Acta Materialia 55 (2007) 46574665 www.elsevier.

com/locate/actamat

3-D Raman spectroscopy measurements of the symmetry of residual stress elds in plastically deformed sapphire crystals
Thomas Wermelinger, Cesare Borgia, Christian Solenthaler, Ralph Spolenak
Laboratory for Nanometallurgy, Department of Materials, ETH Zurich, Wolfgang-Pauli-Strasse 10, 8093 Zurich, Switzerland Received 19 February 2007; received in revised form 10 April 2007; accepted 13 April 2007 Available online 12 June 2007

Abstract The development of methods to characterize materials in three dimensions, such as tomography by X-rays, focused ion beam and electrons, has led to progress in the understanding of materials properties. Recently, even stress and deformation tensors could be measured in three dimensions. Specically the stress elds around indents in metals were studied by three-dimensional (3-D) X-ray stress microscopy. In this paper, we investigate the 3-D residual stress eld around a microindent using confocal Raman microscopy with a lateral resolution of 300 nm and a depth resolution of 600 nm. The model system investigated was single crystalline sapphire, which was indented normal to its basal c(0 0 0 1) plane. A cross-section of the indent was studied by transmission electron microscopy to visualize the deformed microstructure. The major result is that the geometry of the indenter has no direct inuence on the symmetry of the resulting residual stress eld. Residual stresses directly depend on the crystal symmetry and the defect structures formed during indentation. Confocal Raman microscopy is a powerful method for analyzing 3-D stress elds and the corresponding defect structures (by peak width analysis) with a resolution in the submicron range. 2007 Acta Materialia Inc. Published by Elsevier Ltd. All rights reserved.
Keywords: Raman spectroscopy; Residual stress; Phase transformation; Microindentation; Ceramics

1. Introduction Novel applications in the microelectronics industry have intensied interest in sapphire as a substrate [1]. Various sapphire-based devices, such as fast neutron lters and high-pressure magnetic resonance cells, have been developed. Specically, the high strength and heat resistance of sapphire attracts a lot of attention. Therefore a profound knowledge of the mechanical properties is required. In this context, the deformation of sapphire has been extensively studied by several groups [112]. As a result, the deformation and fracture mechanisms of the sapphire single crystal are well known. In contrast, magnitude and distribution of residual internal stresses after plastic deformation, i.e. the properties of residual stress elds, still

Corresponding author. E-mail address: ralph.spolenak@mat.ethz.ch (R. Spolenak).

remain unclear. Since these stresses may cause failure, and hence aect the lifetime of devices, it is important to measure and to visualize them, ideally in three dimensions (3-D). At the moment, only a few methods are known to perform 3-D measurements with a resolution at or below the micrometer range. One method is 3-D X-ray diraction microscopy, which has a resolution in the micrometer range [13]. Another is Xray microbeam diraction, which is an experimental option to measure the strain tensor with a submicrometer resolution [1416] in three dimensions. The 3-D distribution of the local crystalline phase, the texture and the elastic strain tensor can all be measured with a resolution below 1 lm. These instruments combine ultraintense synchrotron Xray sources and advanced X-ray optics to probe crystalline materials. Only a few such systems are available world wide, and until now only a few such experiments have been performed to date. Consequently, alternative methods on the laboratory scale are needed.

1359-6454/$30.00 2007 Acta Materialia Inc. Published by Elsevier Ltd. All rights reserved. doi:10.1016/j.actamat.2007.04.036

4658

T. Wermelinger et al. / Acta Materialia 55 (2007) 46574665

This paper demonstrates that another accurate method for the 3-D measurement of stresses is micro-Raman spectroscopy. The method is based on the well known eect that Raman signals (phonon energies) are aected by internal stresses. In a Raman spectrum of deformed material, peak positions are shifted relative to the peak positions obtained from the stress free material. Quantifying this shift allows determining sign and magnitude of internal stress. As already proven in many 2-D studies [1721], Raman microscopy can successfully be applied in the analysis of stresses in silicon devices and other Raman active materials, such as polymers, semiconductors and ceramics. In transparent Raman active materials the method could also be used for the 3-D case provided a confocal microscope is available that provides precise positioning of the focus spot. Such measurements at a microindent in a sapphire single crystal are reported and discussed below. The crystal structure of sapphire is hexagonal rhombohedral, space group D6 3d , with two molecular Al2O3 groups per unit cell [22,23]. This structure involves seven Raman active phonon modes: 2A1g + 5Eg. According to Louden [24], the Raman tensors of the active optical vibrations of sapphire are     a 0 0  c c d          1 A1g  0 a 0 ; Eg  c c d      0 0 b  d d o  The Raman frequencies (given in relative wavenumbers) of stress free sapphire are 417 and 646 cm1 for the A1g modes and 380, 432, 451, 578 and 751 cm1 for the Eg modes [22]. The shift of these modes as a function of applied stress has been characterized by several groups [25,26]. Watson et al. applied uniaxial pressure up to 1 GPa on macroscopic sapphire pillars and measured the corresponding peak shifts [23]. The results of these experiments are listed in Table 1. The same experiment but on sapphire bers with a diameter in the micrometer range showed slightly dierent results [27]. Here, the shifts were found to be about 1.8 and 1.1 cm1 GPa1 for the peaks at 417 and 380 cm1, respectively. Gallas et al. [26] used a diamond anvil high-pressure cell for the calibration of the Raman eect in sapphire. They received similar results for the peak at 380 cm1; however, for the peak at a rela-

tive wavelength of 417 cm1 they found a pressure dependence of 2.2 cm1 GPa1. Shin and Raccah reported the eect of uniaxial stress on the Raman frequencies in sapphire [28]. Recently, a confocal Raman microscope was used to perform an in-depth analysis of residual stresses in polycrystalline alumina coating [29]. 2. Experimental A sapphire single crystal (size 10 10 1 mm) was plastically deformed by microindentation. The indent was placed on the basal (0 0 0 1) plane with a Vickers microindenter at room temperature. A maximal indentation force of 0.4 N at ambient pressure was applied under constant load rate of 0.05 N s1. The radius of the indenter tip varied from one direction to the other from 0.16 to 1.1 lm. The shape of the indent was determined using an atomic force microscope (AFM) (CRAFM 200, WITec GmbH, Germany). The residual stress eld around the indent was measured with a confocal Raman microscope (CRM 200, WITec GmbH, Germany) equipped with a heliumcadmium laser with a wavelength k of 442 nm. The microscope was operated in the backscattering mode with a 100 objective and a numerical aperture NA of 0.9. According to the Rayleigh equation kLaser 2 2 NA the lateral resolution d of the microscope is about 300 nm, while the depth resolution is on the order of 600 nm. As the microscope is confocal only information from the focal spot is collected. Confocality combined with the possibility of a precise positioning of the focus spot in the x-, y- and zdirections allows for measuring stress elds in three dimensions. Starting from the surface of the sapphire crystal, a stack of planar Raman scans parallel to the basal (0 0 0 1) plane was recorded. Measurements were made always within the same x- and y-coordinates but changing z-position. The distance between two planar scans was 0.4 lm. In total, 19 scans were recorded down to depth of 7.6 lm. Each planar scan was 12 12 lm and consisted of 48 48 spectra. As the Raman signal of sapphire was weak compared with signals from silicon, an integration time of 5 s per spectrum was chosen to obtain a suciently large signal-to-noise ratio. In order to gain some direct information on the deformation structure, a cross-section trough the indent was examined by means of transmission electron microscopy (TEM). For this purpose a TEM lamella was cut perpendicular to the (0 0 0 1) basal plane through the center of the indent (Fig. 2a) with a FEI Nano Lab 600 focused ion beam microscope (FIB). The TEM lamella was about 2.5 15 lm in-plane and 100150 nm deep; it was examined with a FEI Tecnai G2 F20X-Twin TEM operating at 200 kV. d 1:22

Table 1 Pressure dependence of Raman frequencies in sapphire Frequency (cm1) Peak shift (cm1/GPa) Shin et al. [27] 2.3 0.2 1.7 0.1 1.8 0.1 1.0 0.2 2.7 0.3 5.0 0.4 2.5 0.3 Watson et al. [22] 1.37 0.06 2.11 0.06 2.95 0.08 1.66 0.10 2.77 0.12 4.80 0.20 Gallas et al. [25] 1.10 0.10 2.20 0.07 Jia and Yen [26] 1.10 1.80 1.16 0.55 0.00 1.70

380 417 432 451 578 646 751

T. Wermelinger et al. / Acta Materialia 55 (2007) 46574665

4659

3. Results Fig. 1 shows the Raman eect of undeformed sapphire for two dierent crystal orientations and corresponding dierent polarization of the incoming laser light. The black spectrum (rectangular symbols) was measured with an incident laser beam polarized perpendicular to the c(0 0 0 1) axis and features six major lines at the wavenumbers 378, 417, 428, 447, 575 and 748 cm1. The gray spectrum (triangular symbols) was recorded with an incident laser beam polarized parallel to the c(0 0 0 1) axis and features three prominent lines at 378, 417, 645 cm1, plus a faint peak at 748 cm1. These measurements with the actual experimental set-up are in agreement with literature [22]. The peak at 378 cm1 gets stronger, compared with the other peaks, with increasing numerical aperture of the objective. Therefore, this particular vibration must have a stronger in-plane component than the other phonons due to the fact that a parallel laser beam only interacts with out-of-plane phonons. Fig. 2a shows an AFM map of the topography of the Vickers microindent on the c(0 0 0 1) plane of the sapphire single crystal. The indent has a diameter of about 4.5 5 lm. No pile-ups are visible at the edges of the indent. Though there are some slightly asymmetric features, such as the tip radius, the shape of the plastically deformed area corresponds to the fourfold symmetry of the indenter tip. A cross-section through the middle of the indent shows that the indent has a depth of about 0.5 lm (see Fig. 2b). All measured Raman peaks from the indented area are shifted due to residual stresses. Fig. 3ad shows the shift of four dierent peaks measured 0.4 lm below the surface of the sample. While a shift to higher wavenumbers indicates compressive stresses, shifts to lower wavenumbers are due to tensile stresses relative to the direction of the phonon probed. The map in Fig. 3a contains a mixture

Fig. 2. (a) AFM image of the Vickers microindent in the basal (0 0 0 1) plane of the sapphire single crystal. The indent has a diameter of about 4.55 lm and clearly shows a fourfold symmetry. The trace T marks the position of the TEM sample. (b) Cross-section through the middle of the indent which has a depth of approximately 0.5 lm.

Fig. 1. Raman spectrum of sapphire for dierent polarization of the incoming light. Black spectrum (rectangular): polarization perpendicular to c(0 0 0 1) axis; gray spectrum (triangular): polarization parallel to c(0 0 0 1) axis. Objective 100 magnication, numerical aperture 0.9.

between in-plane and out-of-plane stresses; the other maps in Fig. 3 represent out-of-plane stresses only. All peaks show a similar threefold stress distribution. The clarity of the pattern is obviously aected by the signal-to-noise ratio of the dierent peaks. The peak at the wavenumber at 417 cm1 has the best signal-to-noise ratio and is therefore selected for the further investigations. Fig. 4 shows the stress distribution over the indent area in terms of the shift of the peak near 417 cm1 for increasing depth, starting form the surface of the crystal down to 3.6 lm. Bright colors correspond to compressive stresses while dark colors correspond to tensile stresses in the outof-plane direction. Threefold symmetry is apparent in all scans. The measurements were made down to 7.6 lm but the threefold symmetry disappears 3.6 lm below the surface. Image processing tools such as ImageG allow the calculation of a 3-D image from a stack of 2-D maps. Fig. 5

4660

T. Wermelinger et al. / Acta Materialia 55 (2007) 46574665

Fig. 3. Stress maps (12 12 lm) at the indent 0.4 lm below the surface, for the peaks at wavenumbers (a) 378 cm1, (b) 417 cm1, (c) 575 cm1 and (d) 748 cm1. In maps (b)(d) bright colors correspond to compressive stresses while dark colors correspond to tensile stresses in the out-of-plane direction. Map (a) contains an in-plane as well as an out-of-plane component.

Fig. 4. Shift of the peak at 417 cm1 wavenumbers around the indent. High wavenumbers correspond to compressive stress while low wavenumbers correspond to tensile stress. The map distance is 0.4 lm.

shows the 3-D stress eld around the Vickers indent calculated from the stack of images in Fig. 4. Only out-of-plane stresses are shown. As every color represents a certain

stress, image processing makes it possible to calculate the volume of dierent stress components. In principle, the volume of phase-transformed regions could also be deter-

T. Wermelinger et al. / Acta Materialia 55 (2007) 46574665

4661

Fig. 5. (a) The calculated 3-D stress eld (20 20 2 lm) around the microindent. Dark colors refer to tensile stresses, bright colors to compressive stresses. (b) The tensile stresses above a certain threshold are transparent. Only areas with neutral or compressive stresses are visible in this volume (12 12 4 lm).

Fig. 6. (a) Peak position of three dierent line scans in the z-direction. Measured in the unstressed part of the sample (see Fig. 3, arrow I), in the middle of the indent where tensile stresses are dominating (Fig. 3, arrow II) and in the middle of the indent where compressive stresses are prominent (Fig. 3, arrow III). (b) The peak width instead of the position.

mined. However, in the current case the volume in the particular area was found to be too small for quantication. As a further result of the stack of images, one can extract single line depth scans. Fig. 6 visualizes the stress distribution of three dierent line scans in z-direction. These scans show the variation of the peak shift with increasing depth. Scans were taken (i) far away from the indent (Fig. 3, arrow I) as well as (ii) within the indented area dominated by tensile stresses (Fig. 3, arrow II) and (iii) within the indent dominated by compressive stresses (Fig. 3e, arrow III). The peak position was determined by tting the measured intensity to a Gaussian function; error bars in Fig. 6 refer to the standard deviation of the tted curve. At a depth of about 2 lm the dierences of the peak position between the three lines nearly vanish. But down to 4 lm the values from the tensile region are slightly below the other two curves. With increasing depth the values of

the line scan far away from the indent are slightly shifted to lower wavenumbers. Fig. 6b shows the variation of the peak width for the same line scans. As before the major dierences are visible down to a depth of about 2 lm. Although the absolute value of the peak shifts for the tensile and the compressive stress are comparable, there are signicant dierences in peak width. As the peak width is inuenced not only by the stress but also by the defect density, it can be deduced that regions of compressive stress exhibit higher defect densities. At the side walls of the indent pronounced and strong changes are observed in the Raman spectrum (Fig. 7). The peaks are broadening and their position is shifted to higher wavenumbers. For example, the peak at 748 cm1 is shifted to 773 cm1. This property is found locally just beneath the side walls of the indent and disappears at a depth of about 1.0 lm below the sample surface.

4662

T. Wermelinger et al. / Acta Materialia 55 (2007) 46574665

Fig. 7. Raman spectra of sapphire taken parallel to the c(0 0 0 1) axis. The gray spectrum (triangles) is taken from the unstressed part of the crystal. The black spectrum (squares) is taken form the middle of the indent.

The TEM contrast image Fig. 8 gives an overview over the very complex deformation structure of the indented area and of the transition region towards the un-deformed material. The sample is a thin lamella cut out through the middle of the indent (trace T in Fig. 2) by means of FIB. Arrow a marks the center of the indent. Dark bands b along the sample surface are platinum layers deposited for sample preparation in order to provide for the required mechanical stability of the lamella. Obviously, the deformed zone is wide and deep and the defect density is high. At least three morphologically dierent areas are observed: (i) an area of intersecting elongated cells oblique

to the c-axis and not parallel to the slopes of the indent (area c); (ii) an area of elongated cells perpendicular to the c-axis, almost without any intersection (area d); and (iii) an almost contrast-free area just below one of the slopes of the indent (area e). From this phenomenological picture only a few general but important conclusions can be drawn as follows. The complex stress distribution seems to activate simultaneously both slip and twinning as competing deformation mechanisms. Directly beneath the indent, both mechanisms seem to become activated with similar probability: interaction of all activated deformation systems leads to the observed intersecting cell structures. More distant from the indent center, deformation is essentially concentrated on the basal plane. Again, the mechanism seems to be a combination of both slip and twinning. The process shears material away from the indent so that a stack of horizontal, parallel cells is formed. The width of these cells increases with the distance from the surface. The tip of the arrow d points to a small horizontal crack. The low-contrast area e features no clear traces from any deformation structure, neither from c nor from d. Therefore, deformation must here involve an additional mechanism, such as, for example, phase transformation (see the following section). Near the surface of the sample a thin layer of high dislocation density is observed (area f) which is obviously not due to the indentation experiment. It is well known that polishing sapphire may induce such a thin layers of defects [2]. 4. Discussion 4.1. Mechanical aspects of indenting sapphire In the previous section it is demonstrated that 3-D Raman measurements of the residual stress eld at a Vickers microindent clearly reveal a threefold stress distribution, though the geometrical shape of the indent is of course fourfold symmetrical. This result is surprising and interesting. It can be explained with the crystallographic anisotropy of plastic deformation in hexagonal crystals and the probability Tc for activating the dierent deformation systems as follows. Several groups investigated the room temperature plasticity of sapphire [37,9,10]. Nowak et al. used a spherical microindenter for the analysis of plastic deformation. According to their work, the probability Tc of activating the cth slip or twinning system depends on the shear stress value sc acting on the particular slip or twinning plane, a constraint factor Kc, denoting the orientation of the indented crystal surface, and the critical shear stress sCRc [1,5]. It is given by sc Kc 3 Tc sCRc Table 2 lists the available deformation systems and critical shear stresses for sapphire with a c/a axis ratio of 2.73.

Fig. 8. TEM image of the FIB lamella which was cut out through the indent. (a) Center of the indent. (b) Platinum layer. (c) Elongated cells. (d) Cell structure perpendicular to the c-axis. (e) Almost contrast-free area. (f) Thin layer of dislocations originating from polishing.

T. Wermelinger et al. / Acta Materialia 55 (2007) 46574665 Table 2 Possible twinning and slip systems of sapphire crystal [5] Symbol 1 2 3 4 5 6 Twinning or slip system  1iK1f0 1 1 2g h0 1 1 h1  1 0 0iK10 0 0 1 h2  1 1 0if0  1 1 2g h2  1 1 0i0 0 0 1 h1 0  1 0if 12 1 0g h1 0  1 0if 1 1 0 1g h1 0  1 0if 1 0 1 2g h1 0  1 0if 1 1 2 3g Description Rhombohedral twinning Basal twinning Rombohedral slip Basal slip Prismatic slip Pyramidal slip Critical shear stress (GPa) 0.111 0.148 3 17 1.2 18

4663

For a spherical indentation with an opening angle a of 130 (180 2 * w, with w = 25), which is comparable to the opening angle of a Vickers indent (a = 136), the probability Tc is shown in Fig. 9 for an indentation on the c(0 0 0 1) plane and a force of 0.218 N. Rhombohedral twinning (see Table 2, mechanism 1) has the highest probabilities. This mechanism has six equivalent glide systems, three of which are favored at this opening angle a. The same is true for mechanism 3 (rhombohedral glide), which also shows a preference for three of its six glide systems, though at a lower probability level. The opening angle a has a strong inuence on the symmetry of deformation and is the main reasons for the appearance of a threefold symmetry in the residual stress eld. A at punch indent, for example, should activate dierent deformation systems and will probably lead to a sixfold symmetry of the residual stress eld. The TEM image Fig. 8 shows that in the actual experimental set-up rhombohedral twinning and rhombohedral slip are the dominating deformation mechanism. This behavior must be due to the very complex stress distribution beneath the indent. According to Fig. 9, the dominat-

ing deformation systems involve threefold symmetry. Therefore it is reasonable to assume that the same property is to be expected for the residual stress eld. The strong changes of the Raman spectra observed directly below the indent cannot be explained with slip and twinning mechanisms alone. However, they might well be due to a phase transformation. Colomban and Havel found similar peak shifts in their work [30]. They stated that, in molecular approximation, the 417 cm1 band originates from an AlO stretching mode of the AlO4 tetrahedron. The shift of this peak to higher wavenumbers indicates a compressed structure. As already explained, this property is detectable down to 1.0 lm. The TEM image shows a bright zone directly under the indent, where the contrast of the image is dierent from the other deformed areas. It highlights that this zone is only visible on one side of the indent. The zone is about 0.5 lm deep. The focus spot has in z-direction a diameter of about 0.6 lm. If one compares the cut out direction of the lamella and the stress distribution, one sees that the asymmetry of the deformation structures is also visible in the stress maps. Thus it seems possible to measure a small signal of the phase-transformed area even if the microscope is focused 1 lm below the surface, which corresponds to the ndings of the measurement. 4.2. 3-D Raman microscopy The lateral resolution of the microscope is restricted by the wavelength of the laser and the numerical aperture. Even in the case for deep blue lasers and oil immersion objectives, which have the highest numerical aperture, the lateral resolution is in the order of 150 200 nm. In comparison with a 3-D X-ray crystal microscope [14], which also has a resolution in the submicrometer range, the lateral resolution of an optimized confocal Raman microscope is slightly better. Other advantages of the confocal Raman microscope are the straightforward sample preparation, the fast measuring method and the good accessibility. Another feature is the direct measurement of several components of the stress tensor. As the dierent peaks belong to dierent phonon vibrations of the crystal, it is possible to relate dierent peaks to particular crystallographic directions. Therefore it is also possible to assign dierent peak shifts to dierent directions. However, some restrictions also have to be taken into account. Every crystal class features dierent Raman active phonon vibrations; thus it is not possible to predict in general how many components of the stress tensor are accessible. Silicon, for example, exhibits a single Raman peak which belongs to triply degenerated optical phonons [31]. By examining the polarization and the direction of the scattered light, all components of the stress tensor can be obtained [32]. For sapphire analyzed in the actual experimental set-up, it is possible to measure an out-of-plane component from the peaks at 417, 578 and 751 cm1

Fig. 9. Probability for activating a slip or twinning system for all directions l around a spherical indent [33]. Numbers 13 correspond to symbols in Table 2.

4664

T. Wermelinger et al. / Acta Materialia 55 (2007) 46574665

Fig. 10. Stresses in the middle of the indent at dierent depth levels.

and a mixed out-of-plane and in-plane component from the peak at 380 cm1. Therefore only the out-of-plane component could be measured directly. As not all components of the stress tensor could be measured, it was crucial make certain assumptions about the stress state in order to calculate stresses. While the surface was assumed to be stress free, below the indent the stress can be assumed to be hydrostatic. The following gure (Fig. 10) shows the calculated stresses in the center of the indent at dierent depth levels based on the results of Watson et al. [23]. This calculation was chosen due to the fact that all three peaks, which belong to out-of-plane stresses, show similar stress values, as to be expected. At the surface of the sapphire the assumption of hydrostatic pressure is not valid. As expected, the out-of-plane stress, which can be estimated from the shift of the peaks at 417, 578 and 751 cm1, is almost zero. In contrast, the peak at the wavenumber 380 cm1, which also has an inplane component, exhibits a strong compressive stress of about 1.7 GPa. Below the surface, where the pressure is assumingly hydrostatic, the out-of-plane stress seems to be in the tensile regime. The peak at 380 cm1 shows no signicant shift, which means that the tensile stress of the out-of-plane component is compensated by the in-plane stress. Only transparent and Raman active materials are suitable for 3-D stress measurements. These are ceramics, diamond and most polymers. This is a drawback in comparison with the 3-D X-ray stress microscope, which can be used for any crystalline material. In the case of ceramics, single crystal samples are preferred because polycrystalline samples involve peak broadening, which makes an accurate determination of the peak position more dicult. Fully amorphous samples are still Raman active, but peaks are in general too broad to allow for an accurate determination of the peak position. 5. Summary and conclusions The residual stress distribution around a Vickers indent in sapphire was analyzed with a confocal Raman micro-

scope. Interestingly, the residual stress eld is not directly inuenced by the geometry of the indent. Although the indent shape had a fourfold symmetry, the residual stress eld was found to be of threefold symmetry. This result can be explained by the property of sapphire. Sapphire has a hexagonal crystal structure and therefore shows an anisotropic deformation. The probability for activating a certain slip or twinning system is strongly orientation dependent. The dominating slip and twinning systems, namely rhombohedral twinning and slip, exhibit a clear threefold symmetry. This symmetry leads to a threefold symmetry of the residual stresses. Moreover, experimental evidence for a phase transformation at the side walls of the indent was found. The exact structure of this proposed phase was not investigated and must be determined in future research. It is shown that confocal Raman microscopy is a powerful tool for analyzing 3-D stress elds with a spatial resolution in the submicrometer range. Depending on the Raman tensors, it is possible to get access to several components of the stress tensor. In the case of sapphire, only an out-of-plane component was measurable due to the fact that the experimental set-up only allowed to analyze out-of-plane vibrating phonons or phonons which had an out-of-plane as well as an in-plane component. In principle, the method can be applied for all Raman active transparent materials at modest expenses and short timescales compared with synchrotron based techniques. Acknowledgements The authors thank Kyburz AG for donating us a single crystal sapphire, and Dr. Steve Reyntjens and Dr. Erwan Soutry from FEI Application Research Lab., Eindhoven, for the preparation of the TEM lamella as well as the TEM analysis. This work was supported by ETH Research Grant TH -39/05-1. References
[1] Nowak R, Manninen T, Li CL, Heiskanen K, Hannula SP, Lindroos V, et al. Anomalous surface deformation of sapphire claried by 3-D FEM simulation of the nanoindentation. Jsme Int J Ser ASolid Mech Mater Eng 2003;46:265. [2] Hirayama TSaT. Lattice strain and dislocations in polished surfaces on sapphire. J Am Ceram Soc 2005;88:2277. [3] Pirouz P, Lawlor BF, Geipel T, BildeSorensen JB, Heuer AH, Lagerlof KPD. On basal slip and basal twinning in sapphire (alphaAl2O3). 2. A new model of basal twinning. Acta Mater 1996;44:2153. [4] Nowak R, Sakai M. The anisotropy of surface deformation of sapphire continuous indentation of triangular indenter. Acta Metall Mater 1994;42:2879. [5] Nowak R, Sekino T, Niihara K. Surface deformation of sapphire crystal. Philos Mag A 1996;74:171. [6] Nowak R, Sekino T, Niihara K. Non-linear surface deformation of the (1 0 (1)over-bar-0) plane of sapphire: identication of the linear features around spherical impressions. Acta Mater 1999;47:4329. [7] Inkson BJ. Dislocations and twinning activated by the abrasion of Al2O3. Acta Mater 2000;48:1883.

T. Wermelinger et al. / Acta Materialia 55 (2007) 46574665 [8] BildeSorensen JB, Lawlor BF, Geipel T, Pirouz P, Heuer AH, Lagerlof KPD. On basal slip and basal twinning in sapphire (alphaAl2O3). 1. Basal slip revisited. Acta Mater 1996;44:2145. [9] Tymiak NI, Daugela A, Wyrobek TJ, Warren OL. Acoustic emission monitoring of the earliest stages of contact-induced plasticity in sapphire. Acta Mater 2004;52:553. [10] Page TF, Oliver WC, McHargue CJ. The deformation-behavior of ceramic crystals subjected to very low load (nano)indentations. J Mater Res 1992;7:450. [11] Lloyd SJ, Molina-Aldareguia JM, Clegg WJ. Deformation under nanoindents in sapphire, spinel and magnesia examined using transmission electron microscopy. Philos Mag A 2002;82:1963. [12] Kollenberg W. Plastic-deformation of Al2O3 single-crystals by indentation at temperatures up to 750-degrees-C. J Mater Sci 1988;23:3321. [13] Poulsen HF, Nielsen SF, Lauridsen EM, Schmidt S, Suter RM, Lienert U, et al. Three-dimensional maps of grain boundaries and the stress state of individual grains in polycrystals and powders. J Appl Cryst 2001;34:751. [14] Ice GE, Larson BC. 3-D X-ray crystal microscope. Adv Eng Mater 2000;2:643. [15] Larson BC, Yang W, Ice GE, Budai JD, Tischler JZ. Threedimensional X-ray structural microscopy with submicrometre resolution. Nature 2002;415:887. [16] Tamura N, MacDowell AA, Spolenak R, Valek BC, Bravman JC, Brown WL, et al. Scanning X-ray microdiraction with submicrometer white beam for strain/stress and orientation mapping in thin lms. J Synchrotron Radiat 2003;10:137. [17] Loechelt GH, Cave NG, Menendez J. Polarized o-axis Raman spectroscopy: a technique for measuring stress tensors in semiconductors. J Appl Phys 1999;86:6164. [18] Dewolf I, Vanhellemont J, Romanorodriguez A, Norstrom H, Maes HE. Micro-Raman study of stress-distribution in local isolation structures and correlation with transmission electron-microscopy. J Appl Phys 1992;71:898. [19] De Wolf I, Jian C, van Spengen WM. The investigation of microsystems using Raman spectroscopy. Opt Lasers Eng 2001;36:213.

4665

[20] Bonera E, Fanciulli M, Batchelder DN. Combining high resolution and tensorial analysis in Raman stress measurements of silicon. J Appl Phys 2003;94:2729. [21] Bonera E, Fanciulli M, Batchelder DN. Raman spectroscopy for a micrometric and tensorial analysis of stress in silicon. Appl Phys Lett 2002;81:3377. [22] Porto SPS, Krishnan RS. Raman eect of corundum. J Chem Phys 1967;47:1009. [23] Watson GH, Daniels WB, Wang CS. Measurements Of Raman intensities and pressure-dependence of phonon frequencies in sapphire. J Appl Phys 1981;52:956. [24] Loudon R. Raman eect in crystals. Adv Phys 1964;13:423. [25] Shin SH, Pollak FH, Raccah PM. Eects of uniaxial stress on Raman frequencies Of Ti2O3 and Al2O3. Bull Am Phys Soc 1974;19:536. [26] Gallas MR, Chu YC, Piermarini GJ. Calibration of the Raman eect in alpha-Al2O3 ceramic for residual-stress measurements. J Mater Res 1995;10:2817. [27] Jia W, Yen WM. Raman-scattering from sapphire bers. J Raman Spectrosc 1989;20:785. [28] Shin FHP S, Raccah PM. Proceedings of the third international conference on light scattering in solids. In: Balanski RCCLaSPSP M, editor. Third international conference on light scattering in solids, 1976. [29] Ohtsuka S, Zhu W, Tochino S, Sekiguchi Y, Pezzotti G. In-depth analysis of residual stress in an alumina coating on silicon nitride substrate using confocal Raman piezo-spectroscopy. Acta Mater 2007;55:1129. [30] Colomban P, Havel M. Raman imaging of stress-induced phase transformation in transparent ZnSe ceramic and sapphire single crystals. J Raman Spectrosc 2002;33:789. [31] De Wolf I. Stress measurements in Si microelectronics devices using Raman spectroscopy. J Raman Spectrosc 1999;30:877. [32] Loechelt GH, Cave NG, Menendez J. Measuring the tensor nature of stress in silicon using polarized o-axis raman-spectroscopy. Appl Phys Lett 1995;66:3639. [33] Abramof PG, Ferreira NG, Beloto AF, Ueta AY. Investigation of nanostructured porous silicon by Raman spectroscopy and atomic force microscopy. J Non-Cryst Solids 2004;338:139.