- Email: [email protected]
Strain mapping in MOSFETS by high-resolution electron microscopy and electron holography
Materials Science and Engineering B 154–155 (2008) 221–224
Contents lists available at ScienceDirect
Materials Science and Engineering B journal homepage: www.elsevier.com/locate/mseb
Strain mapping in MOSFETS by high-resolution electron microscopy and electron holography Florian Hüe, Martin Hytch ∗ , Florent Houdellier, Etienne Snoeck, Alain Claverie CEMES-CNRS, 29 rue Jeanne Marvig, 31055 Toulouse, France
a r t i c l e
i n f o
Article history: Received 24 May 2008 Received in revised form 11 September 2008 Accepted 8 October 2008 Keywords: Strained silicon Strain metrology Electron microscopy Electron holography
a b s t r a c t We present two methods for mapping strains in MOSFETs at the nanometer scale. Aberration-corrected high-resolution transmission electron microscopy (HRTEM) coupled with geometric phase analysis (GPA) provides sufficient signal-to-noise to accurately determine strain fields across the active regions of devices. Finite element method (FEM) simulations are used to confirm our measurements. The field of view is however limited to about 100 nm2 . To overcome this, we have developed a new technique called dark-field holography based on off-axis electron holography and dark-field imaging. This new technique provides us a better strain resolution than HRTEM, a spatial resolution of 4 nm and a field of view of 1 m. © 2008 Elsevier B.V. All rights reserved.
1. Introduction Strained silicon is now an integral feature of the latest generation of transistors and electronic devices because of the associated enhancement in carrier mobility [1]. Tensile strain leads to a splitting of the degeneracy of the conduction band minima in the 0 0 1 direction resulting in reduced intervalley scattering. Electron mobility in biaxially strained-Si (s-Si) grown on virtual substrates of Si1−x Gex can be increased by a factor 2 for x > 20% [2]. Furthermore, strain lowers the energy of the heavy hole and spin-orbit bands relative to the light hole band. For x > 40% the hole mobility in s-Si can be increased by more than two [3]. Other techniques can also be used to introduce strain into the channel such as embedded SiGe or Si-C source and drain or with nitride stressors. These techniques lead to an improvement of the carrier mobility for both electrons and holes. Recent analysis showed that uniaxial strain is 3 times more efficient than biaxial strain [4]. With the continuous reduction in scale of devices, strain has become increasingly difficult to measure. Developing methods of strain measurement at the nanoscale has therefore been a major goal of recent years but has proven illusive in practice [5]. Raman spectroscopy or X-ray diffraction techniques can map strain only at the micron scale. At smaller scale, transmission electron microscopy (TEM) techniques such as convergent-beam electron
∗ Corresponding author. Tel.: +33 562257883; fax: +33 562257897. E-mail address: [email protected] (M. Hytch). 0921-5107/$ – see front matter © 2008 Elsevier B.V. All rights reserved. doi:10.1016/j.mseb.2008.10.020
diffraction (CBED) [6] and nano beam diffraction (NBD) [7] can be used. However, CBED can sometimes be too sensitive to foil bending in the highly strained areas typical of devices [8] and none of the techniques combine the necessary spatial resolution, precision and field of view. An alternative method for measuring local strains is highresolution transmission electron microscopy (HRTEM) combined with the image processing technique of geometric phase analysis (GPA) [9,10]. We have shown recently how strains can be mapped across an entire MOSFET including source, channel and drain region [11]. However, because of the relatively high magnifications necessary to image the atomic lattice, the field of view is limited. We will show an example of such HRTEM analysis followed by preliminary results for a new technique, based on electron holography, capable of circumventing all the problems currently plaguing strain characterisation [12].
2. Experimental methods Lamellas of uniform thickness were prepared for TEM observations by mechanical polishing (tripod method) followed by focused-ion beam (FIB) using a Cross Beam XB 1540 (Zeiss): a scanning electron microscope (Gemini Zeiss) combined with a FIB column (Orsay Physics). Specimens are thinned to have the two surface normals parallel to the [1 1¯ 0] zone axis and are typically 80 nm thick for HRTEM observations, and 200 nm thick for electron holography experiments.
222
F. Hüe et al. / Materials Science and Engineering B 154–155 (2008) 221–224
TEM experiments were performed on the SACTEM-Toulouse, a Tecnai F20 ST (FEI) fitted with an imaging aberration corrector (CEOS), 2048 × 2048 CCD camera (Gatan) and a rotatable electrostatic biprism (FEI). HRTEM images were obtained at the [1 1¯ 0] zone axis in nominally zero spherical aberration conditions, and at relatively low magnification (×200,000) in order to maximize the field of view to about 100 nm square. For electron holography, specimens were oriented to a two-beam condition close to the [1 1¯ 0] zone axis. The microscope was operated in Lorentz mode [13] at a nominal magnification ×20,000 to obtain a field of view of 1 m. The voltage applied to the electron biprism was 80 V producing holographic fringes of 2 nm spacing and overlap width of 250 nm. Images were analysed using a modified version of the software package GPA Phase 2.0 (HREM Research Inc.) a plug-in for DigitalMicrograph (Gatan). Image processing uses a mask in Fourier space which determines the real-space spatial resolution of the measurements: 5 nm for HRTEM measurements and 4 nm for electron holography. A detailed explanation of geometric phase analysis can be found in ref. [9]. Images were corrected for the geometrical distortions introduced by the CCD camera and the projector lenses of the microscope [14]. Strains were simulated in the p-MOSFET by the finite element method (FEM) based on linear anisotropic elastic theory. Domains of different chemical composition are distinguished by their elastic coefficients and lattice parameters (determined by applying Vergard’s law to the bulk values for Si and Ge). The geometry of the model was based on the bright-field images of the observed structures. Epitaxy is treated as a thermal expansion problem [15]. Relaxation of the different domains is principally governed by the
elastic tensor in each domain and the boundary conditions. Along the x-axis, in the source–drain direction, the lamella is considered to be infinite by using periodic boundary conditions. In the z-direction of growth, the lower boundary in the substrate is held fix and the upper surface treated as a free surface. For the bulk simulation, the y direction (electron beam direction) is infinite with periodic boundary conditions. Simulations were carried out for a 100 nm thick lamella with two free surfaces (corresponding to the TEM samples) and an infinitely thick sample (corresponding to the bulk structure). 3. Experimental results 3.1. High-resolution electron microscopy An example of an HRTEM image of a p-MOSFET with a dummy gate and Si80 Ge20 source and drain (schematically shown in Fig. 1a) is shown in Fig. 1b. The contrast is poor compared to typical HRTEM because of the thickness of the specimen and the damage layers caused by FIB preparation. Indeed without the aberration correction, the images would not have sufficient signal-to-noise to carry out meaningful analysis [11]. The 2-dimensional strain components were determined using GPA with the x-axis defined by the source–drain direction. The map of the εxx component (Fig. 1d) has a spatial resolution of 5 nm and a precision of 0.3% (determined by the standard deviation of the variations within the substrate). The reference used for the undeformed lattice was chosen in the substrate, as far from the gate as possible but still within the image. Any strain in this region will add a constant value to the measured
Fig. 1. HRTEM strain mapping of a uniaxial 45 nm strained silicon channel pinched between Si80 Ge20 source (S) and drain (D): (a) schematic representation of the p-MOSFET device; (b) experimental HRTEM image; (c) finite element method (FEM) simulation for thick specimen of εxx strain component parallel to source–drain direction; (d) experimental εxx strain component.
F. Hüe et al. / Materials Science and Engineering B 154–155 (2008) 221–224
strains, thus introducing a systematic error. We therefore compare our observation with our calibrated FEM model taking into account the geometry of the device, the boundaries conditions and the elastic constants of the different materials. Experimental and simulated values are in good agreement (the colour scale is identical for Fig. 1c and d). The channel is in a compression state in x (εxx = −1%) reach-
223
ing −1.2% just below the gate. A detailed analysis of all the strain components can be carried out in this way [11]. Despite the success of the HRTEM technique there are drawbacks such as the noise due to the thickness of the specimen or the amorphous layers on both sides of the lamella from the FIB preparation. A more serious problem is the limited field of view. 3.2. Dark-field electron holography We have recently invented a new technique capable of measuring strain to high precision, with nanometre spatial resolution and for micron fields of view [16]. The method combines the advantages of the conventional moiré technique with the flexibility of off-axis electron holography and is applicable to standard FIB prepared samples. Two diffracted beams (one from the bulk region, another from the region of interest) are interfered with the aid of an electrostatic biprism located below the specimen. Analysis of the interference fringes allow us to determine the local variations in the lattice parameter [12]. We have performed experiments with similar samples as shown previously for HRTEM (Fig. 2). The components of the 2dimensional strain tensor can be mapped across a wide field of view (Fig. 2b and c). The results have been confirmed with FEM simulations [12]. The major interest of this new technique is the large field of view that we can obtain (depending of the biprism voltage). In our case, it is possible to image three, even four transistors side by side and compare strain mapping between each other. Dark-field holography is highly accurate even close to the gate area of the transistor (Fig. 2d). 4. Conclusions Our results show that the distribution of strain within the active area of a transistor can be imaged and measured to high accuracy with a spatial resolution of about 4 nm. We present two methods of strain mapping, one based on HRTEM and another based on electron holography. The former is easier to apply but restricted to small areas. The latter is very accurate with a larger field of view but requires a biprism and a special lens mode for the TEM. However, this new method is the only technique combining high accuracy and large field of view and will probably be extensively used due to the growing role of strain in future devices. Acknowledgements F. Hüe is grateful to the CEA-Leti for financial support. This work was partially supported by the European Union through the projects PullNano (Pulling the limits of nanoCMOS electronics, IST026828) and ESTEEM (Enabling Science and Technology through European Electron Microscopy, IP3: 0260019). We thank P. Mooney (Gatan) for supplying the CCD camera calibration data and H. Bender (IMEC) for the samples. References
Fig. 2. Holographic dark-field strain mapping of short channel p-MOSFET with Si80 Ge20 source and drain: (a) conventional bright-field TEM image; (b) experimentally measured mean dilatation; (c) experimentally measured εxx strain component parallel to source–drain direction; (d) experimental profile of εxx strain as a function of distance from the gate, averaged over channel width.
[1] ITRS, International Technology Roadmap for Semiconductors 2005 Edition. Available online at http://www.itrs.net/reports.html. [2] M.L. Lee, E.A. Fitzgerald, M.T. Bulsara, M.T. Currie, A. Lochtefeld, J. Appl. Phys. 97 (2005) 011101. [3] C.W. Leitz, M.T. Currie, M.L. Lee, Z.Y. Cheng, D.A. Antoniadis, E.A. Fitzgerald, J. Appl. Phys. 92 (2002) 3745. [4] S.E. Thompson, G.Y. Sun, Y.S. Choi, T. Nishida, IEEE Trans. Electron Devices 53 (2006) 1010. [5] B. Foran, M.H. Clark, G. Lian, Future Fab Int. 20 (2006) 127. [6] P. Zhang, A.A. Istratov, E.R. Weber, C. Kisielowski, H.F. He, C. Nelson, J.C.H. Spence, Appl. Phys. Lett. 89 (2006) 161907.
224
F. Hüe et al. / Materials Science and Engineering B 154–155 (2008) 221–224
[7] K. Usuda, T. Numata, T. Irisawa, N. Hirashita, S. Takagi, Mater. Sci. Eng. B 124 (2005) 143. [8] L. Clément, R. Pantel, L.F.T. Kwakman, J.-L. Rouvière, Appl. Phys. Lett. 85 (2004) 651. ¨ [9] M.J. Hytch, E. Snoeck, R. Kilaas, Ultramicroscopy 74 (1998) 131. ¨ [10] M.J. Hytch, J.-L. Putaux, J.-M. Pénisson, Nature 423 (2003) 270. ¨ [11] F. Hüe, M.J. Hytch, H. Bender, F. Houdellier, A. Claverie, Phys. Rev. Lett. 100 (2008) 156602. ¨ [12] M.J. Hytch, F. Houdellier, F. Hüe, E. Snoeck, Nature 453 (2008) 1086.
[13] E. Snoeck, P. Hartel, H. Mueller, M. Haider, P.C. S Tiemeijer, Proceedings of 16th International Microscopy Congress, 2, Japanese Society of Microscopy, Sapporo, 2006, p. 730. ¨ [14] F. Hüe, C.L. Johnson, S. Lartigue-Korinek, G. Wang, P.R. Buseck, M.J. Hytch, J. Electron Microsc. 54 (2005) 181. [15] S. Christiansen, M. Albrecht, H.P. Strunk, H.J. Maier, Appl. Phys. Lett. 64 (1994) 3617. ¨ [16] M.J. Hytch, F. Houdellier, F. Hüe, E. Snoeck, Patent Application FR N 07 06711 (2007).
Contents lists available at ScienceDirect
Materials Science and Engineering B journal homepage: www.elsevier.com/locate/mseb
Strain mapping in MOSFETS by high-resolution electron microscopy and electron holography Florian Hüe, Martin Hytch ∗ , Florent Houdellier, Etienne Snoeck, Alain Claverie CEMES-CNRS, 29 rue Jeanne Marvig, 31055 Toulouse, France
a r t i c l e
i n f o
Article history: Received 24 May 2008 Received in revised form 11 September 2008 Accepted 8 October 2008 Keywords: Strained silicon Strain metrology Electron microscopy Electron holography
a b s t r a c t We present two methods for mapping strains in MOSFETs at the nanometer scale. Aberration-corrected high-resolution transmission electron microscopy (HRTEM) coupled with geometric phase analysis (GPA) provides sufficient signal-to-noise to accurately determine strain fields across the active regions of devices. Finite element method (FEM) simulations are used to confirm our measurements. The field of view is however limited to about 100 nm2 . To overcome this, we have developed a new technique called dark-field holography based on off-axis electron holography and dark-field imaging. This new technique provides us a better strain resolution than HRTEM, a spatial resolution of 4 nm and a field of view of 1 m. © 2008 Elsevier B.V. All rights reserved.
1. Introduction Strained silicon is now an integral feature of the latest generation of transistors and electronic devices because of the associated enhancement in carrier mobility [1]. Tensile strain leads to a splitting of the degeneracy of the conduction band minima in the 0 0 1 direction resulting in reduced intervalley scattering. Electron mobility in biaxially strained-Si (s-Si) grown on virtual substrates of Si1−x Gex can be increased by a factor 2 for x > 20% [2]. Furthermore, strain lowers the energy of the heavy hole and spin-orbit bands relative to the light hole band. For x > 40% the hole mobility in s-Si can be increased by more than two [3]. Other techniques can also be used to introduce strain into the channel such as embedded SiGe or Si-C source and drain or with nitride stressors. These techniques lead to an improvement of the carrier mobility for both electrons and holes. Recent analysis showed that uniaxial strain is 3 times more efficient than biaxial strain [4]. With the continuous reduction in scale of devices, strain has become increasingly difficult to measure. Developing methods of strain measurement at the nanoscale has therefore been a major goal of recent years but has proven illusive in practice [5]. Raman spectroscopy or X-ray diffraction techniques can map strain only at the micron scale. At smaller scale, transmission electron microscopy (TEM) techniques such as convergent-beam electron
∗ Corresponding author. Tel.: +33 562257883; fax: +33 562257897. E-mail address: [email protected] (M. Hytch). 0921-5107/$ – see front matter © 2008 Elsevier B.V. All rights reserved. doi:10.1016/j.mseb.2008.10.020
diffraction (CBED) [6] and nano beam diffraction (NBD) [7] can be used. However, CBED can sometimes be too sensitive to foil bending in the highly strained areas typical of devices [8] and none of the techniques combine the necessary spatial resolution, precision and field of view. An alternative method for measuring local strains is highresolution transmission electron microscopy (HRTEM) combined with the image processing technique of geometric phase analysis (GPA) [9,10]. We have shown recently how strains can be mapped across an entire MOSFET including source, channel and drain region [11]. However, because of the relatively high magnifications necessary to image the atomic lattice, the field of view is limited. We will show an example of such HRTEM analysis followed by preliminary results for a new technique, based on electron holography, capable of circumventing all the problems currently plaguing strain characterisation [12].
2. Experimental methods Lamellas of uniform thickness were prepared for TEM observations by mechanical polishing (tripod method) followed by focused-ion beam (FIB) using a Cross Beam XB 1540 (Zeiss): a scanning electron microscope (Gemini Zeiss) combined with a FIB column (Orsay Physics). Specimens are thinned to have the two surface normals parallel to the [1 1¯ 0] zone axis and are typically 80 nm thick for HRTEM observations, and 200 nm thick for electron holography experiments.
222
F. Hüe et al. / Materials Science and Engineering B 154–155 (2008) 221–224
TEM experiments were performed on the SACTEM-Toulouse, a Tecnai F20 ST (FEI) fitted with an imaging aberration corrector (CEOS), 2048 × 2048 CCD camera (Gatan) and a rotatable electrostatic biprism (FEI). HRTEM images were obtained at the [1 1¯ 0] zone axis in nominally zero spherical aberration conditions, and at relatively low magnification (×200,000) in order to maximize the field of view to about 100 nm square. For electron holography, specimens were oriented to a two-beam condition close to the [1 1¯ 0] zone axis. The microscope was operated in Lorentz mode [13] at a nominal magnification ×20,000 to obtain a field of view of 1 m. The voltage applied to the electron biprism was 80 V producing holographic fringes of 2 nm spacing and overlap width of 250 nm. Images were analysed using a modified version of the software package GPA Phase 2.0 (HREM Research Inc.) a plug-in for DigitalMicrograph (Gatan). Image processing uses a mask in Fourier space which determines the real-space spatial resolution of the measurements: 5 nm for HRTEM measurements and 4 nm for electron holography. A detailed explanation of geometric phase analysis can be found in ref. [9]. Images were corrected for the geometrical distortions introduced by the CCD camera and the projector lenses of the microscope [14]. Strains were simulated in the p-MOSFET by the finite element method (FEM) based on linear anisotropic elastic theory. Domains of different chemical composition are distinguished by their elastic coefficients and lattice parameters (determined by applying Vergard’s law to the bulk values for Si and Ge). The geometry of the model was based on the bright-field images of the observed structures. Epitaxy is treated as a thermal expansion problem [15]. Relaxation of the different domains is principally governed by the
elastic tensor in each domain and the boundary conditions. Along the x-axis, in the source–drain direction, the lamella is considered to be infinite by using periodic boundary conditions. In the z-direction of growth, the lower boundary in the substrate is held fix and the upper surface treated as a free surface. For the bulk simulation, the y direction (electron beam direction) is infinite with periodic boundary conditions. Simulations were carried out for a 100 nm thick lamella with two free surfaces (corresponding to the TEM samples) and an infinitely thick sample (corresponding to the bulk structure). 3. Experimental results 3.1. High-resolution electron microscopy An example of an HRTEM image of a p-MOSFET with a dummy gate and Si80 Ge20 source and drain (schematically shown in Fig. 1a) is shown in Fig. 1b. The contrast is poor compared to typical HRTEM because of the thickness of the specimen and the damage layers caused by FIB preparation. Indeed without the aberration correction, the images would not have sufficient signal-to-noise to carry out meaningful analysis [11]. The 2-dimensional strain components were determined using GPA with the x-axis defined by the source–drain direction. The map of the εxx component (Fig. 1d) has a spatial resolution of 5 nm and a precision of 0.3% (determined by the standard deviation of the variations within the substrate). The reference used for the undeformed lattice was chosen in the substrate, as far from the gate as possible but still within the image. Any strain in this region will add a constant value to the measured
Fig. 1. HRTEM strain mapping of a uniaxial 45 nm strained silicon channel pinched between Si80 Ge20 source (S) and drain (D): (a) schematic representation of the p-MOSFET device; (b) experimental HRTEM image; (c) finite element method (FEM) simulation for thick specimen of εxx strain component parallel to source–drain direction; (d) experimental εxx strain component.
F. Hüe et al. / Materials Science and Engineering B 154–155 (2008) 221–224
strains, thus introducing a systematic error. We therefore compare our observation with our calibrated FEM model taking into account the geometry of the device, the boundaries conditions and the elastic constants of the different materials. Experimental and simulated values are in good agreement (the colour scale is identical for Fig. 1c and d). The channel is in a compression state in x (εxx = −1%) reach-
223
ing −1.2% just below the gate. A detailed analysis of all the strain components can be carried out in this way [11]. Despite the success of the HRTEM technique there are drawbacks such as the noise due to the thickness of the specimen or the amorphous layers on both sides of the lamella from the FIB preparation. A more serious problem is the limited field of view. 3.2. Dark-field electron holography We have recently invented a new technique capable of measuring strain to high precision, with nanometre spatial resolution and for micron fields of view [16]. The method combines the advantages of the conventional moiré technique with the flexibility of off-axis electron holography and is applicable to standard FIB prepared samples. Two diffracted beams (one from the bulk region, another from the region of interest) are interfered with the aid of an electrostatic biprism located below the specimen. Analysis of the interference fringes allow us to determine the local variations in the lattice parameter [12]. We have performed experiments with similar samples as shown previously for HRTEM (Fig. 2). The components of the 2dimensional strain tensor can be mapped across a wide field of view (Fig. 2b and c). The results have been confirmed with FEM simulations [12]. The major interest of this new technique is the large field of view that we can obtain (depending of the biprism voltage). In our case, it is possible to image three, even four transistors side by side and compare strain mapping between each other. Dark-field holography is highly accurate even close to the gate area of the transistor (Fig. 2d). 4. Conclusions Our results show that the distribution of strain within the active area of a transistor can be imaged and measured to high accuracy with a spatial resolution of about 4 nm. We present two methods of strain mapping, one based on HRTEM and another based on electron holography. The former is easier to apply but restricted to small areas. The latter is very accurate with a larger field of view but requires a biprism and a special lens mode for the TEM. However, this new method is the only technique combining high accuracy and large field of view and will probably be extensively used due to the growing role of strain in future devices. Acknowledgements F. Hüe is grateful to the CEA-Leti for financial support. This work was partially supported by the European Union through the projects PullNano (Pulling the limits of nanoCMOS electronics, IST026828) and ESTEEM (Enabling Science and Technology through European Electron Microscopy, IP3: 0260019). We thank P. Mooney (Gatan) for supplying the CCD camera calibration data and H. Bender (IMEC) for the samples. References
Fig. 2. Holographic dark-field strain mapping of short channel p-MOSFET with Si80 Ge20 source and drain: (a) conventional bright-field TEM image; (b) experimentally measured mean dilatation; (c) experimentally measured εxx strain component parallel to source–drain direction; (d) experimental profile of εxx strain as a function of distance from the gate, averaged over channel width.
[1] ITRS, International Technology Roadmap for Semiconductors 2005 Edition. Available online at http://www.itrs.net/reports.html. [2] M.L. Lee, E.A. Fitzgerald, M.T. Bulsara, M.T. Currie, A. Lochtefeld, J. Appl. Phys. 97 (2005) 011101. [3] C.W. Leitz, M.T. Currie, M.L. Lee, Z.Y. Cheng, D.A. Antoniadis, E.A. Fitzgerald, J. Appl. Phys. 92 (2002) 3745. [4] S.E. Thompson, G.Y. Sun, Y.S. Choi, T. Nishida, IEEE Trans. Electron Devices 53 (2006) 1010. [5] B. Foran, M.H. Clark, G. Lian, Future Fab Int. 20 (2006) 127. [6] P. Zhang, A.A. Istratov, E.R. Weber, C. Kisielowski, H.F. He, C. Nelson, J.C.H. Spence, Appl. Phys. Lett. 89 (2006) 161907.
224
F. Hüe et al. / Materials Science and Engineering B 154–155 (2008) 221–224
[7] K. Usuda, T. Numata, T. Irisawa, N. Hirashita, S. Takagi, Mater. Sci. Eng. B 124 (2005) 143. [8] L. Clément, R. Pantel, L.F.T. Kwakman, J.-L. Rouvière, Appl. Phys. Lett. 85 (2004) 651. ¨ [9] M.J. Hytch, E. Snoeck, R. Kilaas, Ultramicroscopy 74 (1998) 131. ¨ [10] M.J. Hytch, J.-L. Putaux, J.-M. Pénisson, Nature 423 (2003) 270. ¨ [11] F. Hüe, M.J. Hytch, H. Bender, F. Houdellier, A. Claverie, Phys. Rev. Lett. 100 (2008) 156602. ¨ [12] M.J. Hytch, F. Houdellier, F. Hüe, E. Snoeck, Nature 453 (2008) 1086.
[13] E. Snoeck, P. Hartel, H. Mueller, M. Haider, P.C. S Tiemeijer, Proceedings of 16th International Microscopy Congress, 2, Japanese Society of Microscopy, Sapporo, 2006, p. 730. ¨ [14] F. Hüe, C.L. Johnson, S. Lartigue-Korinek, G. Wang, P.R. Buseck, M.J. Hytch, J. Electron Microsc. 54 (2005) 181. [15] S. Christiansen, M. Albrecht, H.P. Strunk, H.J. Maier, Appl. Phys. Lett. 64 (1994) 3617. ¨ [16] M.J. Hytch, F. Houdellier, F. Hüe, E. Snoeck, Patent Application FR N 07 06711 (2007).