Nanoscale characterization of strained silicon by tip ...

3 downloads 0 Views 246KB Size Report
1 shows the schematic of the -Si (Renesas Technology Corp.). The. -Si sample is prepared by forming germanium (Ge) doped. Si layer on a pure Si substrate.


Nanoscale characterization of strained silicon by tip-enhanced Raman spectroscope in reflection mode Y. Saito, M. Motohashi, and N. Hayazawaa兲 RIKEN (The Institute of Physical and Chemical Research), 2-1 Hirosawa, Wako-shi, Saitama 351-0198 Japan

M. Iyoki SII NanoTechnology Inc., 36-1, Takenoshita, Oyama-cho, Sunto-gun, Shizuoka 410-1393, Japan

S. Kawata RIKEN (The Institute of Physical Chemical Research), 2-1 Hirosawa, Wako-shi, Saitama 351-0198, Japan and Department of Applied Physics, Osaka University, 2-1 Yamadaoka, Suita, Osaka 565-0871, Japan

共Received 12 December 2005; accepted 18 February 2006; published online 4 April 2006兲 We observe localized strains in strained silicon by tip-enhanced near-field Raman spectroscope in reflection mode. The tip-enhanced Raman spectra show that the Raman frequency and intensity of strained silicon were different within a crosshatch pattern induced by lattice mismatch. Micro-Raman measurements, however, show only uniform features because of averaging effect due to the diffraction limit of light. Nanoscale characterization of strained silicon is essential for developing reliable next generation integrated circuits. This technique can be applicable not only to strained silicon but also to any other crystals. © 2006 American Institute of Physics. 关DOI: 10.1063/1.2191949兴 Near-field scanning optical microscopy 共NSOM兲 provides nanometer scale structural and chemical information, and development of apertureless NSOM, using a metallic probe, presents many advantages.1–6 For example, the potential for improved spatial resolution is tremendous because an apertureless probe may be considered as one or a few dipoles that interact locally with a sample surface. Moreover, the inclusion of a metallic probe tip in a system based on surface-enhanced Raman scattering7 共SERS兲 yields encouraging results as the probe can act as a surface enhancer. This concept has been recently recognized as tip-enhanced Raman spectroscopy 共TERS兲.8–15 SERS and TERS are expected to be promising tools for nanoscale characterization of surface of materials, e.g., analyzing stress in silicon.16 There are two configurations in TERS, transmission and reflection modes. In transmission mode,9–11,14 the metallic probe tip is illuminated through the sample using an objective lens. This configuration has been widely used in TERS because of its high efficiency in both illumination and collection especially when using high numerical aperture 共NA兲 objective lens. However, transmission mode TERS is not applicable to opaque or thick samples. In a reflection-mode TERS,15 the tip is illuminated directly from the same side of the probe tip, which is advantageous and promising for observing opaque samples. In this letter, we construct a reflection-mode TERS and successfully applied to nanoscale characterization of strained silicon 共␧-Si兲. This nondestructive characterization technique with a nanospatial resolution is essential for fabricating high performance electronic devices using ␧-Si substrates. The schematic of the system is shown in Fig. 1. An incident cw yttrium aluminum garnet laser 共YAG兲 laser 共532 nm, JDS Uniphase兲 is directed to the optical setup using a single mode optical fiber. From the fiber output, the beam

diameter of the laser is expanded 共⬃10 mm兲 while setting the polarization to p polarization, which is parallel to the tip axis. This polarized light is focused on the sample using a long-working-distance 共LWD兲 objective lens 共Mitsutoyo, NA= 0.28, ⫻20, WD= 30.5 mm兲. The apex of a cantilever 共Nanosensor ATEC-CONT兲 is adjusted onto the focused spot using the same principle as a contact-mode operation of atomic force microscope 共AFM兲. The power of the incident laser is set to deliver ⬃1 mW at the sample position. The same LWD objective lens collects the tip-enhanced Raman signal. A dichroic mirror directs the Raman signal output via a multimode optical fiber to a spectrometer 共Photon Design PPDT3-640兲. The spectrum is detected using a liquid-nitrogen-cooled charge coupled device 共CCD兲 camera 共Princeton Instruments兲. Raman signal was obtained with an integration time of 120 s. The inset in Fig. 1 shows the schematic of the ␧-Si 共Renesas Technology Corp.兲. The ␧-Si sample is prepared by forming germanium 共Ge兲 doped Si layer on a pure Si substrate. The concentration of Ge is gradually increased up to 25% to expand the lattice constant


FIG. 1. 共Color online兲 Experimental setup of reflection-mode TERS. The inset shows the structure of strained silicon on Si1−xGex 共x = 0.25兲.

Author to whom correspondence should be addressed; electronic mail: [email protected]

0003-6951/2006/88共14兲/143109/3/$23.00 88, 143109-1 © 2006 American Institute of Physics Downloaded 07 Apr 2006 to Redistribution subject to AIP license or copyright, see


Saito et al.

FIG. 2. 共Color online兲 Tip-enhanced near-field Raman spectra of ␧-Si. Red line and blue line correspond to near-field and far-field spectra, respectively. Black line is the background-corrected spectrum derived by the difference between near-field and far-field spectra. Dotted red lines are the Lorentzian curve fitting for the background-corrected spectrum.

of Si–Si. The 30-nm-thick strained silicon layer is epitaxially grown on the buffer SiGe layer 共thickness: 1 ␮m兲 in which the concentration is kept at 25%. To attain an efficient tip-enhancement effect in the reflection-mode TERS, the probe is coated with silver. Coated tips are prepared by thermal evaporation process. To prepare the tip 5-nm-thick gold palladium alloy 共AuPd兲 is deposited on a tip to modify the surface of a Si tip. Then, silver is evaporated with the thickness of 40 nm at a rate of 0.5 Å / s. Figure 2 shows the TERS spectra of ␧-Si. The near-field spectrum 共red line兲 in Fig. 2 is obtained when the silver tip is positioned close to the surface of the strained silicon. On the other hand, the far-field spectrum 共blue line兲 is obtained without the tip. We derive a background-corrected spectrum by taking the difference between the near-field and far-field spectra. The background-corrected spectrum contains nanometer scale information of localized strains. In the far-field spectrum, the ␧-Si peak is recognized only as a shoulder of the strong background signal generated from the underlying SiGe layer. The near-field spectrum, however, shows a clear and distinct peak and we owe this effect to tip enhancement. Since the capped ␧-Si layer is in the order of several tens of nanometers, a sufficient amount of incident light penetrates through the SiGe layer. The background signal from the underlying SiGe buffer layer is still significant. In order to clarify the band peaks, the background-corrected spectrum is decomposed into three Lorentzian functions 共dotted red lines兲 showing peaks at 503, 514, and 520 cm−1. These correspond to Si–Si stretch from the underlying SiGe layer, the ␧-Si layer, and the tip, respectively. The backgroundcorrected spectrum does not contain the unassigned peak at 481 cm−1, which can be considered as stray light from the optical fiber. The displacement of the Si–Si Raman peaks of ␧-Si from the inherent peak of Si 共520 cm−1兲 共Ref. 17兲 provides us with plenty of information such as local stress.17,18,20,21 The estimated enhancement factor for ␧-Si is

Appl. Phys. Lett. 88, 143109 共2006兲

FIG. 3. 共Color online兲 共a兲 Bright-field microscopic image of the strained silicon surface. 共b兲 AFM image of the ␧-Si surface. Red crosses indicate the tip positions where TERS mapping was carried out. 共c兲 and 共d兲 are Lorentzian curve fitting results of position dependent TERS spectra of ␧-Si and SiGe, respectively. a – f correspond to the positions shown in 共b兲.

2.2⫻ 104, calculated based on the size of the diffraction limited focused spot 共␾: 3 ␮m兲 and the enhanced electric field corresponding to the tip diameter 共␾: 40 nm兲. Lateral heterogeneity of strain in strained Si/ SiGe structures is a serious problem when fabricating high-mobility electronic devices assembled on ␧-Si substrates. Here, we demonstrate the position dependence of near-field spectra for detailed characterization. Figure 3共a兲 shows the bright-field microscopic image of the ␧-Si surface. The crosshatch patterns induced by lattice mismatch are seen on the epitaxial surface of the strained silicon. Figure 3共b兲 illustrates the contact AFM image 共3 ⫻ 3 ␮m2兲 of the crosshatch pattern. The six red crosses indicate positions where the tip-enhanced measurements were carried out. Figure 3共c兲 shows the background-corrected ␧-Si peak extracted by Lorentzian curve fitting, as shown in Fig. 2. Based on our previous work,19 each ␧-Si spectrum is curve fitted with a 4 cm−1 spectral width. From Fig. 3共c兲, one can observe that the position and the intensity of the ␧-Si peaks fluctuate. The fluctuation is due to variations in thickness of the ␧-Si layer based on an underlying surface topography. The pattern in the underlayer depends both on the Ge concentration of the relaxed material and on any residual, unrelaxed strain developed in the SiGe resulting from lattice-mismatched heteroepitaxial growth of SiGe on Si.20,21 The bright areas indicated in locations a, d, e, and f show a fairly intense Raman signal compared with the dark areas indicated in locations b and c. From the cross sectional view of Fig. 3共b兲, the step height between the bright and the dark area is ⬃16 nm. Figure 3共c兲 also illustrates the slight shift 共⬃2 cm−1兲 of the ␧-Si peaks. While the average deviation of the peak of the silicon tip stays within the range of 520± 0.04 cm−1 共data not shown兲, its value in the ␧-Si spectrum is 514± 0.8 cm−1. Raman shift of Si–Si vibration basically provides a measure of the interatomic Si–Si spacing in the ␧-Si layer. This spac-

Downloaded 07 Apr 2006 to Redistribution subject to AIP license or copyright, see


Appl. Phys. Lett. 88, 143109 共2006兲

Saito et al.

ing causes fluctuation of channel strains, and subsequently affects the performance of the device. It has been reported that lateral strain variations is mostly associated with the crosshatched pattern and its variation is in the order of ⬃1 cm−1.21 Calculated from the shift in Raman peaks, the strain imposed on the capped ␧-Si layer is about 共3 ± 0.4兲 ⫻ 109 Pa.20 Figure 3共d兲 shows the SiGe peak extracted by Lorentzian curve fitting. Si–Si vibration in the underlying SiGe layer did not show much spectral shift during the spectrum mapping except for a slight shift observed in spectrum d. This can be attributed to the local density of doped Ge, which is higher around position d. Accordingly, Si–Si in ␧-Si layer observed at d in Fig. 3共c兲 exhibits the highest spectral shift, and the nearby spectra obtained at c and e also have slight spectral shifts due to a relatively higher concentration of defects induced by the underlying heterogeneity of the concentration of Ge. This is the direct evidence of the fact that the Si–Si spacing in ␧-Si is affected by the Si–Si structures in the underlying SiGe layer. All these observed nanometer scale information on localized strain cannot be detected by microSERS experiments due to averaging effect within a diffraction limited focused spot.19 Using TERS has successfully revealed localized strain hidden in nanometer scale. However, the observed spectral shifts in the strained Si 共⬃2 cm−1兲 are higher than those in the SiGe layer 共⬃1 cm−1兲. One of the possible reasons could be due to the different contributions to the enhanced spectra shown in Figs. 3共c兲 and 3共d兲. In the case of the spectrum of SiGe, the enhanced electric field of excitation light localized at the tip cannot reach much of the underlying SiGe layer while the far-field signal generated from the diffraction limited focused spot is secondary enhanced by the tip since the penetration depth of the excitation light is much higher than the thickness of the ␧-Si layer.21 As a result, the near-field information of SiGe is mostly overwhelmed or averaged by the secondary-enhanced far-field signals in Fig. 3共d兲. This tendency was also observed with the different Ge concentrations and the thicknesses such as 20% with 17.5 nm thickness or 15% with 20 nm thickness. Further investigations on these observations are necessary. In previous works,22 varying the excitation from 351 to 514 nm spans the optical

penetration depth in ␧-Si from 5 to 300 nm. Reducing the excitation wavelength allows us to obtain background-free epilayer spectra leading to a feasible imaging of ␧-Si substrate. This work is now under development. The authors thank Dr. Vincent Daria of the University of the Philippines for the critical advice on this letter. The authors are also grateful to H. Naruoka and N. Hattori of Renesas Technology Corp. for the sample preparation and valuable comments. U. C. Fischer and D. W. Pohl, Phys. Rev. Lett. 62, 458 共1989兲. Y. Inouye and S. Kawata, Opt. Lett. 19, 159 共1994兲. 3 R. Bachelot, P. Gleyzes, and A. C. Boccara, Appl. Opt. 36, 2160 共1997兲. 4 B. Knoll and F. Keilmann, Nature 共London兲 399, 134 共1999兲. 5 M. S. Anderson and W. T. Pike, Rev. Sci. Instrum. 73, 1198 共2002兲. 6 D. Roy, S. H. Leong, and M. E. Welland, J. Korean Phys. Soc. 47, S140 共2005兲. 7 R. K. Chang and T. E. Furtak, Surface Enhanced Raman Scattering 共Plenum, New York, 1981兲. 8 R. M. Stöckle, Y. D. Suh, V. Deckert, and R. Zenobi, Chem. Phys. Lett. 318, 131 共2000兲. 9 N. Hayazawa, Y. Inouye, Z. Sekkat, and S. Kawata, Opt. Commun. 183, 333 共2000兲. 10 N. Hayazawa, Y. Inouye, Z. Sekkat, and S. Kawata, Chem. Phys. Lett. 335, 369 共2001兲. 11 D. S. Bulgarevich and M. Futamata, Appl. Spectrosc. 58, 757 共2004兲. 12 B. Pettinger, B. Ren, G. Picardi, R. Schuster, and G. Ertl, Phys. Rev. Lett. 92, 086101 共2004兲. 13 A. Hartschuh, E. J. Sánchez, X. S. Xie, and L. Novotny, Phys. Rev. Lett. 90, 095503 共2003兲. 14 T. Ichimura, N. Hayazawa, M. Hashimoto, Y. Inouye, and S. Kawata, Phys. Rev. Lett. 92, 220801 共2004兲. 15 D. Mehtani, N. Lee, R. D. Hartschuh, A. Kisliuk, M. D. Foster, A. P. Sokolov, and J. F. Maguire, J. Raman Spectrosc. 36, 1068 共2005兲. 16 S. Webster, D. N. Batchelder, and D. A. Smith, Appl. Phys. Lett. 72, 1478 共1998兲. 17 J. C. Tsang, P. M. Mooney, F. Dacol, and J. O. Chu, J. Appl. Phys. 75, 8098 共1994兲. 18 A. Atkinson and S. C. Jain, J. Raman Spectrosc. 30, 885 共1999兲. 19 N. Hayazawa, M. Motohashi, Y. Saito, and S. Kawata, Appl. Phys. Lett. 86, 263115 共2005兲. 20 J. McCarthy, S. Bhattacharya, T. S. Perova, R. A. Moore, H. Gamble, and B. M. Armstrong, Scanning 26, 235 共2004兲. 21 G. G. Goodman, V. Pajcini, S. P. Smith, and P. B. Merrill, Mater. Sci. Semicond. Process. 8, 225 共2005兲. 22 M. Holtz, W. M. Duncan, S. Zollner, and R. Liu, J. Appl. Phys. 88, 2523 共2000兲. 1 2

Downloaded 07 Apr 2006 to Redistribution subject to AIP license or copyright, see

Suggest Documents