JOURNAL OF APPLIED PHYSICS 102, 123110 共2007兲

Characterization of long-range surface plasmon-polariton in stripe waveguides using scanning near-field optical microscopy Ildar Salakhutdinova兲 and Jagdish S. Thakurb兲 Department of Electrical and Computer Engineering, Wayne State University, Detroit, Michigan 48202, USA

Kristjan Leossonc兲 Institute of Science, University of Iceland, Reykjavik IS-107, Iceland

共Received 27 August 2007; accepted 24 October 2007; published online 28 December 2007兲 Propagation characteristics of long-range surface plasmon-polariton 共LRSPP兲 guiding along thin gold stripes embedded in polymer cover layers are investigated by scanning near-field optical microscopy 共SNOM兲. It is shown that the characterization of samples with cover layers up to 10 ␮m is feasible in the optical communication wavelength range. We found that the spatial dimension of the optical signal is directly related to the geometrical dimension of the guiding layer, and the light collected by the SNOM is scattered light from the surface and not the evanescent field. We also discuss the limitations of the SNOM technique for the characterization of LRSPP modes. © 2007 American Institute of Physics. 关DOI: 10.1063/1.2826910兴 I. INTRODUCTION

The use of nanotools like focused ion beam 共FIB兲 milling and e-beam lithography for fabrication of devices has recently gained a huge interest for the development of optical nanotechnology which is also characterized by nanocharacterization tools like the scanning near-field optical microscope 共SNOM兲 technique.1 Determination of the local electric-field distribution supported by nanoscale structures working below the diffraction limit is a very important aspect related to the functionalities of these devices. The SNOM technique has the potential to determine the electricfield distribution confined to nanometer scales, and can be applied to investigate the characteristics of different structures: dielectric waveguides, whispering gallery modes, and surface plasmon-polariton 共SPP兲 from their evanescent fields.2,3 The mapping of a guiding structure having extra layers between the guiding medium and air is a complex problem, and in this paper we have investigated the possibility of using the SNOM technique to map long-range surface plasmon-polariton 共LRSPP兲 modes propagating along a thin metal stripe embedded in the polymer layers. LRSPP nanostructures, which are a relatively new class of guided structures, have been investigated for more than 25 years because of their larger propagation length of the light signal compared to SPP-based structures.4,5 In design, the Sarid’s structure for the generation of SPP modes, which consists of a thin metal film suspended between two dielectrics, is quite simple. However, the propagation length for SPP modes, which is limited by the ohmic losses, is smaller than 100 ␮m in the near-infrared region. On the contrary, the propagation length of LRSPP modes is quite large and can become of the order of centimeters in the same wavelength range. Due to the larger propagation lengths of LRa兲

Electronic mail: [email protected]. Electronic mail: [email protected]. c兲 Electronic mail: [email protected]. b兲

0021-8979/2007/102共12兲/123110/5/$23.00

SPP, these structures can be utilized in the areas of largescale integrated optics, nonlinear optics devices,6 and sensors.7 Initially LRSPP structures were proposed for optical communication.8,9 This application provided a new approach to use LRSPP structures as basic elements for both passive and active integrated optical components. Later on, it was shown that polymer-based LRSPP structures can also be used for optical communications.10 Because of a wide variation in the optical parameters of polymers, they can be used in different optoelectronic applications including nanoimprinting technology––a very effective tool for creation of nanosize objects.11 This technology also has an additional advantage of mass production in comparison with other nanotechnologies like FIB milling and e-beam lithography.

II. TECHNOLOGY AND EXPERIMENTS

Investigation of the LRSPP structures by the SNOM technique is a challenging subject because of the very specific nature of these guiding modes. In the case of dielectric waveguides and SPP structures, the guiding structure makes a direct contact with surface. However, in the case of the LRSPP structure, light propagates under a thick cover layer whose thickness can vary from 2 to 10 ␮m. To minimize the energy loss of modes due to leakage, the buffer layer must be thick enough to ensure that the intensity of the evanescent tail in the guided field is negligible at the interface between the polymer film layer and silicon substrate. This requires a relatively thick cover layer whose thickness normally varies from 10 to 15 ␮m due to the high refractive index of the silicon substrate.12 The plasmon field of the LRSPP mode mainly remains confined to the surrounding dielectric media, and only a small fraction of it decays exponentially with distance from the metal.10 As a result, only the tail of the evanescent wave decays in the cover dielectric layer which can be detected in

102, 123110-1

© 2007 American Institute of Physics

Downloaded 31 Dec 2007 to 141.217.43.133. Redistribution subject to AIP license or copyright; see http://jap.aip.org/jap/copyright.jsp

123110-2

J. Appl. Phys. 102, 123110 共2007兲

Salakhutdinov, Thakur, and Leosson

FIG. 2. 共a兲 Far-field output view. 共b兲 SNOM images of light propagating in the stripe LRSPP waveguides with cover layer thickness equal to 2 ␮m 共I兲, 4 ␮m 共II兲, 8 ␮m 共III兲, 10 ␮m 共IV兲. The stripe width is 5 ␮m for 共I兲, 8 ␮m for 共II兲 and 共III兲; the stripe is 10 ␮m for 共IV兲.

FIG. 1. 共Color online兲 共a兲 Configuration of the LRSPP waveguide structure. 共b兲 Scheme of the SNOM experiment. 1-input PM fiber, 2-LRSPP sample, 3-fiber tip, 4-piezo-transducer, 5-InGaAs pin-PD, 6-feedback control electronics, 7-computer control and display.

the air/dielectric layer boundary. The penetration depth d p of the evanescent field can be calculated3 using the following formula: dp =

␭

, 2␲冑n21 sin2 ␪ − n22

共1兲

where ␪ is the incidence angle, ␭ represents the wavelength, n1 and n2 are refractive indices for the bordering media. In our case n1 = 1.535, n2 = 1.0, and using ␪ = 90°, we get d p of about 212 nm for ␭ = 1550 nm. The field distribution for LRSPP is very different from the field distribution for SPP. The field amplitude reached maxima at the “metal-dielectric” boundary decaying toward the “dielectric-air” boundary. This property affects, for example, the sensitivity of the potential LRSPP sensor, although the resolution of such a sensor is high.13 We have investigated samples whose thickness of the substrate layer 关Fig. 1共a兲兴 varies from 12 to 20 ␮m. These samples were fabricated by spin-coating method on silicon substrate with a layer of CYCLOTENE 共Dow Chemical Co., Midland, MI兲 and a coating layer of UV photoresist. Stripes were patterned using UV exposure. After the deposition of a 10 nm thick gold layer, the photoresist was removed using a lift-off procedure. We then deposited the top cover layer by spin-coating of CYCLOTENE. The LRSPP stripe modes were excited by a tunable laser source 共1500–1640 nm range兲 delivered by a single-mode polarization-maintaining fiber via end-fire coupling. The polarization of the laser light was orthogonal to the metal layer. SNOM measurements were made by an uncoated dielectric fiber tip in the collection mode using a DualScope™ DS95–200 SNOM scan head

共Danish Micro Engineering A/S, Herlev, Denmark兲. The ultrasensitive InGaAs femtowatt photoreceiver model 2153 共New Focus, San Jose, CA兲 has been used for light collection. The schematic of the experimental setup is shown in Fig. 1. Samples with cover layer thickness of 2, 4, 8, and 10 ␮m were investigated. Figure 2 presents the SNOM images of the structures having different cover layer thickness measured at ␭ = 1570 nm. It is interesting to note that the contrasts of the SNOM pictures do not show any noticeable dependence on the thickness of the cover layer. In all the cases, we observed a strong signal in the stripe area, and the size of the light stripe in the SNOM picture is basically equal to the actual size of the metal stripe. This behavior continues to exist, even in our thickest sample having a 10 ␮m cover layer thickness. The far-field measurements show that the shape of the mode and its field distribution depend weakly on the geometrical asymmetry of the surrounding BCB layers, as almost the same pictures have been observed for both strongly asymmetric and near-symmetric structures. To understand the behavior of light propagation in the LRSPP stripe structures, we made measurements for different distances of the fiber tip from the waveguide surface. From Eq. 共1兲 we found that the penetration depth of the evanescent field for such geometry is around 212 nm. If we assume that the light collected by the fiber probe originates from the evanescent field, then one expects the strong signal decaying with distance from the waveguide surface and eventually becoming zero for distances larger than 212 nm. Figure 3 presents SNOM pictures for different distances varying from 0 to 1000 nm from the waveguide surface. Contrary to the expectation, strong optical signals are observed from the stripe area for all the distances up to 1000 nm. The signal cross sections showing their intensity distributions are presented for distances h = 0 and 500 nm on the right side of Fig. 3. It is interesting to note that the signal

Downloaded 31 Dec 2007 to 141.217.43.133. Redistribution subject to AIP license or copyright; see http://jap.aip.org/jap/copyright.jsp

123110-3

J. Appl. Phys. 102, 123110 共2007兲

Salakhutdinov, Thakur, and Leosson

FIG. 4. The illustration of specular 共a兲 and diffuse 共b兲 reflection in the guiding layer. 共Courtesy of Professor E. Fred Schubert from the Rensselaer Polytechnic Institute, Troy, NY.兲

imperfections are causing diffuse reflection 关Fig. 4共b兲兴 leading to the scattering of light. The probability of light extraction following a reflection event is14 then given by p=R

FIG. 3. 共Color online兲 共a兲 SNOM images of light collected for different distances from the surface varying from 0 to 1000 nm. 共b兲 Signal cross section for h = 0 共i兲, and h = 500 nm 共II兲. Cover layer thickness is 2 ␮m; the stripe width is 5 ␮m. Signal amplitude varies from 1.77 to 4.15 V for h = 0, and from 1.42 to 3.88 V for h = 500 nm.

from the stripe area with h = 500 nm is stronger and less noisy compared to those measured near the waveguide surface. The signals collected by the fiber probe do not show any correlation with the evanescent wave, which clearly suggests that we are measuring the scattered light signal rather than the signal from the evanescent wave. III. LIGHT EXTRACTION FROM WAVEGUIDE

We have observed above that the signals collected by the fiber tip originate from the guiding structure. This can be further verified by considering another guiding structure, namely the waveguide structure used in light-emitting diode omnidirectional reflectors.14 The field distributions are different for the conventional dielectric waveguide 关Fig. 4共a兲兴 and guiding structure supporting LRSP modes, but the distribution of the electromagnetic wave decaying into the surrounding low-index medium is the same. In the case of specular reflection, the light signal propagates in the waveguide area and the only signal connected with the evanescent wave is the one that spills into the surrounding medium; its extent is defined by the penetration depth. The light trapped in the guiding area cannot leak into the surrounding medium. It was shown that, only in the case of diffuse reflection, light can be extracted from the guiding layer but not in the case of specular reflection.14 So, in our case it seems the surface

兰0␪cIdiff cos ␪ sin ␪␲d␪ Pdiff , Pdiff + Pspec 兰0␲/2Idiff cos ␪ sin ␪2␲d␪

共2兲

where R is the reflectivity, Pdiff and Pspec are the powers of diffuse and specular reflection, respectively. Idiff is the intensity of diffuse reflection along the normal direction. As one can see, the probability of such a reflection does not depend on the wavelength or propagation losses. According to this model 关Eq. 共2兲兴 the nonzero value of p means that the scattered light should be recorded by the SNOM measurements. We clearly observed strong signals indicating that the light collected by the probe is a scattered light due to diffuse reflection from the metal surface. If light scattering is specular, similar to the one observed in Fig. 4, our experimental setup can measure a small signal from the evanescent wave. However, note that we have a strong scattering on the metal surface. The electromagnetic field at the metal-dielectric boundary is much stronger than at the dielectric-air boundary. Thus, a signal originating from the scattered light can be detected by the SNOM technique. An interesting point is that this signal is directly related to the guiding light. IV. DISCUSSION

Our experimental results have shown that the SNOM technique can be used to investigate the propagation characteristics of light signals from LRSPP structures. We discuss the SNOM measurement response to different excitation wavelengths propagating through the LRSPP structures. Investigation of the signal dependence on wavelength is also useful to learn about the nature of light scattering: whether it is elastic 共Raleigh scattering兲 or inelastic 共Raman or Brillouin scattering兲. Results of the SNOM measurements for different wavelengths shown in Figs. 5 and 6 did not show any wavelength dependence when inspected visually and from the intensity distribution in the signal cross section. Although there were small changes in signal values, they could not be associated with the wavelength changes. These

Downloaded 31 Dec 2007 to 141.217.43.133. Redistribution subject to AIP license or copyright; see http://jap.aip.org/jap/copyright.jsp

123110-4

J. Appl. Phys. 102, 123110 共2007兲

Salakhutdinov, Thakur, and Leosson

FIG. 7. 共Color online兲 Images of guided wave propagating in the stripe waveguide at wavelength ␭ = 1570 nm after different treatment of SNOM signal by IMAGEJ. FIG. 5. 共Color online兲 Wavelength dependence of light distribution in the stripe area for the spectral range from 1500 to 1620 nm. Cover layer thickness is 2 ␮m; the stripe width is 5 ␮m. The size of the SNOM pictures is 50⫻ 50 ␮m.

observations do not contradict our hypothesis about the light extraction from the stripe LRSPP guiding structures by the SNOM technique. The fundamental property of the SNOM technique for the measurement of LRSPP is that it only measures the optical signal and not the topography signal in contrast to other cases of guiding waves 共like dielectric waveguides, or SPP兲, where a substantial part of the evanescent field can be collected by an optical fiber. One should note that the topographical measurements cannot be employed for the verification of optical signal from these structures. It has been observed earlier15,16 that many artifacts in the SNOM measurements are associated with the measurements of evanescent field which decreases exponentially with distance from the surface. In this case a small fluctuation in the distance between the tip and surface can lead to a large fluctuation in the signal intensity. However, in the case of LRSPP, the evanescent field decays inside the cover layer and thus cannot be characterized by any topographical measurements. The editing software tools used in almost all the SNOM pictures can generate misleading results. We have discussed this issue in Fig. 7, where different images of light propagation in the stripe waveguide are obtained by performing different treat-

ments by IMAGEJ software on the original SNOM image. It is clear from the figures that the characteristic features of the signal are the same for all the treatments. It was found17,18 that interaction between the different modes in a multilayer nanostructure leads to the creation of new types of guiding modes, called the bulk plasmonpolariton 共BPP兲 modes.17 These modes are highly localized in nanosize layers and cannot reach the surface. We believe that SNOM optical characterization, which does not require any direct contact with the surface, can be applied for the characterization of BPP guiding modes. However, the diffuse scattering mechanism for the BPP mode needs to be confirmed experimentally. V. CONCLUSIONS

We demonstrated that the SNOM technique can be used for the characterization of long-range surface plasmon structures, particularly for their light propagating properties. We showed that the mapping of LRSPP stripe waveguides by the SNOM technique is possible even for the cover layer with thickness up to 10 ␮m. We observed that the spatial dimension of the optical signal is directly related to the geometrical size of the guiding structure layer, thus clearly demonstrating the confinement of the signal to the guiding layer. We showed that light collected by the SNOM technique from the LRSPP structures is not the evanescent light but scattered light in the guiding area. One of the important advantages of

FIG. 6. 共Color online兲 Cross sections of light distribution for different wavelengths in the range from 1500 to 1620 nm for pictures in Fig. 5. The SNOM pictures are only given for wavelengths 1500 and 1620 nm. Signal amplitude varies from 1.53 to 9.82V for ␭ = 1500 nm, and from 1.27 to 8.92 V for ␭ = 1620 nm.

Downloaded 31 Dec 2007 to 141.217.43.133. Redistribution subject to AIP license or copyright; see http://jap.aip.org/jap/copyright.jsp

123110-5

the SNOM technique is that it can be used for characterization of guiding structures where the guiding layer is embedded deep in the structure, for example a BPP guiding structure. ACKNOWLEDGMENTS

The authors would like to thank Dr. Valentin Volkov, University of Aalborg, Denmark, and Dr. E. Fred Schubert, Rensselaer Polytechnic Institute, Troy, NY, for useful discussions. Our special thanks to Dr. Sergey Bozhevolnyi, University of Aalborg, Denmark, who inspired this work. P. N. Prasad, Nanophotonics 共Wiley-Interscience, New York, 2004兲. H. Raether, Surface Plasmons on Smooth and Rough Surfaces and on Gratings 共Springer, Berlin, 1988兲. 3 F. De Fornel, Evanescent Waves 共Springer, Berlin, 2001兲. 4 D. Sarid, Phys. Rev. Lett. 47, 1927 共1981兲. 5 J. J. Burke, G. I. Stegeman, and T. Tamir, Phys. Rev. B 33, 5186 共1986兲. 6 H. J. Simon, Y. Wang, L. B. Zhou, and Z. Chen, Opt. Lett. 17, 1268 共1992兲. 1 2

J. Appl. Phys. 102, 123110 共2007兲

Salakhutdinov, Thakur, and Leosson 7

G. G. Nenninger, P. Tobiska, J. Homola, and S. S. Yee, Sens. Actuators B 74, 145 共2001兲. 8 P. Berini, Opt. Lett. 24, 1011 共1999兲. 9 R. Charbonneau, P. Berini, E. Berolo, and E. Lisicka-Shrzek, Opt. Lett. 25, 844 共2000兲. 10 T. Nikolajsen, K. Leosson, I. Salakhutdinov, and S. I. Bozhevolnyi, Appl. Phys. Lett. 82, 668 共2003兲. 11 L. J. Guo, J. Phys. D 37, R123 共2004兲. 12 B. P. Pal, in Progress in Optics, edited by E. Wolf 共North-Holland, Amsterdam, 1993兲, Vol. XXXII, p. 20. 13 F. Pigeon, I. F. Salakhutdinov, and A. V. Tishchenko, J. Appl. Phys. 90, 852 共2001兲. 14 J. K. Kim, H. Luo, Y. Xi, J. M. Shah, T. Gessmann, and E. F. Schubert, J. Electrochem. Soc. 153, G105 共2006兲. 15 B. Hecht, H. Bielefeldt, Y. Inouye, D. W. Pohl, and L. Novotny, J. Appl. Phys. 81, 2492 共1997兲. 16 L. Billot, M. Lamy de la Chapelle, D. Barchiesi, S.-H. Chang, S. K. Gray, J. A. Rogers, A. Bouhlier, P.-M. Adam, J.-L. Bijeon, G. P. Widerrecht, R. Bachelot, and P. Royer, Appl. Phys. Lett. 89, 023105 共2006兲. 17 J. Elser, V. A. Podolskiy, I. Salakhutdinov, and I. Avrutsky, Appl. Phys. Lett. 90, 191109 共2007兲. 18 I. Avrutsky, I. Salakhutdinov, J. Elser, and V. A. Podolskiy, Phys. Rev. B 75, 241402 共2007兲.

Downloaded 31 Dec 2007 to 141.217.43.133. Redistribution subject to AIP license or copyright; see http://jap.aip.org/jap/copyright.jsp