Scannable Plasmonic Trapping Using a Gold Stripe - Harvard University

Aug 17, 2010 - Mirrors M1,2 and lens L1,2 sit on rotation and ... The dimensions of the elliptical spot refer to ..... tion with a standard deviation ...
0 downloads 0 Views 2MB Size
pubs.acs.org/NanoLett

Scannable Plasmonic Trapping Using a Gold Stripe Kai Wang,* Ethan Schonbrun, Paul Steinvurzel, and Kenneth B. Crozier* School of Engineering and Applied Sciences, Harvard University, Cambridge, Massachusetts 02138 ABSTRACT Using counterpropagating surface plasmon polaritons (SPPs) on a gold stripe, we demonstrate a scannable integrated optical tweezer. We demonstrate the trapping of individual fluorescent beads on the stripe, which supports a single quasi-transverse magnetic (TM) mode at the metal-water interface. The beads are localized to the stripe center, with a standard deviation of 51 nm transverse to the stripe, corresponding to a trap stiffness of 1.7 pN/µm. The localization along the stripe is achieved by balancing the scattering forces from the two counterpropagating SPPs excited by prism coupling. The particle position along the stripe can be controlled by varying the relative intensity of the two input beams. This work adds an important new capability to plasmonic optical tweezers, that of scanning. We anticipate that this will broaden the range of applications of plasmonic optical manipulation. KEYWORDS surface plasmon polaritons, plasmonic waveguide, optical trapping, microfluidic device

S

ince the pioneering work of Arthur Ashkin,1 the field of optical trapping has grown tremendously due to its successful implementation in the life and physical sciences, with applications ranging from force spectroscopy in single molecule studies to optical sorting in microfluidic systems2-5 This success largely lies in the noninvasive nature of optical forces, which can be exerted on micro- and nanoscale particles to trap, manipulate, and measure forces accurately without disturbing the environment. While the well-established technique of far field optical trapping, usually called optical tweezers, is continually being refined,6,7 there is a growing trend toward using trapping and manipulation using optical near fields.8-16 This is due to the flexibility that nanostructures afford for the realization of favorable field distributions and their small footprints, potentially enabling massively parallel integration into lab-on-a-chip devices. Making use of the field enhancement near high Q spherical and photonic crystal resonators,10,11 optical trapping has been observed with reduced laser power. The enhancement of gradient force due to subwavelength field confinement near nanoapertures has also been proposed and demonstrated.12,13 Moreover, plasmon resonance structures, which can provide both strong field enhancement and field confinement,14,15 are gaining increasing attention for the optical trapping of nanoparticles and biological materials.16-20 The trapping of micrometer sized dielectric particles and cells has been demonstrated with localized surface plasmon (LSP) resonance structures at illumination intensities much smaller than those needed for far-field trapping.16,17 More recently, enhanced optical forces on metallic nanoparticles from surface plasmon polaritons have been observed.18,19

The use of plasmonic trapping is motivated largely by its potential for considerably enhanced optical forces. Absorption-induced thermal convection from plasmonic structures, however, usually limits the maximum optical power that can be used, thereby limiting the magnitude of the optical force or trapping stiffness that can be achieved.16,17 Rather than exciting SPPs on continuous gold films, we excite surface plasmon polaritons (SPPs) on gold stripes to trap micrometersized polystyrene particles. By reduction in the amount of metal, absorption, and therefore heating, is minimized. Here, we demonstrate plasmonic trapping using a thin micrometer scale gold stripe. Particles are tightly localized to the stripe with a large transverse trapping stiffness of 1.7 pN/µm and an estimated trapping potential of 88 kBT (T ) 300 K). Moreover, with this scheme, trapped particles can be scanned over large distances along the gold stripe. This flexibility may be useful for manipulation and sorting applications. As illustrated in Figure 1a, the plasmonic trap consists of a gold stripe sitting on a glass slide in a microfluidic channel. The stripe is fabricated by electron beam lithography and the lift-off process. The gold stripes are 40 nm thick and have widths of 1 and 2 µm. A PDMS microfluidic channel is sealed onto the glass slide such that the gold stripes are perpendicular to the channel axis. The channel is made only ∼5 µm deep to suppress thermal convection.20 Diluted solutions of 1 or 2 µm diameter polystyrene particles (Invitrogen, Inc.) are loaded into the microfluidic channel and flow to the trapping site on the gold stripe. Experiments are performed with a Kretschmann prism-coupling geometry, as shown in Figure 1b. The optical setup is shown in Figure 1c. The laser source is a fiber pigtailed semiconductor diode laser (λ0 ) 980 nm). The beam output from the fiber is collimated and split into two by a polarizing cube beamsplitter. The halfwave plate, which controls the polarization of the beam entering the beamsplitter, is used to adjust the splitting ratio

* Corresponding authors: Kenneth B. Crozier [email protected] and Kai Wang [email protected]. Received for review: 05/10/2010 Published on Web: 08/17/2010 © 2010 American Chemical Society

3506

DOI: 10.1021/nl101653n | Nano Lett. 2010, 10, 3506–3511

FIGURE 2. (a) Time sequence of scattered light images of particles in the microfluidic channel, as recorded by the CCD camera. (b) Experimentally measured propulsion velocity of polystyrene particles at different incident angles when illumination intensity is 40 µW/µm2. Key: W, gold stripe width; D, particle diameter.

FIGURE 1. (a) Schematic of plasmonic trapping device, consisting of a gold stripe in a microfluidic channel formed on a microscope glass slide. (b) Schematic of prism coupling of counterpropagating SPPs on gold stripes. (c) Schematic of optical setup: LP filter, long pass filter; SP filter, short pass filter.

angles considered in our investigation (62-65°), as predicted by Gaussian beam theory and observed experimentally. When the input power is 100 mW, at which we perform most measurements, the average illumination field intensity over the central portion of the focused spot, in which the intensity is no smaller than 1/e2 times the peak intensity, is ∼40 µW/µm2. The gold stripe is perpendicular to the microfluidic channel. Because the laminar fluid flow in the channel is purely along the length of the channel, the particles’ movement along the gold stripe is purely due to optical forces. The flow rate in the microfluidic channel is kept low (∼10 µm/s) to increase the probability of transverse trapping by the gold stripes. As the first experiment, we launch only one laser beam into the prism, in order to propel particles using the SPPs on the gold stripe. Particles scatter light when they come into the evanescent field generated by total internal reflection (TIR) at the glass-water interface, away from the gold stripe. Stronger scattering of light by particles is observed when they get close to the gold stripe, as shown in Figure 2a (also see movie 1 in Supporting Information). The particles, comprising 1 µm diameter polystyrene beads, are then trapped transversely and propelled along the gold stripe. The measured velocities of particles propelled by optical forces, from TIR at the glass-water interface and by SPPs excited on the gold stripes, are plotted in Figure 2b as a function of angle of incidence. Data are presented for stripes with widths of 1 and 2 µm and for polystyrene spheres with diameters

of the two output beams. A second half-wave plate is placed after the beamsplitter, to rotate the polarization of one of the beams, so that both beams in the prism are incident on the stripe with TM polarization. The two TM-polarized beams are loosely focused onto the prism with the same incident angle and focusing conditions, meaning that the spot sizes are identical. Mirrors M1,2 and lens L1,2 sit on rotation and translation stages to allow accurate control of incident angles and focusing conditions. The slide on which the gold stripes and microfluidic channel are formed is placed on top of the prism with immersion oil in between. When the incident angle is greater than the critical angle, counterpropagating evanescent waves result on the glass surface inside the microfluidic channel. A microscope objective (50×, NA ) 0.55) is used to form images of the particles on a charge coupled device (CCD) camera. The images formed are either darkfield images of the light scattered by the particles from the evanescent field or fluorescence images. The fluorescence from the particles is excited by a green laser (λ0 ) 532 nm). To couple the power into surface plasmon polaritons on the gold stripe efficiently, we limit the divergence of the incident beam to ∼0.6°. This results in an elliptical spot incident on the glass-water interface with dimensions 40 µm × 80 µm. The dimensions of the elliptical spot refer to the extent of the 1/e2 intensity contour over the range of © 2010 American Chemical Society

3507

DOI: 10.1021/nl101653n | Nano Lett. 2010, 10, 3506-–3511

of 1 and 2 µm. All experiments are performed with an average illumination field intensity of 40 µW/µm2. The critical angle for TIR at the glass-water interface is 61.7°. When there is no gold stripe (i.e., TIR), the propulsion velocity of particles decreases rapidly with increasing incident angle (black curve in Figure 2b). This has been observed previously,21 and is mainly due to reduction of the evanescent field depth. The largest propulsion velocity is observed when incident angle is below the critical angle, meaning that the fields in the water are propagating, rather than evanescent. In this case, there is no optical force that draws particles onto the glass water interface, so stable optical trapping is not achieved. In assessing the performance of particle propulsion using the TIR and surface plasmon approaches, therefore, it is the evanescent field regime, when the incident angle is above critical angle, that should be considered. From Figure 2b, it can be seen that in this regime, for the same angle of incidence, the velocities are higher for particles propelled by SPPs on the gold stripes than TIR at the glass-water interface. The excitation of SPPs on the gold stripes results in enhanced fields, which in turn increases the optical force on the particles. This leads to the enhanced propulsion velocities. Interestingly, the width of the gold stripe produces a marked change in the dependence of propulsion velocity with incident angle. Gold stripes with widths of 1 µm can trap polystyrene particles for incident angles between 62° and 63°, with the velocity peaking at 62°. Over this range, the propulsion velocity is enhanced by on average ∼2 times, compared to the TIR case. Gold stripes with widths of 2 µm can trap polystyrene particles for incident angles in the range 62.5-63°, with a highest propulsion velocity enhancement factor of ∼5 times at 63.5°. For gold stripes wider than 2 µm, strong thermal convection is observed and we get very unstable trapping results. We have previously demonstrated the propulsion of gold nanoparticles with SPPs on an unpatterned gold film.18 The propulsion of dielectric particles, however, requires higher laser powers due to their smaller polarizability. The larger amount of metal present in the wider stripes results in the strong thermal convection we observe. We also tried narrower gold stripes with widths of 600 and 800 nm in experiments, but no transverse trapping and enhanced propulsion was observed. To gain insight into these phenomena, we employ the finite difference time domain (FDTD) method to calculate the H field intensity enhancement (|Hsurface|2/|Hin|2) on the gold stripes as a function of incident angle of the plane wave illumination. The dielectric constant of gold at λ0 ) 980 nm is taken as -40.1 + 2.73i in the calculations.22 As shown in Figure 3, the field enhancement peaks at different incident angles for the 1 and 2 µm wide gold stripes. This is due to the dispersive characteristic of the SPPs modes being excited. Since we use a Kretschmann prism-coupling configuration, we excite leaky modes with an effective refractive index neff satisfying nwater < neff < nglass.23 Using the finite © 2010 American Chemical Society

FIGURE 3. Calculated H field intensity enhancement (|Hsurface|2/|Hin|2) on glass surface for TIR at the glass-water interface and on the gold stripe surface for SPP excitation. Widths (W) of gold stripes are 1 and 2 µm. Results for an unpatterned gold film are also shown.

FIGURE 4. (a). Electric field component |E|2 of first-order quasi-TM mode on gold stripe (1 µm wide and 40 nm thick). (b) Real part of the effective refractive index (left axis) and corresponding excitation angle of incidence (AOI, right axis) of quasi-TM modes on gold stripe calculated by FEM.

element method (FEM), we calculate the mode profile and effective refractive index, as shown in Figure 4. These modes are termed quasi-TM modes because they are similar, but not identical, to the TM modes on unpatterned metal films. As the width of the gold stripe decreases, the number of the modes supported by the gold stripe decreases. When the width is smaller than ∼1 µm, cut off is reached, meaning that no quasi-TM modes are supported. It should be noted, however, that the stripe still supports other modes.24 For 1 and 2 µm wide gold stripes used in our experiments, only the first-order quasi-TM mode is supported and can be excited by prism coupling. According to the effective refractive index plotted in Figure 4b, the k matching conditions are satisfied when incident angles are 61.7° and 63.25° for the 1 and 2 µm wide stripes, respectively, which confirms 3508

DOI: 10.1021/nl101653n | Nano Lett. 2010, 10, 3506-–3511

that the field enhancement found in FDTD simulations is due to effective excitation of the first-order quasi-TM mode. The variation in the angle of incidence at which the field enhancement peaks with stripe width, as shown in Figure 3, is consistent with the experimental results of Figure 2. From this figure, it is clear that the measured propulsion velocity peaks at different angles for the 1 and 2 µm wide stripes. Additionally, the experimental observation that no trapping can be obtained with gold stripes narrower than 1 µm can be reasonably explained by the cut off of the quasi-TM modes. From Figure 3, it can be seen that the field enhancement of the 2 µm wide gold stripe is not superior to that of the 1 µm wide stripe. On the other hand, 2 µm wide gold stripes produce more heating. In the following experiments, therefore, we employ 1 µm wide gold stripes. An incident angle of 62° is used, as this angle gives the highest propulsion velocity in the experiments of Figure 2 for the 1 µm wide stripe. Trapping and manipulation are achieved when two counterpropagating beams are launched into the prism at the same time. When a small displacement is introduced between two focused spots of these two beams, which leads to a displacement between +z and -z propagating SPPs’ maximum intensity positions, individual particles can be trapped in the longitudinal direction along the gold stripe due to the scattering forces from the two counterpropagating SPPs being balanced, as shown in Figure 5a. Furthermore, Figure 5a also illustrates that displacements of the particles from the trap center will result in a restoring force toward the center. This would not be the case, were the beams not displaced from one another along the axis of the stripe (z direction). When the power of these two beams is adjusted, the equilibrium position, at which the scattering forces from the two beams cancel and the particle is trapped, can be changed, as shown in Figure 5b. This would again not be possible, were the beams not displaced from one another. In this way, the particle can be translated freely along the gold stripe. A typical trajectory of a 1 µm diameter polystyrene particle being trapped and scanned along the 1 µm wide gold stripe is shown in Figure 5c (see also movie 2 in Supporting Information). In this experiment, the laser diode power is not changed, but the powers of the two beams incident on the prism are modulated by manually rotating the wave plate before the polarizing beam splitter. The particle is initially trapped on the left side of the stripe (z ∼ -10 µm) by adjusting the powers of the laser beams so that the scattering forces cancel. At t ∼ 4 and t ∼12 s, the left and right laser beams’ power are increased and decreased, respectively, resulting in particle propulsion to the right (+z direction). Similarly, the particle is propelled to the left at t ∼ 9 and t ∼ 16 s, when the right laser beam power is increased and the left laser beam power is decreased. The particle can be scanned over a wide region (∼30 µm), which is only limited by the focal spot size of laser beams. In this example, the particle is propelled at the maximum possible © 2010 American Chemical Society

FIGURE 5. (a, b) Schematic illustration of scanning mechanism of plasmonic trapping for equal (part a) and unequal (part b) beam intensities. IL, IR: intensities of +z and -z propagating SPPs excited by left and right laser beams, respectively. Key: Fz, total scattering force in +z direction; A, B, centers of laser beams incident from left and right, respectively; O, equilibrium position of trapped particle; (c) measured scanning trajectory of a trapped 1 µm particle in x-z plane.

velocity (∼15 µm/s) by completely turning off one of two beams. Accurate positioning along the gold stripe can be achieved by the mechanism described in parts a and b of Figure 5. The time taken for a particle to reach its equilibrium position is proportional to the trapping stiffness in the z direction and related to the gradient of the SPPs’ intensity profile, which is a combined effect of intensity gradient of focused laser spot and the decay with distance of the SPPs excited on the gold stripe. As increasing the intensity gradient requires the spot size to be reduced, there is a trade off between the scanning range and the trapping stiffness in this method of scannable plasmonic trapping. Over the entire scanning range, the particle is trapped stably in the direction transverse to the gold stripe (x direction). Variation in the x position is minimal and is due to Brownian motion. Using fluorescence microscopy (Figure 1b), we are able to measure the trajectory of a trapped 1 µm diameter polystyrene particle on a 1 µm wide gold stripe. The uncertainty of the particle tracking system is below 5 nm, as found by measuring the standard deviation of a fixed bead. As shown in Figure 6a, the trapping potential is highly anisotropic in the x and z directions. Strong confinement and trapping stiffness are achieved in the x direction due to the field confinement on the stripe. Along the z direction, the 3509

DOI: 10.1021/nl101653n | Nano Lett. 2010, 10, 3506-–3511

factors of the scattering optical force Fz and of the trapping force Fxy are 3.4 and 7.1, respectively, relative to the case of TIR at the glass-water boundary. In order to compare simulation with experiment, the illumination intensity is taken to be the same as that used in experiment (40 µW/ µm2) in the following calculations. When the particle is 40 nm above the gold stripe, the maximum value of the force Fx (1.14 pN) occurs when the particle center is ∼390 nm along the x axis from the center of the gold stripe. The trapping stiffness for small displacements about the center is found to be 3.73 pN/µm, which is in reasonable agreement with the experimentally measured stiffness of 1.7 pN/µm. The integration of the optical force with position over the trapping region gives a calculated transverse trapping potential depth of ∼192 kBT (T ) 300 K). The actual trapping potential depth can then be estimated by multiplying the numerically predicted potential depth by the ratio of the measured trapping stiffness to the numerically predicted stiffness. This yields a value of ∼88 kBT, which represents reasonably stable optical trapping. Although the quasi-TM leaky modes used in this work cut off when gold stripe is narrower than ∼1 µm, we anticipate that it will be possible to apply our method for scannable plasmonic trapping to much narrower structures, such as square, wedge-shaped stripes, or V-grooves in a thick metal film.25-27 It may be possible to achieve trapping along the long axes of these structures through the intensity gradient resulting from the surface plasmon decay length. To summarize, a scannable integrated optical tweezer using counterpropagating surface plasmon polaritons on a gold stripe is proposed and experimentally demonstrated. The parameters for optical trapping are optimized by numerical simulations of the SPP modes and by particle propulsion experiments. Taking advantage of the field enhancement provided by the SPP mode on the gold stripe, a trapping stiffness of 1.7 pN/µm for 1 µm diameter polystyrene particles is achieved at an illumination intensity of 40 µW/µm2. A scanning range of over 30 µm, which is only limited by the

FIGURE 6. (a) Measured positions of the trapped particle, recorded over a period of 1 min at 30 measurements s-1. (b) Histogram of a particle’s position in the transverse direction (x direction), from data shown in part a.

trapping is weaker by comparison because it is the result of the intensity gradients of the focused laser beams. These intensity gradients are small because the beams are only weakly focused. The histogram of particle position along the x axis is plotted as Figure 6b. It exhibits a Gaussian distribution with a standard deviation of 51 nm. The trapping stiffness in this direction is then found to be 1.7 pN/µm based on equipartition, using (1/2)kBT ) (1/2)k〈∆x2〉. This represents a large trapping stiffness for plasmonic trapping and is achieved at a relatively low illumination intensity. Along the stripe direction (z axis), standard deviation in particle position is ∼800 nm, corresponding to a trapping stiffness of 27 fN/µm. To further characterize the optical trapping, we carry out 3D FDTD analysis to calculate the optical force using the Maxwell stress tensor method. The optical force experienced by a 1 µm diameter polystyrene particle at different positions above the 1 µm wide gold stripe is calculated and shown in Figure 7. When the particle is 40 nm above the surface of the gold stripe, along its center, the enhancement

FIGURE 7. 3D FDTD calculated power flow distributions for a 1 µm diameter polystyrene particle that is 40 nm above the surface of a 1 µm gold stripe, along its center, on (a) x ) 0 cross section and (b) z ) 0 cross section. (c) Optical forces and their directions for different particle positions, as found by 3D FDTD simulations and the Maxwell stress tensor method (d1) R + 80 nm, d2 ) R + 160 nm, d3 ) R + 240 nm, R ) 500 nm). An illumination intensity of 40 µW/µm2 is assumed. The base of each arrow is situated at the particle center. Blue arrows show magnitude and direction of force Fxy in xy plane. For force Fz, downward arrows represent forces in the +z direction. © 2010 American Chemical Society

3510

DOI: 10.1021/nl101653n | Nano Lett. 2010, 10, 3506-–3511

(7)

focused laser spot size, is demonstrated experimentally. We expect that the flexibility provided by the scanning capability we introduce here would be advantageous for particle manipulation and sorting applications. In addition, it should be possible to reduce the structure to subwavelength dimensions through the use of bound surface plasmon modes on gold nanowires. We predict that the increased field confinement resulting from reducing the size of the structure will find applications for flexible nanoparticle manipulation.

(8) (9) (10) (11) (12) (13) (14) (15)

Acknowledgment. This work was supported by the Defense Advanced Research Project Agency (DARPA) and by the National Science Foundation (NSF). Fabrication work was carried out in the Harvard Center for Nanoscale Systems, which is supported by the NSF.

(16) (17)

Supporting Information Available. FDTD calculation of optical forces and videos showing propulsion and scanning of polystyrene particles. This material is available free of charge via the Internet at http://pubs.acs.org.

(18) (19) (20)

REFERENCES AND NOTES (1) (2) (3) (4) (5) (6)

(21) (22) (23)

Ashkin, A. Phys. Rev. Lett. 1970, 24, 156–159. Merkel, R.; Nassoy, P.; Leung, A.; Ritchie, K.; Evans, E. Nature 1999, 397, 50–53. Moffitt, J. R.; Chemla, Y. R.; Aathavan, K.; Grimes, S.; Jardine, P. J.; Anderson, D. L.; Bustamante, C. Nature 2009, 457, 446–450. Cui, Y.; Bustamante, C. Proc. Natl. Acad. Sci. U.S.A. 2000, 97, 127– 132. MacDonald, M. P.; Spalding, G. C.; Dholakia, K. Nature 2003, 426, 421–424. Lang, M. J.; Fordyce, M. P.; Engh, A. M.; Neuman, K. C.; Block, S. M. Nat. Methods 2004, 1, 1–7.

© 2010 American Chemical Society

(24) (25) (26) (27)

3511

Moffitt, J. R.; Chemla, Y. R.; Izhaky, D.; Bustamante, C. Proc. Natl. Acad. Sci. U.S.A. 2006, 103, 9006–9011. Kawata, S.; Tani, T. Opt. Lett. 1996, 21, 1768–1770. Yang, A. H. J.; Moore, S. D.; Schmidt, B. S.; Klug, M.; Lipson, M.; Erickson, D. Nature 2009, 457, 71–75. Arnold, S.; Keng, D.; Shopova, S. I.; Holler, S.; Zurawsky, W.; Vollmer, F. Opt. Express 2009, 17, 6230–6238. Mandal, S.; Serey, X.; Erickson, D. Nano Lett. 2010, 10, 99–104. Okamoto, K.; Kawata, S. Phys. Rev. Lett. 1999, 83, 4534–4537. Kwak, E. S.; Onuta, T. D.; Amarie, D.; Potyrailo, R.; Stein, B.; Jacobson, S. C.; Schaich, W. L.; Dragnea, B. J. Phys. Chem. B 2004, 108, 13607–13612. Nie, S.; Emory, S. R. Science 1997, 275, 1102–1106. Novotny, L.; Bian, R. X.; Xie, X. S. Phys. Rev. Lett. 1997, 79, 645– 648. Righini, M.; Ghenuche, P.; Cherukulappurath, S.; Myroshnychenko, V.; Garcia de Abajo, F. J.; Quidant, R. Nano Lett. 2009, 9, 3387–3391. Righini, M.; Volpe, G.; Girard, C.; Petrov, D.; Quidant, R. Phys. Rev. Lett. 2008, 100, 186804. Wang, K.; Schonbrun, E.; Crozier, K. B. Nano Lett. 2009, 9, 2623– 2629. Zhang, W.; Huang, L.; Santschi, C.; Martin, O. J. F. Nano Lett. 2010, 10, 1006–1011. Garces-Chavez, V.; Quidant, R.; Reece, P. J.; Badenes, G.; Torner, L.; Dholakia, K. Phys. Rev. B 2006, 73, No. 085417. Kawata, S.; Sugiura, T. Opt. Lett. 1992, 17, 772–774. Johnson, P. B.; Christy, R. W. Phys. Rev. B 1972, 6, 4370–4379. Zia, R.; Selker, M. D.; Brongersma, M. L. Phys. Rev. B 2005, 71, 165431. Steinvurzel, P.; Yang, T.; Crozier, K. B. Proc. SPIE 2010, 7604, 760417. Verhagen, E.; Spasenovic, M.; Polman, A.; Kuipers, L. Phys. Rev. Lett. 2009, 102, 203904. Jung, J.; Sondergaard, T.; Bozhevolnyi, S. I. Phys. Rev. B 2007, 76, No. 035434. Boltasseva, A.; Volkov, V. S.; Nielsen, R. B.; Moreno, E.; Rodrigo, S. G.; Bozhevolnyi, S. I. Opt. Express 2008, 16, 5252–5260.

DOI: 10.1021/nl101653n | Nano Lett. 2010, 10, 3506-–3511