- Email: [email protected]
Chemical functionalization of atomically flat cantilever surfaces
Microelectronic Engineering 86 (2009) 1200–1203
Contents lists available at ScienceDirect
Microelectronic Engineering journal homepage: www.elsevier.com/locate/mee
Chemical functionalization of atomically flat cantilever surfaces Valeria Toffoli a, Friederich Esch a, Mauro Melli a,b, Alessandro Pozzato a,c, Massimo Tormen a, Marco Lazzarino a,* a
CNR-INFM Laboratorio Nazionale TASC, Area Science Park, Basovizza, 34012 Trieste, Italy Scuola Internazionale di Studi Superiori Avanzati (SISSA/ISAS), Trieste, 34014, Italy c Università di Padova, Padova, Italy b
a r t i c l e
i n f o
Article history: Received 29 September 2008 Received in revised form 4 December 2008 Accepted 4 December 2008 Available online 24 December 2008 Keywords: MEMS Cantilever Chemical functionalization Atomically flat surface
a b s t r a c t In this paper, we present a novel approach toward the fabrication of mechanical oscillators with a local chemical functionalization with subnanometric control. Our method exploits the reactivity of the freshly cleaved surfaces that form when a monocrystalline silicon microstructure is cleaved. The surfaces that form after the cleave expose an atomically smooth Si(1 1 1) surface which, if a suitable chemical environment is provided, could undergo to a cycloaddition process that create stable, local and specific chemical bonds. Here, we demonstrated the feasibility of such a scheme on a twin cantilevers geometry, we prove the effective selectivity and stability of the cycloaddition process and we provide experimental evidence that below a critical size, the cleavage procedure creates step free atomically flat silicon surfaces. Ó 2008 Elsevier B.V. All rights reserved.
1. Introduction Mechanical oscillators represent a promising research approach to biomolecule manipulation and detection, capable of single molecule [1] and single atom detection [2]. In order to achieve the best molecular sensitivity, the location where molecules are adsorbed should be known with high precision and the sticking coefficient should be as close as possible to one. The usual experimental approach consists in calculating the mass as uniformly distributed on the mechanical oscillator [2]. However, some groups defined lithographically a gold probe area, which anchors the target molecules in a specific location to maximize the mechanical response. In systems with higher complexity, like two terminal devices such as twin cantilever systems [3,4], or break junction-like systems [5], the localization requirements are more stringent. Here, molecules should be immobilized across a gap that is few nanometres in width with a well defined orientation. So far, molecules have been attached randomly using a post-processing statistical approach to discard the results from molecule not correctly bound. In order to implement single molecule manipulation and detection, a method to functionalize two terminal nanodevices with two distinct functional groups would represent a big improvement.
* Corresponding author. Tel.: +39 040 3756434; fax: +39 040 226767. E-mail address: [email protected] (M. Lazzarino). 0167-9317/$ - see front matter Ó 2008 Elsevier B.V. All rights reserved. doi:10.1016/j.mee.2008.12.017
Here, we demonstrate this specific functionalization on two silicon surfaces that are separated by a gap, controllable with nanometric precision, through a cycloaddition reaction which exploits the reactivity of the freshly cleaved surfaces that form when the cantilever gap is created. Moreover, we analysed the surface quality of the cleaved areas and we found that below a threshold value, the cleaved surfaces are free from defects and are monoatomically flat, thus representing and ideal substrate for molecular anchor for single or few molecule experiments.
2. Fabrication In order to demonstrate our approach we adopted a twin cantilever structure similar to that described in Ref. [3]. Silicon on insulator (SOI) wafers composed of a 2.5 lm thick monocrystalline silicon layer with h1 1 0i or h1 0 0i surface orientation, 2 lm thick buried oxide layer and 500 lm thick handle silicon layer are the starting point. We deposit 50 nm thick nickel coating as a mask for the subsequent inductively coupled plasma (ICP) etching that produces vertical walls in silicon etching. A double clamped bridge geometry (500 lm long, and 20 lm wide with a central region smoothly narrowed to a notch with a width of 10 lm) is reproduced on the Ni coating by standard optical lithography and subsequent wet etching. Finally the oxide sacrificial layer is removed with a HF based wet etching and the sample is dried with a supercritical point dryer process in order to avoid the stiction collapse of the suspended structure [6]. At this stage the twin cantilevers are
V. Toffoli et al. / Microelectronic Engineering 86 (2009) 1200–1203
1201
in contact to each other. In order to open the gap for molecular experimentations and to tune it with nanometric precision, we apply a bending moment to the wafer as described in Refs. [3,7]. 3. Cleaved surface analysis
Fig. 1. Illustration of the formation of twin cantilever structures by mechanical cleaving a silicon bridge in a reactive droplet, by means of an AFM tip.
not yet formed and the structure has the shape of a continuous bridge. By applying a lateral mechanical stress at the notch with the help of a sharp tip such as that of an AFM (see schematic illustration in Fig. 1) the monocrystalline bridge breaks along a crystallographic plane generating two independent cantilevers. The two cleavage planes are, in principle, two defect free, atomically flat surfaces. Since silicon cleaves along the (1 1 1) surface, one obtains a cleavage plane tilted at 55 °C with respect to the surface when starting with h1 0 0i wafer while vertical cleavage planes are obtained when starting with h1 1 0i wafers. If no residual stress is present in the topmost silicon layer, the two cleaved surfaces stay
The surface quality of the two facing surfaces generating by the cleavage process is, in principle, much better than any other surface produced through a technological process, such as those based on plasma etching or focused ion beam lithography where edge roughness is typically in the range of several nanometres. Having two atomically flat surfaces facing each other separated by a gap of few nanometers represents an incredible opportunity for a large amount of chemical and physical experiments spanning from wear and friction in confined molecular layers [8] to Casimir forces in microsystems [9]. For this reason we investigated in detail the formation of the gap and the quality of the surface involved. In order to establish the actual absence of atomic steps on a cleaved surface, scanning probe microscopes offer the ideal resolution; unfortunately, in our case, the surfaces of interest are vertical with respect to the optical surface of the sample and therefore they are not accessible to this kind of analysis. Therefore we fabricated by ICP and electron beam lithography several pillar structures of various size on a h1 1 1i silicon wafer and then we cleaved them close to the substrate in order to investigate the surfaces by AFM and SEM and then compare their performances. In Fig. 2a we show an SEM image of a cleaved surface showing several cleavage steps. In Fig. 2b we show an AFM image of the area highlighted by the square in Fig. 2a. In Fig. 2c we display the height profile obtained along the line drawn in Fig. 2b. Several steps ranging from 0.8 nm to 2.7 nm are observed, most of which can be clearly visualized also on the SEM image. We can conclude that SEM analysis can be used to assess the quality of cleaved surface down to atomic scale. We fabricated some hundreds of bridges with three different notch size: 2.2 3.0 lm2; 2.2 2.2 lm2 and 1.5 1.5 lm2. We then cleaved the bridges with the procedure described in the fab-
Fig. 2. Comparison of SEM and AFM images of cleavage defects. (a) SEM image of a pillars 2 4 lm2 mechanically cleaved along the (1 1 1). Some steps are clearly resolved an otherwise flat area. Scale bar 1 lm. (b) AFM zoom on the 1 1 lm2 area highlighted by the square in panel (a). Several flat terraces are observed. (c) Height profile along the green arrow in panel (b). All the steps are smaller than 3 nm.
1202
V. Toffoli et al. / Microelectronic Engineering 86 (2009) 1200–1203
rication section. We removed one of the resulting cantilevers in order to facilitate the subsequent SEM imaging. We obtained three different typologies of surfaces illustrated in Fig. 3. In Fig. 3a we show a 2.2 2.2 lm2 cleavage surface in which no defects are visible. In Fig. 3b we show an example of a surface with a bunch of steps indicated by the arrows. Finally in Fig. 3c we show a surface with a defected cleavage in which one side of the surface presents a large protrusion, indicated by the arrow. The latter geometry hinders the mutual movements of the two faces and represents a failure of a twin cantilever structure. It is worth to stress here the
Fig. 4. Bright field (a) and fluorescence (b) optical images of a cantilever functionalized with FITC. The opposite cantilever has been mechanically removed for facilitate the imaging. On the fluorescence image (b) the cantilever contour has been reported.
amazing difference between the surface roughness of the ICPetched walls and the cleaved surfaces. Fig. 3d displays the statistical analysis for the defect free surfaces (open symbols – red line) and the failed surfaces (full symbols – black line) plotted versus the surface area. At 2.2 lm2, 50% of the surfaces are free of defects, and following the trend sketched in Fig. 3d, a further reduction of the surface area would provide a majority of defect-free surfaces, opening the way for a possible engineering of this fabrication process. Finally it is worth to stress here that so far in literature the defect free cleavage procedures refers to solids whose area was in the few nm2 range [10]. 4. Functionalization process
Fig. 3. SEM images of cleaved cantilevers with section 2.2 2.2 lm2 showing different surface quality. All scale bars are 1 lm. (a) Defect free surface. (b) Surface with a shallow cleavage step indicated by the arrows. (c) Surface with a micrometer sized protrusion indicated by an arrow. (d) Plot of the percentage of defect free (open circles) and damaged (solid circles) cleavage as a function of cleavage area.
In order to exploit the reactivity of the freshly cleaved silicon surface we performed the cleavage procedure in a highly reactive ambient using fluorescein isothiocyanate (FITC) in a fashion with reminds for some aspects the chemomechanical patterning developed by the group of Matthew Lindsay [11]. FITC consists of a green emitting fluorescent head and of a C@N@S reactive tail, which is know to undergo a cycloaddition process when brought in contact with clean unsaturated silicon surfaces [12]. FITC is soluble in water and other polar solvents, which are in the majority also good oxidizing agents, and are therefore competing with FITC for the reaction with the freshly cleaved silicon surface. Therefore we dissolved the FITC in dimethyl sulfoxide (0.13 M), directly in a sealed chamber in which the sample and the cleaving mechanism were prealigned, in order to reduce the oxygen contamination of the solution and subsequent oxidation of the cleaved surfaces. After cleaving, the samples have been rinsed in abundant methanol and water in order to remove all the FITC that was non-specifically
V. Toffoli et al. / Microelectronic Engineering 86 (2009) 1200–1203
absorbed on the surface. The samples have been investigated by fluorescence optical microscopy. In Fig. 4a we display a representative bright field optical image of a cantilever after cleaving, and in Fig. 4b the fluorescent counterpart on which the edges of the cantilever have been artificially superimposed. We observe the presence of a two fluorescence spots in correspondence of the cleaved surface while the rest of the cantilever and the substrate surfaces do not show any fluorescence. This result, although not as uniform as expected, indicates that the functionalization process has been effective and that we bound our target molecules only on the cleaved areas while the non specific absorption has been minimized. A further optimization of the process is required in order to achieve a complete and uniform functionalization over the whole cleaved area. 5. Conclusions We demonstrated that our approach provides a new route to selective functionalization of silicon-based microelectromechanical devices, with high chemical sensitivity. The surfaces which are functionalized are flat on the atomic scale. Moreover, due the peculiar process, the localization of the functionalized regions on the structures of interest can be realized with subnanometer accuracy.
1203
Acknowledgements We thank Giacinto Scoles for valuable discussions. The partial support by the EC-funded project NaPa (Contract No. NMP4-CT2003-500120) is gratefully acknowledged.
References [1] B. Ilic, Y. Yang, K. Aubin, R. Reichenbach, S. Krylov, H.G. Craighead, Nanoletters 5 (2005) 925. [2] K. Jensen, K. Kwanpyo, A. Zettl, Nat. Nanotechnol. 3 (2008) 533. [3] M. Lazzarino, E. de Marchi, M. Bressanutti, L. Vaccari, S. Cabrini, C. Schmid, R. Poetes, G. Scoles, Microelectron. Eng. 83 (2006) 1309. [4] A. Qazi, D. Nonis, A. Pozzato, M. Tormen, M. Lazzarino, S. Carrato, G. Scoles, Appl. Phys. Lett. 92 (2007) 173118. [5] L. Venkataraman, J.E. Klare, C. Nuckolls, M.S. Hybertsen, M.L. Steigerwald, Nature 442 (2006) 904. [6] A. Show, I. Jafri, H. Busta, S. Walsh Critical, Proc. SPIE 3880 (1999) 51. [7] V. Toffoli, F. Esch, M. Melli, F. Cataruzza, A. Pozzato, S. Carrato, G. Scoles, M. Tormen, M. Lazzarino, Nanotechnology 19 (2008) 445502. [8] U. Raviv, J. Klein, Science 297 (2002) 1540. [9] H.B. Chan, V.A. Aksyuk, R.N. Kleiman, D.J. Bishop, Federico Capasso, Science 291 (2001) 1941. [10] A. Wei-Qiang Han, A. Zettl, Appl. Phys. Lett. 84 (2004) 2644. [11] Yang Li, Y.Y. Lua, M.R. Linford, Acc. Chem. Res. 38 (2005) 933. [12] M.D. Ellison, R.J. Hamers, J. Phys. Chem B 103 (1999) 6243.
Contents lists available at ScienceDirect
Microelectronic Engineering journal homepage: www.elsevier.com/locate/mee
Chemical functionalization of atomically flat cantilever surfaces Valeria Toffoli a, Friederich Esch a, Mauro Melli a,b, Alessandro Pozzato a,c, Massimo Tormen a, Marco Lazzarino a,* a
CNR-INFM Laboratorio Nazionale TASC, Area Science Park, Basovizza, 34012 Trieste, Italy Scuola Internazionale di Studi Superiori Avanzati (SISSA/ISAS), Trieste, 34014, Italy c Università di Padova, Padova, Italy b
a r t i c l e
i n f o
Article history: Received 29 September 2008 Received in revised form 4 December 2008 Accepted 4 December 2008 Available online 24 December 2008 Keywords: MEMS Cantilever Chemical functionalization Atomically flat surface
a b s t r a c t In this paper, we present a novel approach toward the fabrication of mechanical oscillators with a local chemical functionalization with subnanometric control. Our method exploits the reactivity of the freshly cleaved surfaces that form when a monocrystalline silicon microstructure is cleaved. The surfaces that form after the cleave expose an atomically smooth Si(1 1 1) surface which, if a suitable chemical environment is provided, could undergo to a cycloaddition process that create stable, local and specific chemical bonds. Here, we demonstrated the feasibility of such a scheme on a twin cantilevers geometry, we prove the effective selectivity and stability of the cycloaddition process and we provide experimental evidence that below a critical size, the cleavage procedure creates step free atomically flat silicon surfaces. Ó 2008 Elsevier B.V. All rights reserved.
1. Introduction Mechanical oscillators represent a promising research approach to biomolecule manipulation and detection, capable of single molecule [1] and single atom detection [2]. In order to achieve the best molecular sensitivity, the location where molecules are adsorbed should be known with high precision and the sticking coefficient should be as close as possible to one. The usual experimental approach consists in calculating the mass as uniformly distributed on the mechanical oscillator [2]. However, some groups defined lithographically a gold probe area, which anchors the target molecules in a specific location to maximize the mechanical response. In systems with higher complexity, like two terminal devices such as twin cantilever systems [3,4], or break junction-like systems [5], the localization requirements are more stringent. Here, molecules should be immobilized across a gap that is few nanometres in width with a well defined orientation. So far, molecules have been attached randomly using a post-processing statistical approach to discard the results from molecule not correctly bound. In order to implement single molecule manipulation and detection, a method to functionalize two terminal nanodevices with two distinct functional groups would represent a big improvement.
* Corresponding author. Tel.: +39 040 3756434; fax: +39 040 226767. E-mail address: [email protected] (M. Lazzarino). 0167-9317/$ - see front matter Ó 2008 Elsevier B.V. All rights reserved. doi:10.1016/j.mee.2008.12.017
Here, we demonstrate this specific functionalization on two silicon surfaces that are separated by a gap, controllable with nanometric precision, through a cycloaddition reaction which exploits the reactivity of the freshly cleaved surfaces that form when the cantilever gap is created. Moreover, we analysed the surface quality of the cleaved areas and we found that below a threshold value, the cleaved surfaces are free from defects and are monoatomically flat, thus representing and ideal substrate for molecular anchor for single or few molecule experiments.
2. Fabrication In order to demonstrate our approach we adopted a twin cantilever structure similar to that described in Ref. [3]. Silicon on insulator (SOI) wafers composed of a 2.5 lm thick monocrystalline silicon layer with h1 1 0i or h1 0 0i surface orientation, 2 lm thick buried oxide layer and 500 lm thick handle silicon layer are the starting point. We deposit 50 nm thick nickel coating as a mask for the subsequent inductively coupled plasma (ICP) etching that produces vertical walls in silicon etching. A double clamped bridge geometry (500 lm long, and 20 lm wide with a central region smoothly narrowed to a notch with a width of 10 lm) is reproduced on the Ni coating by standard optical lithography and subsequent wet etching. Finally the oxide sacrificial layer is removed with a HF based wet etching and the sample is dried with a supercritical point dryer process in order to avoid the stiction collapse of the suspended structure [6]. At this stage the twin cantilevers are
V. Toffoli et al. / Microelectronic Engineering 86 (2009) 1200–1203
1201
in contact to each other. In order to open the gap for molecular experimentations and to tune it with nanometric precision, we apply a bending moment to the wafer as described in Refs. [3,7]. 3. Cleaved surface analysis
Fig. 1. Illustration of the formation of twin cantilever structures by mechanical cleaving a silicon bridge in a reactive droplet, by means of an AFM tip.
not yet formed and the structure has the shape of a continuous bridge. By applying a lateral mechanical stress at the notch with the help of a sharp tip such as that of an AFM (see schematic illustration in Fig. 1) the monocrystalline bridge breaks along a crystallographic plane generating two independent cantilevers. The two cleavage planes are, in principle, two defect free, atomically flat surfaces. Since silicon cleaves along the (1 1 1) surface, one obtains a cleavage plane tilted at 55 °C with respect to the surface when starting with h1 0 0i wafer while vertical cleavage planes are obtained when starting with h1 1 0i wafers. If no residual stress is present in the topmost silicon layer, the two cleaved surfaces stay
The surface quality of the two facing surfaces generating by the cleavage process is, in principle, much better than any other surface produced through a technological process, such as those based on plasma etching or focused ion beam lithography where edge roughness is typically in the range of several nanometres. Having two atomically flat surfaces facing each other separated by a gap of few nanometers represents an incredible opportunity for a large amount of chemical and physical experiments spanning from wear and friction in confined molecular layers [8] to Casimir forces in microsystems [9]. For this reason we investigated in detail the formation of the gap and the quality of the surface involved. In order to establish the actual absence of atomic steps on a cleaved surface, scanning probe microscopes offer the ideal resolution; unfortunately, in our case, the surfaces of interest are vertical with respect to the optical surface of the sample and therefore they are not accessible to this kind of analysis. Therefore we fabricated by ICP and electron beam lithography several pillar structures of various size on a h1 1 1i silicon wafer and then we cleaved them close to the substrate in order to investigate the surfaces by AFM and SEM and then compare their performances. In Fig. 2a we show an SEM image of a cleaved surface showing several cleavage steps. In Fig. 2b we show an AFM image of the area highlighted by the square in Fig. 2a. In Fig. 2c we display the height profile obtained along the line drawn in Fig. 2b. Several steps ranging from 0.8 nm to 2.7 nm are observed, most of which can be clearly visualized also on the SEM image. We can conclude that SEM analysis can be used to assess the quality of cleaved surface down to atomic scale. We fabricated some hundreds of bridges with three different notch size: 2.2 3.0 lm2; 2.2 2.2 lm2 and 1.5 1.5 lm2. We then cleaved the bridges with the procedure described in the fab-
Fig. 2. Comparison of SEM and AFM images of cleavage defects. (a) SEM image of a pillars 2 4 lm2 mechanically cleaved along the (1 1 1). Some steps are clearly resolved an otherwise flat area. Scale bar 1 lm. (b) AFM zoom on the 1 1 lm2 area highlighted by the square in panel (a). Several flat terraces are observed. (c) Height profile along the green arrow in panel (b). All the steps are smaller than 3 nm.
1202
V. Toffoli et al. / Microelectronic Engineering 86 (2009) 1200–1203
rication section. We removed one of the resulting cantilevers in order to facilitate the subsequent SEM imaging. We obtained three different typologies of surfaces illustrated in Fig. 3. In Fig. 3a we show a 2.2 2.2 lm2 cleavage surface in which no defects are visible. In Fig. 3b we show an example of a surface with a bunch of steps indicated by the arrows. Finally in Fig. 3c we show a surface with a defected cleavage in which one side of the surface presents a large protrusion, indicated by the arrow. The latter geometry hinders the mutual movements of the two faces and represents a failure of a twin cantilever structure. It is worth to stress here the
Fig. 4. Bright field (a) and fluorescence (b) optical images of a cantilever functionalized with FITC. The opposite cantilever has been mechanically removed for facilitate the imaging. On the fluorescence image (b) the cantilever contour has been reported.
amazing difference between the surface roughness of the ICPetched walls and the cleaved surfaces. Fig. 3d displays the statistical analysis for the defect free surfaces (open symbols – red line) and the failed surfaces (full symbols – black line) plotted versus the surface area. At 2.2 lm2, 50% of the surfaces are free of defects, and following the trend sketched in Fig. 3d, a further reduction of the surface area would provide a majority of defect-free surfaces, opening the way for a possible engineering of this fabrication process. Finally it is worth to stress here that so far in literature the defect free cleavage procedures refers to solids whose area was in the few nm2 range [10]. 4. Functionalization process
Fig. 3. SEM images of cleaved cantilevers with section 2.2 2.2 lm2 showing different surface quality. All scale bars are 1 lm. (a) Defect free surface. (b) Surface with a shallow cleavage step indicated by the arrows. (c) Surface with a micrometer sized protrusion indicated by an arrow. (d) Plot of the percentage of defect free (open circles) and damaged (solid circles) cleavage as a function of cleavage area.
In order to exploit the reactivity of the freshly cleaved silicon surface we performed the cleavage procedure in a highly reactive ambient using fluorescein isothiocyanate (FITC) in a fashion with reminds for some aspects the chemomechanical patterning developed by the group of Matthew Lindsay [11]. FITC consists of a green emitting fluorescent head and of a C@N@S reactive tail, which is know to undergo a cycloaddition process when brought in contact with clean unsaturated silicon surfaces [12]. FITC is soluble in water and other polar solvents, which are in the majority also good oxidizing agents, and are therefore competing with FITC for the reaction with the freshly cleaved silicon surface. Therefore we dissolved the FITC in dimethyl sulfoxide (0.13 M), directly in a sealed chamber in which the sample and the cleaving mechanism were prealigned, in order to reduce the oxygen contamination of the solution and subsequent oxidation of the cleaved surfaces. After cleaving, the samples have been rinsed in abundant methanol and water in order to remove all the FITC that was non-specifically
V. Toffoli et al. / Microelectronic Engineering 86 (2009) 1200–1203
absorbed on the surface. The samples have been investigated by fluorescence optical microscopy. In Fig. 4a we display a representative bright field optical image of a cantilever after cleaving, and in Fig. 4b the fluorescent counterpart on which the edges of the cantilever have been artificially superimposed. We observe the presence of a two fluorescence spots in correspondence of the cleaved surface while the rest of the cantilever and the substrate surfaces do not show any fluorescence. This result, although not as uniform as expected, indicates that the functionalization process has been effective and that we bound our target molecules only on the cleaved areas while the non specific absorption has been minimized. A further optimization of the process is required in order to achieve a complete and uniform functionalization over the whole cleaved area. 5. Conclusions We demonstrated that our approach provides a new route to selective functionalization of silicon-based microelectromechanical devices, with high chemical sensitivity. The surfaces which are functionalized are flat on the atomic scale. Moreover, due the peculiar process, the localization of the functionalized regions on the structures of interest can be realized with subnanometer accuracy.
1203
Acknowledgements We thank Giacinto Scoles for valuable discussions. The partial support by the EC-funded project NaPa (Contract No. NMP4-CT2003-500120) is gratefully acknowledged.
References [1] B. Ilic, Y. Yang, K. Aubin, R. Reichenbach, S. Krylov, H.G. Craighead, Nanoletters 5 (2005) 925. [2] K. Jensen, K. Kwanpyo, A. Zettl, Nat. Nanotechnol. 3 (2008) 533. [3] M. Lazzarino, E. de Marchi, M. Bressanutti, L. Vaccari, S. Cabrini, C. Schmid, R. Poetes, G. Scoles, Microelectron. Eng. 83 (2006) 1309. [4] A. Qazi, D. Nonis, A. Pozzato, M. Tormen, M. Lazzarino, S. Carrato, G. Scoles, Appl. Phys. Lett. 92 (2007) 173118. [5] L. Venkataraman, J.E. Klare, C. Nuckolls, M.S. Hybertsen, M.L. Steigerwald, Nature 442 (2006) 904. [6] A. Show, I. Jafri, H. Busta, S. Walsh Critical, Proc. SPIE 3880 (1999) 51. [7] V. Toffoli, F. Esch, M. Melli, F. Cataruzza, A. Pozzato, S. Carrato, G. Scoles, M. Tormen, M. Lazzarino, Nanotechnology 19 (2008) 445502. [8] U. Raviv, J. Klein, Science 297 (2002) 1540. [9] H.B. Chan, V.A. Aksyuk, R.N. Kleiman, D.J. Bishop, Federico Capasso, Science 291 (2001) 1941. [10] A. Wei-Qiang Han, A. Zettl, Appl. Phys. Lett. 84 (2004) 2644. [11] Yang Li, Y.Y. Lua, M.R. Linford, Acc. Chem. Res. 38 (2005) 933. [12] M.D. Ellison, R.J. Hamers, J. Phys. Chem B 103 (1999) 6243.