Strain Stability in Nanoscale Patterned Strained Silicon-On-Insulator O

ECS Transactions, 33 (6) 511-522 (2010) 10.1149/1.3487581 © The Electrochemical Society

Strain Stability in Nanoscale Patterned Strained Silicon-On-Insulator O. Moutanabbira,b, M. Reichea, A. Hähnela, W. Erfurtha, A. Tarun b, N. Hayazawa b, and S. Kawata b, F. Naumannc, M. Petzoldc a

Max Planck Institute of Microstructure Physics, Weinberg 2, D 06120 Halle (Saale), Germany b Nanophotonics Laboratory, RIKEN Advanced Science Institute, Hirosawa, Wako, Saitama 351-0198, Japan c Fraunhofer for Mechanics of Materials, Walter-Hülse-Strasse 1, D 06120 Halle (Saale), Germany Nowadays, strain engineering plays a key role in boosting the performance of Si-based nanoelectronics. So far a tremendous progress has been made in establishing methods for strain manipulation in Si nanodevices. This has brought up further challenges in terms of development of reliable probes to characterize the strain on the nanoscale. In this paper, we discuss strain imaging using multi-wavelength micro-Raman spectroscopy. As a model system, we investigate the strain behavior upon nanoscale patterning of ultrathin strained silicon-on-insulator (SSOI). We show that valuable details on the strain depth distribution can be obtained by combining deep UV and visible Raman microprobes. Additionally, we also demonstrate that strain mapping with high lateral resolution can be achieved using UVRaman with glycerin-immersed high numerical aperture objective lens. Results from nano-beam electron diffraction and peak-pairs analysis of high-angle annular dark field images are also presented. Detailed 3D finite element simulations of edge-induced strain relaxation in SSOI nanostructures augment our experimental studies.

Introduction

The precise control of the amount and distribution of strain in Si devices has emerged as a powerful strategy to insure the continuity in the scaling of CMOS technology (1). Indeed, since the 65 nm technology node strain has been used to improve the carrier transport in Si-based CMOS devices. It was demonstrated that the introduction of a compressive and/or tensile strain in the Si channel can improve the mobility of holes and electrons and, consequently, increase the n- and p-MOSFET drive currents (2, 3). Compressive strain is usually introduced by a shallow trench isolation toward channels longitudinally and laterally (4). To induce tensile strain in the channel, SiGe virtual substrates (5), tensile films (6), and mechanical forces (7, 8) are among the widely used methods. In this case, the strain is generated locally during transistor processing. This approach, known as local strain process, utilizes SiGe deposition on specific active regions beside the channel. The strains induced in this case are typically uniaxial. However, with the technology scaling, the efficiency of this local strain tends to drop (9).

511 Downloaded 01 Oct 2010 to 192.108.69.253. Redistribution subject to ECS license or copyright; see http://www.ecsdl.org/terms_use.jsp

ECS Transactions, 33 (6) 511-522 (2010)

An alternative method is to build transistors directly from a strained Si material. For this approach, known as global strain process, materials engineering is as critical as the device design. The global strain can be generated on the wafer level by the growth of ultrathin Si layer on a relaxed SiGe virtual substrate. In this case, the induced tensile strain is biaxial and its magnitude depends on the content of Ge in the virtual substrate. The biaxial tensile strain breaks both the sixfold degeneracy in the ellipsoidal valley of the conduction band as well as the degeneracy between heavy and light hole bands (10). This translates into a reduction of the intervalley scatterings in the conduction band and the electron effective transport mass, which leads to an enhancement in the electron mobility (11). However, the presence of the SiGe layer was found to cause several complications in the final device such as high leakage current, Ge diffusion, and enhanced n-type dopant diffusion. The obvious remedy to these problems would be the elimination of the SiGe layer. This can be accomplished by using ultrathin strained Si layer directly on a SiO2/Si substrate (i.e., the SiO2 layer is sandwiched between strained Si layer and Si substrate) (12, 13). This novel SiGe-free structure combines the advantages of Si-on-insulator technology (14) with the benefits of strained Si. Thus, there is a strong motivation to fabricate and understand the fundamental properties of ultrathin strained Si layer directly on insulator. The potential introduction of SSOI in device fabrication raises fundamental questions about the evolution and stability of the strain during different processing steps. The amount and distribution of the residual strain are very critical to the design and performance of SSOI-based devices. Therefore, it is of compelling importance to probe the local strain in individual SSOI structures. Toward this end, high spatial resolution and less invasive techniques are needed. In this paper, we demonstrate strain imaging using multi-wavelength micro-Raman spectroscopy. We show that valuable details on the strain depth distribution can be obtained by combining deep UV and visible Raman microprobes. Additionally, we also demonstrate that strain mapping with high lateral resolution can be achieved using UV-Raman with glycerin-immersed high numerical aperture objective lens. Results from nano-beam electron diffraction (NBED) and peakpairs analysis of high-angle annular dark field (HAADF) images are also presented. It is important to mention that the techniques developed or optimized in this work can also be applied to different systems other than SSOI-based nanostructures.

Characterization of Strain in SSOI: Micro-Raman vs. Scanning Transmission Electron Microscopy 8-inch SSOI wafers were used in this study. The strain was generated by the heteroepitaxial growth of Si thin film on ~500 nm-thick Si1-xGex buffer layer, which was relaxed by helium implantation and thermal annealing (13). The tensile strained thin layer was then transferred to a SiO2/Si handle wafer by combining wafer bonding and ion-cut process. The amount of the strain is controlled by adjusting the content of Ge in the SiGe buffer layer. In fact, the expected in-plane biaxial strain in these SSOI substrates depends on xGe as follows:

ε xx = ε yy =

§ a − aSi · aSiGe − aSi ¸¸ = xGe × ¨¨ Ge aSi © aSi ¹

[1]



where aSi, aGe, and aSiGe are lattice parameters of bulk Si, Ge, and SiGe alloy respectively. The initial thickness of the strained layer is in the range of 15-30 nm. Thicker layers were obtained by an additional homoepitaxy. Figure 1 displays cross-sectional transmission electron microscopy (XTEM) images of thin and thick SSOI substrates. Note the strained films have a uniform thickness and are defect-free as demonstrated by the high resolution images and the electron diffraction patterns.

Figure 1: XTEM images of 20 nm-thick (left) and 70 nm-thick (right) SSOI substrates. 510.7 515.9

xGe = 0.16

Intensity [a.u]

xGe = 0.30

480

490

500

510

520

530

540

-1

Si-Si Raman mode [cm ]

Figure 2: Typical UV-Raman spectra of SSOI substrates fabricated using relaxed SiGe buffer layers having a Ge content of 0.16 (circles) and 0.3 (squares). The solid lines correspond to Voigt function fit. Note that the Si-Si Raman mode peaks at ~520.8 cm-1 in bulk Si.



First, we employed UV micro-Raman to characterize the strain in SSOI substrates. The measurements were performed in backscattering geometry by a LabRam HR800 UV spectrometer with a 325 nm He-Cd laser line corresponding to a penetration depth of ~10 nm in Si. A 40× objective was used to focus the laser beam to a spot of ~800 nm in diameter at the surface of the sample. To avoid local heating effects, the laser power was kept below 2 mW. The backscattered Raman light is diffracted by a 2400 g/mm grating and detected by a charged coupled device camera. The He-Cd plasma line at 854.7 cm-1 was used for the calibration of Raman spectra. The strain values are calculated from the measured wavenumbers of the Si-Si LO phonon using (15):

ω Sistrained = ωSibulk−Si + − Si

· s12 ¨¨ q + p ¸¸ × (ε xx + ε yy ) 2 © s11 + s12 ¹

ω Sibulk−Si §

[2]

bulk strained where ω Si− and ω Si− are the Si-Si Raman shift frequency (in cm-1) of bulk and strained Si Si Si, respectively. İxx and İyy represent the strain in the two in-plane directions. s11 and s12 are the anisotropic elastic compliance tensor elements. p and q are the phonon deformation potentials. Figure 2 shows typical UV Raman spectra of two SSOI substrates fabricated using SiGe relaxed buffer layers with a Ge content of 0.16 (circles) or 0.30 (squares). We note that for xGe = 0.30 the Si-Si Raman mode peaks at ~510.7 cm-1 corresponding to an average in-plane biaxial strain İxx = İyy = 1.24% as estimated from Eq. 2, which agrees with 1.2% expected strain value from the content of Ge (Eq. [1]). Similar agreement is obtained for xGe = 0.16 corresponding to İxx = İyy = ~0.6%. Thus one can conclude that Raman microprobe (based on Eq. [2]) is a reliable method for the evaluation of strain in SSOI materials.

Besides UV Raman spectroscopy, To quantify the strain in the strained layer, nanobeam electron diffraction (NBED) and peak-pairs analysis of high-angle annular dark field (HAADF) images were additionally performed using a probe Cs-corrected FEI-Titan 80-300 microscope, and were correlated with results from Raman spectroscopy [16]. The NBED principle consists in illuminating a nanometer-sized area of the specimen with a nearly parallel electron beam and acquiring series of diffraction patterns at points along previously defined lines. Then the strain components of interest are calculated based on subpixel analyses of the electron diffraction patterns, which include precise information on distances of lattice planes. The spatial resolution of NBED is determined by the beam diameter to a few nanometers, and both, the strain-sensitivity and the accuracy are in the order of 0.1 %. HAADF allows nano-scale strain mapping by use of the known image processing methods for strain analysis from high-resolution electron microscopy (HREM) images. These methods are mainly based on two different approaches: geometric phase analysis and peak-finding, e.g., the peak-pairs analysis (PPA) [16-23]. Processing the images numerically allows estimating the local shifts of intensity maxima (corresponding to the atomic columns) with respect to the positions in a relaxed lattice with a sensitivity of 1×10-3 (in relation to silicon). Figure 3 displays a typical strain map obtained by applying the peak-pairs analysis (PPA) for a SSOI substrate fabricated from a SiGe relaxed buffer layer having a Ge content of 0.16. The reference region of unstrained Si is shown on the left. The in-plane strain İ 110 values were calculated by integrating the corresponding profile in the strain map over a length of about 20 nm along the direction. The estimated average strain was found to be ~0.63 %, which is in perfect agreement with Raman data. Similar strain value was also obtained using NBED



as shown in Figure 4. For NBED analysis, two FIB specimens in the orientation were prepared perpendicular to each other and corresponding to the two in-plane directions. The comparison Raman vs. NBED or PPA/HAADF provides the confidence in using STEM-based methods to evaluate the strain in SSOI ultrathin films. The anticipated strain relaxation during specimen thinning need for STEM analysis appears to be negligible in the present conditions.

Figure 3: Typical İ 110 map (right), calculated by applying PPA to the HAADF image of SSOI layer. The region of interest (ROI) as well as the area used as a reference are marked by white rectangles (The color scale indicates the strain in %. See Ref. 16 for details)

Figure 4: Strain profile in SSOI layer as obtained using NBED. (a) The calibration profile of İ 110 calculated from the diffraction series measured at points along the 300nmline in the μPSTEM image; (b) High-resolution İ 110 profile across the strained Si layer. See Ref. 16 for details



UV-Raman strain imaging in SSOI nanostructure In this section, we focus on the application of UV micro-Raman spectroscopy to map the strain in individual SSOI nanostructures. As mentioned above, the fabrication of SSOI-based devices requires nanoscale patterning of SSOI ultrathin layers. It was demonstrated that thermal annealing at temperatures as high as ~1000 ºC does not influence the strain (24, 25). The formation of free surfaces upon nanoscale patterning was found, however, to lead to a relaxation of the strain. The extent of this phenomenon depends on the size, thickness and geometry of the patterned structures (26-30). Despite these extensive studies, experimental investigations of the strain behavior in single patterned SSOI structures are still missing. In fact, all published reports are based on the analysis of the post-patterning strain averaged over several structures (26-30). In the following, we demonstrate that the strain in a single SSOI patterned structure can be mapped with a fairly good spatial resolution by using confocal UV-Raman scattering spectroscopy. Enhanced spatial resolution was obtained by making a few adjustments to a conventional Raman spectrometer (31). Here a beam from a 355 nm CW laser is introduced into a modified inverted optical microscope (NIKON, TE2000) and focused by a high numerical aperture (NA = 1.4) ×100 UV objective lens. In order to improve the laser focusing and to minimize index mismatch-induced spherical aberrations, the objective and samples were directly immersed in Glycerin with a refractive index of 1.47. The use of Glycerin has the advantage of suppressing the fluorescence background observed in the UV range for the widely used oil immersion, which is typically designed for visible wavelength excitation. According to Rayleigh Criterion the theoretical spatial resolution of this setup is ~155 nm. The microscope is equipped with a PZT driven stage (Physik Instrumente P-517.3CL multi-axis piezo scanner) having an accuracy of 1 nm with a closed-loop feedback system. Raman 2D maps were reconstructed from spectra measured with a 10 or 30 nm-step. For this investigation, a 15 nm-thick SSOI layer under a biaxial tensile strain of 0.8% was used. Order arrays of square-like structures were fabricated using electron beam lithography and reactive ion etching. Figure 5(a) shows a scanning electron microscope image of an array of 500 nm × 500 nm SSOI structures.

(b)

0.75

-1

S i-S i[c m ]

800

515.5

Si-Si

ε

patterned SSOI

514.9

0.70 0.65

515.4 / 0.664 515.4

0.60

514.3

0.55 2.0

Y [nm]

cm-1 / İ [%]

0.80 516.1

1000

ε [% ]

(a)

2.5

3.0 3.5 4.0 position [a.u]

4.5

515.7 / 0.627 515.7 516.1 / 0.578 516.1

5.0

600

516.4 / 0.541 516.4 516.8 / 0.492 516.8

400 517.1 / 0.455 517.1 517.5 / 0.455 517.5

200

517.8 / 0.369 517.8

518.0

0 0

200

400

600

800

1000

X [nm]

Figure 5: (a) A scanning electron microscope image of an array of 500 nm × 500 nm SSOI structures; (b) 2D map of the Si-Si position and the corresponding post-pattering biaxial tensile strain measured by UV-Raman. Note that the dark region between the structures corresponds to no signal as the background from the underlying Si substrate was subtracted. The inset displays the profiles of the Si-Si peak position and the strain across the 500×500 nm2 structure.



Figure 5(b) shows a 1 ȝm × 1 ȝm 2D map of the intrinsic Si-Si Raman peak position and the corresponding biaxial tensile strain measured for the sample shown in Figure 5(a). The map was reconstructed from ~1110 spectra recorded with a ~30 nm-step. We note that the lowest Si-Si peak position is measured at the center of the patterned structure corresponding to ~515.4 cm-1, which is ~1.1 cm-1 above the value in the initial film. This upshift of the Si-Si Raman mode corresponds to a decrease in the biaxial tensile strain from 0.80% to 0.67%. This demonstrates that patterning induces a partial relaxation of the strain in agreement with recent observations (26-30). Note that due to their symmetric shape, the residual strain remains isotropic in both lateral directions (biisotropic) in the investigated structures. The interesting observation here is that the strain in the patterned structure is no longer uniform as the relaxation becomes more pronounced away from the center leading to a tensile strain of ~0.37% near the edges. This is clearly visible in Figure 5(b) (inset) showing the evolution of the Si-Si peak position and the corresponding in-plane strain across the patterned structure in comparison with the values in the initial SSOI substrate. This edge-induced strain relaxation is a result of the inward displacement of the lattice at and near the newly formed free surfaces leading to a strain gradient along the patterned structure. Multi-wavelength Raman microprobe: An in-depth strain probe In the previous section, strain mapping using UV-Raman was demonstrated for a single SSOI nanostructure. In spite of this progress, probing SSOI-based nanostructures using micro-Raman faces, however, some complications due to the fact that the oxide is transparent to the laser light leading to a strong background from the underlying Si substrate in Raman spectra. This severely limits the sensitivity of the technique and projects large uncertainties on the obtained strain values as the signal from the strained structures becomes overwhelmed by the background signal. Fitting and subtracting the background can be accurate to a certain extent provided that the strain is sufficiently high to allow a clear distinction between the Si-Si peaks of the strained structures and the underlying substrate. Obviously, this is not the case for tiny SSOI nanostructures where both the relaxation and the fraction of the laser beam reaching the Si substrate are expected to be more significant. To overcome this limitation, we developed a novel heterostructure that permits the precise analysis of the strain in nanoscale patterned ultrathin SSOI using multi-wavelength micro-Raman spectroscopy. This heterostructure consists of a 15 nm-thick strained Si layer transferred onto SiO2/Ge/Si multilayer as shown in Figure 6(a) (32). The strained layer was fabricated by the epitaxial growth of Si on a ~500 nm-thick Si0.84Ge0.16 buffer layer grown on Si(001) by chemical vapor deposition. The obtained strained-Si/Si0.84Ge0.16/Si wafer was subsequently capped by a ~200 nm-thick SiO2 layer grown by plasma enhanced chemical vapor deposition. A second wafer was prepared by deposition of ~120 nm-thick Ge layer on Si(001) using molecular beam epitaxy. This wafer was also covered by a ~200 nm-thick SiO2 layer. The ultrathin strained Si layer was transferred from the first wafer onto the second one using the ion-cut process and selective chemical etching of SiGe. Owing to the high absorption coefficient of Ge, the introduction of the Ge layer prevents the laser from reaching the Si handle substrate thereby suppressing the background in Raman spectra of SSOI nanostructures. This is clearly demonstrated in Fig. 1b showing Raman signal of the obtained heterostructure recorded using a 488 nm Ar+ laser line. The spectrum is composed of two peaks associated with the Si-Si and Ge-Ge modes from the strained Si



and Ge layers, respectively. No Si-Si mode from the underlying Si substrate is detected because the 488 nm laser used has a limited penetration depth of ~20 nm in Ge. Note that the observed Si-Si peak is centered at ~515.9 cm-1 corresponding to a biaxial tensile strain of ~0.6 % (stress of ~1.1 GPa) as expected from the Ge content in the Si0.84Ge0.16 buffer layer. This indicates that the presence of Ge layer does not affect the strain in the Si layer. To investigate the influence of nanoscale patterning on the strain, ordered arrays of SSOI nanostructures were fabricated using electron beam lithography and reactive ion etching. Three sets of patterned square-shaped nanostructures were prepared with a lateral dimension (L) of 50, 100, or 500 nm. The post-patterning strain was investigated using micro-Raman system equipped with a deep UV frequency-doubled 244 nm laser, a UV He-Cd 325 nm laser, and an Ar+ 488 nm laser having a penetration depth in Si of ~6, ~10 and ~570 nm, respectively. The penetration depth of each laser is superposed on the high resolution TEM image (Fig. 6(a)).

(a)

(b)

325

488

Raman Intensity [a.u.]

244

Si-Si

bulk Si

strained Si/oxide/Ge/Si

Ge-Ge

280 300 320 500 510 520 -1 Raman Shift [cm ]

530

Figure 6: (a) Cross-sectional TEM image of the strained-Si/SiO2/Ge/Si heterostructure used in this study. The strained Si layer was transferred onto Ge/Si substrate using the ion-cut process. The arrow indicates the bonding interface. Inset: High resolution TEM image and electron diffraction pattern of the strained Si layer. The lines in the high resolution TEM denote the penetration depth of each laser lines used in micro-Raman analysis. (b) Raman spectrum of the heterostructure shown in (a). The signal was recorded using a 488 nm laser. The vertical dashed line denotes the Si-Si peak position in bulk Si. For all investigated SSOI nanostructures we found that the spectra recorded using the 244 nm and 325 nm laser lines to be identical for a fixed lateral dimension L. This implies that, regardless of the size, the strain in the top 6 nm and in the 10 nm is the same. However, we observed a strong difference between these spectra and the spectra recorded using the visible laser. This conflicts between UV and visible data is very instructive. Table I summarizes the strain and the corresponding stress measured for SSOI nanostructures at different depths. It is worth pointing out that the obtained results show that relying on UV-Raman alone to probe the strain in SSOI nanostructures can lead to inaccurate conclusions regarding the strain in nanostructures thicker than the penetration depth of the UV laser in Si (~10 nm). In fact, the extrapolation of the UV-Raman data to the whole thickness exaggerates the phenomenon of the patterning-induced relaxation of the strain in SSOI nanostructures. Moreover, it emerges from data in Table I that the



region close to the interface Si/SiO2 becomes under a high tensile strain upon nanoscale patterning. The amount of the post-pattering strain in this region appears to be less sensitive to the lateral dimension. In contrast, the top 10 nm is very sensitive to the size and relaxes rather quickly by reducing L. Indeed, the relaxation was found to involve a pronounced contraction of the region near the surface. This compression or inward motion of the lattice near the surface occurs parallel to an expansion of the lattice near the interface Si/SiO2. The highest difference in the strain between these two regions was observed at a lateral dimension of 100 nm. Interestingly, at L = 50 nm, both compressive and tensile strained regions coexist within the same nanostructure. This phenomenon can provide an additional degree of freedom in the design and fabrication of strained Si nanodevices.

İxx = İyy (ıxx = ıyy)

50 nm

100 nm

500 nm

Top 6 to 10 nm

~-0.10 (~-180)

~0.00 (0)

~0.47 (~830)

Whole 15 nm-thick nanostructure

~0.11 (~195)

~0.27 (~480)

~0.60 (~1065)

Bottom 5 nm

~0.53 (~940)

~0.81 (~1440)

~0.86 (~1530)

Table I: Experimental values of in-plane strain (stress) components in % (MPa) estimated from deep UV and visible micro-Raman probes. L = 50 nm

-0.063

0.098

0.260

0.421

0.583

0.745

0.906

1.067

1.229

L = 100 nm

1.391

-0.066

0.127

0.319

0.512

0.705

0.898

1.091

1.283

1.476

1.669

L = 500 nm

Figure R6: 3D finite element simulations of the distribution of the first principle strain (in %) within the 15nm-thick nanostructures. -0.030

0.073

0.176

0.279

0.381

0.484

0.587

0.690

0.587

0.896



The ensemble of these observations provides the experimental demonstration of calculation-based mechanisms of the edge-induced strain relaxation in nanopatterned SSOI (see e.g., Ref. 28). To gain more insights into this phenomenon, we performed detailed 3D finite element simulations using the program ANSYS V 12.0. The simulations were done in two steps using a reduced quarter model consisting of 3D 20node elements with quadratic displacement behavior. The material properties of the strained Si and the underlying SiO2 were assumed to be anisotropic and isotropic, respectively. Both materials are linearly elastic in the calculations. In a first step, an initial strain of 0.6 %, corresponding to the initial experimental value, was applied on the patterned nanostructures via a virtual biaxial thermal expansion of Si. In order to avoid relaxation in the first step, the edges of the structures were fixed with rigid boundary conditions. SiO2 was initially considered under stress to meet the characteristic of PEVCD-grown oxide used in this work. The relaxation phenomenon was then simulated by removing the boundary conditions in the second step. The obtained 3D maps of the first principle strain for L = 50, 100, and 500 nm are shown in Figure 7. It is clear that the formation of edges leads to a pronounced relaxation of the lattice near the surface. This relaxation is more important for the smallest structures. Independently of the size, the patterning-induced relaxation leads to a out-of-plane distortion of the lattice. It is also important to mention that the strain remains biisotropic (İxx = İyy) in the patterned structures. Interestingly, the Si/SiO2 becomes relatively highly strained upon nanopattering in qualitative agreement with the experimental data shown in Table 1. Conclusion

We have discussed some practical issues in the application of Raman micro-probes, NBED, and PPA to characterize the strain in ultrathin SSOI. We have also described a method to achieve strain mapping with a fairly good spatial resolution using UV microRaman with glycerin-immersed high numerical aperture objective lens. Additionally, we have presented a novel heterostructure to elucidate accurately the strain evolution upon nanoscale patterning of ultrathin SSOI substrates. The introduction of a Ge layer between the handle substrate and the buried oxide suppresses effectively the background and enhances the sensitivity of Raman scattering. By combining UV and visible Raman microprobes unprecedented insights into the phenomenon of edge-induced relaxation in SSOI nanostructures were obtained. We found that the strain in the top 10 nm relaxes quickly and appears to be very sensitive to the size, whereas the region near the Si/SiO2 interface becomes highly strained upon nanopatterning. This disparity in the strain evolution between surface and interface is in a qualitative agreement with numerical calculations based on continuum mechanical approach of nanopatterning-induced relaxation. Acknowledgments We are grateful to U. Doss and H. Blumtritt for their technical help. This work was supported by the German Federal Ministry of Education and Research in the framework of the DECISIF project (contract no. 13 N 9881).



References 1. C. K. Maiti, S. Chattopadhyay, and L. K. Bera, Strained-Si heterostructures field effect devices (Taylor & Francis, 2007). 2. J. Welser, J. L. Hoyt, S. Takagi, and J. F. Gibbons, IEDM Tech. Dig. 373 (1994); K. Rim, J. Welser, J. L. Hoyt, and J. F. Gibbons, IEDM Tech. Dig. 517 (1995). 3. K. Rim. S. Koester, M. Hargrove, J. Chu, P. M. Mooney, J. Ott, T. Kanarsky, P. Ronsheim, M. Leong, A. Grill, and H. –S. Wing, VLSI Technology 59 (2001). 4. V. Chan, K. Rim, M. Ieong, S. Yang, R. Malki, R. Malik, R. W. The, M. Yang, and Q. Ouyang, Proceedings of the IEEE Custom Integrated Circuits Conference 667 (2005). 5. C. K. Maiti, L. K. Bera, and S. Chattopadhyay, Semicond. Sci. Technol. 13, 1225 (1998). 6. A. Shimizu, K. Hachimine, N. Ohki, M. Ohta, Y. Nonka, H. Sato, and F. Ootsuka, IEEE IEDM Tech. Dig. 433 (2001). 7. S. Maikap, C.-Y. Yu, S.-R. Jan, M. H. Lee, and C. W. Liu, IEEE Electron Dev. Lett. 25, 40 (2004). 8. A. Lochtefeld, and D. A. Antoniadis, IEEE Electron Dev. Lett. 22, 591 (2001). 9. S. Eneman, P. Verheyen, R. Rooyackers, F. Nouri, L. Washington, R. Degraeve, B. Kaczer, V. Moroz, A. De Keersgieter, R. Schreutelkamp, M. Kawaguchi, Y. Kim, A. Samoilov, L. Smith, P. P. Absil, K. De Meyer, M. Jurczak, and S. Biesemans, VLSI Technology 22 (2005). 10. G. Abstreiter, H. Brugger, T. Wolf, H. Jorke, and H.-J. Herzog, Phys. Rev. Lett. 54, 2441 (1985); M. L. Lee, E. G. Fitzgerald, M. T. Bulsara, M. T. Currie, and A. Lochtefeld, J. Appl. Phys. 97, 011101 (2005); C. Euaruksakul, Z. W. Li, F. Zheng, F. J. Himpsel, C. S. Ritz, B. Tanto, D. E. Savage, X. S. Liu, and M. G. Lagally, Phys. Rev. Lett. 101, 147403 (2008). 11. F. Schäffler, Semicond. Sci. Technol. 12, 1515 (1997); M. V. Fischetti, F. Gámiz, and W. Hänsch, J. Appl. Phys. 92, 7320 (2002). 12. T. A. Langdo, A. Lochtefel, M. T. Currie, R. Hammond, V. K. Yang, A. Carlin, J. Vineis, G. Braithwaite, H. Badawi, M. T. Bulsara, and E. G. Fitzgerald, Proc. IEEE Int. SOI conf. 211 (2002). 13. M. Reiche, O. Moutanabbir, C. Himcinschi, S. Christiansen, W. Erfurth, U. Gösele, S. Mantl, D. Buca, Q. T. Zhao, R. Loo, D. Nguyen, F. Muster, and M. Petzold, ECS Transactions 16, 311 (2008). 14. C. K. Celler, and S. Cristoloveanu, J. Appl. Phys. 93, 4955 (2003). 15. I. de Wolf, Semi. Sci. Technol. 11, 139 (1996). 16. A. Hähnel, M. Reiche, O. Moutanabbir, H. Blumtritt, H. Geisler, J. Hoentschel , and H.-J. Engelmann, Submitted for publication. 17. R. Bierwolf, H. Hohenstein, F. Philipp, O. Brandt, G.E. Crook, K. Ploog, Ultramicroscopy 49, 273 (1993). 18. A. Rosenauer, Transmission Electron Microscopy of Semiconductor Nanostructures An Analysis of Composition and Strain State, Springer Tracts in Modern Physics, Vol. 182, Springer-Verlag Berlin Heidelberg New York, 2003, ISBN 3-540-00414-9 19. P. Galindo, S. Kret, A.M. Sanchez, J.-Y. Laval,A. Yanez, J. Pizarro, E. Guerrero, T. Ben, S.I. Molina, Ultramicroscopy 107, 186 (2007). 20. D.L. Sales, J. Pizarro, P.L. Galindo, R. Garcia, G. Trevisi, P. Frigeri, L. Nasi, S. Franchi, S.I. Molina, Nanotechnology 18, 475503 (2007).



21. PPA-software, the GatanTM DigitalMicrograph plug-in for High-Resolution Peak Measurement and Strain Mapping Analysis, Version 1.01, HREM research Inc., http://www.hremresearch.com/ 22. M.J. Hÿtch, E. Snoeck, R. Kilaas, Ultramicroscopy 74, 131 (1998) 131. 23. M.J. Hÿtch, T. Plamann, Ultramicroscopy 87, 199 (2001). 24. T. S. Drake, C. Ní Chléirigh, M. L. Lee, A. J. Pitera, E. A. Fitzgerald, D. A. Antoniadis, D. H. Anjum, J. Li, R. Hull, N. Klymko, and J. L. Hoyt, Appl. Phys. Lett. 83,875 (2003). 25. T. A. Langdo, M. T. Currie, A. Lochtefled, R. Hammond, J. A., Carlin, M. E. Erdtmann, G. Braithwaite, V. K. Yang, C. J. Vineis, H. Badawi, and M. T. Bulsara, Appl. Phys. Lett. 82, 4256 (2003). 26. S. Baudot, F. Andrieu, F. Rieutord, and J. Eymery, J. Appl. Phys. 105,114302 (2009). 27. R. Z. Lei, W. Tsai, I. Aberg, T. B. O’Reilly, J. L. Hoyt, D. A. Antoniadis, H. I. Smith, A. J. Paul, M. L. Green, J. Li, and R. Hull, Appl. Phys. Lett. 87, 251926 (2005). 28. O. Moutanabbir, M. Reiche, W. Erfurth, F. Naumann, M. Petzold, and U. Gösele, Appl. Phys. Lett. 94, 243113 (2009). 29. D. Gu, M. Zhu, G. K. Celler, and H. Baumgart, Electrochem. Solid-State Lett. 12, H113 (2009). 30. O. Moutanabbir, M. Reiche, A. Hähnel, W. Erfurth, U. Gösele, M. Motohashi, A. Tarun, N. Hayazawa, and S. Kawata, Nanotechnology 21, 134013 (2010). 31. O. Moutanabbir, M. Reiche, A. Hähnel, W. Erfurth, M. Motohashi, A. Tarun, N. Hayazawa, and S. Kawata, Appl. Phys. Lett. 96, 233105 (2010). 32. O. Moutanabbir, M. Reiche, A. Hähnel, M. Oehme, and E. Kasper, Submitted.
