b). We believe the difference in how the shapes evolve explains the difference in interfacial composition between the two samples. The silicon layer confined by the SiO2 layer cannot evaporate easily as a pure silicon layer but develops a flat surface, as shown in Figs. 7(a) and 8(a). For this reason, considering the Si-assisted SiO2 evaporation of sample A, many Si atoms can combine with SiO2 molecules at the atomic distance of the SiO2 surface because preferential evaporation of Si atoms was retarded, as we described above. In contrast, SiO2 molecules in sample B had to evaporate
unassisted because adjacent Si atoms had already evaporated, as shown in Fig. 8(b). Because of these contrasting circumstances, each interface shows a different interfacial composition. In addition, the thermal behavior of the two materials could enhance the compositional difference. The thermal conductivity of Si is known to be two orders greater than that of SiO2 (Kim et al., 1999). Because of the geometry of the tip and laser, the top surface exposed to the UV laser became SiO2 in sample A and Si in sample B. Note that the diameter of the pulsed laser is large enough to cover the entire area of the tip. In sample B, the heat generated by the pulsed laser rapidly dissipated through the shank or evaporated the surface material. However, the absorbed heat was used to combine Si and SiO2 in sample A. The difference between the left and right interfaces could be explained by the absorption of the UV laser. As we mentioned above, the absorption depends strongly on the illumination direction and occurs only at the top surface (Bachhav et al., 2012; Koelling et al., 2011). This non-uniformity can explain why SiO synthesis occurred mainly at the interface of the laserilluminated side. In order to reinforce this claim, we conducted an additional experiment on the thermal effect of the pulsed UV laser on the tip surface. Fig. 9(a) and (b) are TEM images taken before and after laser illumination on the same STI structure. The power of the laser was the same as in the previous experiment. The sample was milled by a 30-kV Ga ion beam to intentionally damage the tip surface. Indeed, the surface of the tip had been about 25 nm deep so that the single crystalline surface became amorphous, as shown in Fig. 9(a). After laser exposure for 1 h (standing voltage was sustained at 5 kV), epitaxial recrystallization had occurred at the amorphous surface. It should be noted that the epitaxial recrystallization occurred at over 480 ◦ C by thermal annealing (not shown here). Although it is hard to explain the contribution of the standing voltage to the thermal reaction, it is reasonable to claim that the illumination of the pulsed laser possibly caused the thermal reaction. In this work, we observed the STI structure using APT and TEM. The interfacial Si and SiO2 composition was not consistent, but depended on the laser, inner structure of the tip, reaction kinetics, and geometric relationship between the laser and the sample. Because current reconstruction algorithms do not fully reflect the effect of the pulsed laser yet, these conditions should be considered when evaluating the interfacial composition of multilayered structures.
Fig. 8. High-resolution TEM images of (a) sample A and (b) sample B after AP experiment. Tip surface was slightly oxidized in the TEM chamber because of electron beam exposure. Pulsed laser illuminated the right sides of both tips. Red-dotted lines represent the radii of curvature of tip edges. In both samples, the right edges developed a larger radius of curvature due to asymmetric absorption of laser. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)
J.H. Lee et al. / Micron 58 (2014) 32–37
37
Fig. 9. TEM images of the STI structure (a) before pulsed laser exposure and (b) after laser exposure. Since the sample was intentionally damaged by the Ga ion beam, the surface of the Si layer became amorphous. Owing to laser exposure, the laser illuminated area had recrystallized with an epitaxial relationship. The epitaxial recrystallization was observed after laser illumination for 1 h. It is worth mentioning that the recrystallized length from the apex is approximately 350 nm, which is similar to the wavelength of the UV laser (343 nm).
4. Conclusions In summary, an STI structure consisting of Si and SiO2 was studied by APT and TEM. The interface composition was found to change depending on the arrangement of the AP tip. At the interface, SiO2 tends to combine and evaporate with Si when Si atoms are atomically close to the evaporation site. This Si-assisted SiO2 evaporation is attributed to a greater reduction in the activation energy for SiO combination than in that for dissociative evaporation, leading to considerable modification of the interfacial composition during AP analysis. TEM observations proved that the amount of interfacial SiO depends on the shape evolution of the tip. The use of a pulsed laser is essential for AP analysis of insulating materials. However, its role should not be dominant but rather assistive because its influence is not fully understood. Therefore, in compositional analysis of heterogeneous structures using an AP, the laser–matter interaction must be prudently considered to reduce erroneous analyses. Acknowledgements This research was supported by Samsung Electronics Co. The authors thank NINT for supplying the analysis equipment. References Bachhav, M.N., Danoix, R., Vurpillot, F., Hannoyer, B., Ogale, S.B., Danoix, F., 2012. J. Appl. Phys. 111, 064908. Bunton, J.H., Olson, J.D., Lenz, D.R., Kelly, T.F., 2007. Microsc. Microanal. 13, 418. Cavin III, R.K., Herr, D.J.C., Zhirnov, V.V., 2000. J. Nanopart. Res. 1, 213.
Cerezo, A., Clifton, P.H., Gomberg, A., Smith, G.D.W., 2007. Ultramicroscopy 107, 720. Gault, B., Menand, A., de Geuser, F., Deconihout, B., Danoix, R., 2006. Appl. Phys. Lett. 88, 114101. Gault, B., La Fontaine, A., Moody, M.P., Ringer, S.P., Marquis, E.A., 2010a. Ultramicroscopy 110 (9), 1215. Gault, B., Müller, M., La Fontaine, A., Moody, M.P., Shariq, A., Cerezo, A., Ringer, S.P., Smith, G.D.W., 2010b. J. Appl. Phys. 108, 044094. Gault, B., Danoix, F., Hoummada, K., Mangelinck, D., Leitner, H., 2012. Ultramicroscopy 113, 182. Gilbert, M., Vandervorst, W., Koelling, S., Kambham, A.K., 2011. Ultramicroscopy 111, 530. Harp, G.R., Han, Z.L., Tonner, B.P., 1990. Phys. Scripta T31, 23. Kellogg, G.L., Tsong, T.T., Cowan, P., 1978. Surf. Sci. 70, 485. Kelly, T.F., Larson, D.J., Thompson, K., Alvis, R.L., Bunton, J.H., Olson, J.D., Gorman, B.P., 2007. Annu. Rev. Mater. Res. 37, 681. Kim, J.H., Feldman, A., Novotny, D., 1999. J. Appl. Phys. 86, 3959. Koelling, S., Innocenti, N., Schulze, A., Gilbert, M., Kambham, A.K., Vandervorst, W., 2011. J. Appl. Phys. 109, 104909. Larson, D.J., Geiser, B.P., Prosa, T.J., Gerstl, S.S.A., Reinhard, D.A., Kelly, T.F., 2010. J. Microsc. 243, 15. Marquis, E.A., Geiser, B.P., Prosa, T.J., Larson, D.J., 2011a. J. Microsc. 241, 225. Marquis, E.A., Geiser, B.P., Prosa, T.J., Larson, D.J., 2011b. J. Microsc. 241, 225–233. Miller, M.K., Hetherington, M.G., Burke, M.G., 1989. Metall. Trans. A 20 (12), 2651. Muller, D.A., Sorsch, T., Moccio, S., Baumann, F.H., Evans-Lutterodt, K., Timp, G., 1999. Nature 399 (24), 758. Müller, M., Saxey, D.W., Smith, G.D.W., Gault, B., 2011. Ultramicroscopy 111, 487. Müller, M., Smith, G.D.W., Gault, B., Grovenor, C.R.M., 2012. J. Appl. Phys. 111, 064908. Oberdorfer, C., Schmitz, G., 2011. Microsc. Microanal. 17, 15. Sha, G., Cerezo, A., Smith, G.D.W., 2008. Appl. Phys. Lett. 92, 043503. Streit, D.C., Allen, F.G., 1987. J. Appl. Phys. 61, 2894. Thompson, S.E., Parthasarathy, S., 2006. Mater. Today 9 (6), 20. Tromp, R., Rubloff, G.W., Balk, P., LeGoues, F.K., van Loenen, E.J., 1985. Phys. Rev. Lett. 55 (21), 2332. Vurpillot, F., Bostel, A., Blavetter, D., 2000. Appl. Phys. Lett. 76, 3127.