H ANDAI N NOPHOT ONIC S HANDAI NA ANOPHO TONI CS
V o l uu m m ee 2 2
NANOPLASMONICS Fundamentals From Fundamentals to Applications
This Page is Intentionally Left Blank
HANDAI NANOPHOTONICS H ANDAI N ANOPHOTONICS
V o l uu m mee 22
NANOPLASMONICS From Fundamentals to Applications Proceedings of the 2nd International Nanophotonics Symposium Handai July 26-28th 2004, Suita Campus of Osaka University, Osaka, Japan
Edited by
Satoshi Kawata Kawata and Hiroshi Hiroshi Masuhara Masuhara Department of Applied Applied Physics Department Osaka University University Suita, Osaka, Japan
ELSEVIER Amsterdam • Boston •• Heidelberg • London •• New York •• Oxford • Paris San Diego • San Francisco •• Singapore •• Sydney • Tokyo
ELSEVIER B.V. Radarweg 29 211,1000 P.O. Box 211, 1000 AE Amsterdam The Netherlands
ELSEVIER Inc. 525 B Street, Suite 1900 San Diego, CA 92101-4495 USA
ELSEVIER Ltd Langford Lane The Boulevard, Langford Kidlington, Oxford OX5 1GB UK
ELSEVIER Ltd 84 Theobalds Road WC1X8RR London WC1X 8RR UK
© 2006 Elsevier B.V. All rights reserved. This work is protected under copyright by Elsevier B.V., and the following following terms and conditions apply to its use: Photocopying of single chapters may be be made for as allowed by by national copyright laws. Permission of of the Single photocopies of for personal use as of aa fee is is required for all other photocopying, including multiple or or systematic copying, copying for Publisher and payment of for all for puiposes, resale, and all forms of document delivery. deliveiy. Special rates are available for educational institutions advertising or promotional purposes, non-profit educational classroom use. that wish to make photocopies for non-profit Permissions +44) 1865 843830, fax +44) Pel-missions may be be sought directly from Elsevier’s Elsevier's Rights Department in in Oxford, UK: phone ((+44) fax ((+44) 1865 853333, e-mail:
[email protected].
[email protected]. Requests may also be be completed on-line via the Elsevier homepage (http:// 1865 www.elsevier.com/locate/permissions). www.elsevier.com/locate/permissions). the Copyright Clearance Center, Inc., 222 Rosewood In the the USA, users may clear permissions and make payments through the in the UK through the Copyright Drive, Danvers, MA 01923, USA; phone: ((+1) +1) (978) 7508400, fax: ((+1) +1) (978) 7504744, and in 90 Tottenham Court Road, London W1P 0LP, UK; phone: ((+44) +44) Licensing Agency Rapid Clearance Service (CLARCS), 90 20 7631 5555; fax: ((+44) +44) 20 7631 5500. Other countries may have aa local reprographic rights agency for payments.
Derivative Works of contents may be reproduced for of the Publisher is is required for external resale or Tables of for internal circulation, but permission of or of such material. Permission of of the Publisher is is required for all other derivative works, including compilations and distribution of translations. Electronic Storage or Usage Permission of the Publisher is required to store or use electronically any material contained in this work, including any chapter or of aa chapter. part of Except as outlined above, no part of of this work may be reproduced, stored in in aa retrieval system or transmitted in in any form or by any means, electronic, mechanical, photocopying, recording or otherwise, without prior written permission of the Publisher. Elsevier's Rights Department, at the fax and e-mail addresses noted above. Address permissions requests to: Elsevier’s
Notice is assumed by the Publisher for any injury and/or damage to persons or property as aa matter of products liability, No responsibility is or otherwise, or or from from any use or or operation of of any methods, products, instructions or or ideas contained in in the material negligence or of rapid advances in the medical sciences, in particular, independent verification verification of of diagnoses and drug dosages herein. Because of should be made.
First edition 2006 B r i t i of sh Congress Library Cataloging Cataloguing in in Publication Data Library catalogue record is is available from the Library B r i t i of sh Congress. Library. A catalog in Publication Data British Library Cataloguing in is available from the British Library. A catalogue record is ISBN-13: ISBN-10: ISSN:
978 00 444 52249 8 8 0 444 52249 2 0 1574-0641 1574-0641
@ The paper used in this publication meets the requirements requii-ements of ofANSI/NISO Z39.48-1992 (Permanence of ofPaper). Printed in The Netherlands.
Working together to grow libraries in developing countries www.elsevier.com | www.bookaid.org | www.sabre.org
ELSEVIER
BOOK AID International
Sabre Foundation
v
Preface The second volume of the Handai Nanophotonics Book Series features "Nanoplasmonics," a recent hot topic in nanophotonics, impacting a diverse range of research disciplines from information technology and nanotechnology to bioand medical sciences. The interaction between photons and metal nanostractures leads to interesting and extraordinary scientific phenomena and produces new functions for nano materials and devices. Newly discovered physical phenomena include local mode of surface plasmon polariton excited in nanoparticles, hot spots on nano-rods and nano-cones, long range mode of surface plasmons excited on thin metal films, and dispersion relationship bandgaps of surface plasmons in periodic metal structures. These have been applied to, for example, single molecule detection and nano-imaging/spectroscopy, photon accumulation for lasing applications, optical nano-waveguides and nano-circuits. In July 2004, we had a two-day symposium with distinct scientists to discuss the latest progress in this exciting field. The second volume was co-authored by those participants. The book starts with a statement by John Pendry, the pioneer of nanoplasmonics. The first part, the theory of nanoplasmonics, includes four chapters written by Shalaev, Martin-Morenoa, Fukui, and Takahara. The second part, plasmonic enhanced spectroscopy and molecular dynamics, is written by Watanabe, Futamata, Hayashi, Ishida, Kajikawa, Ozaki, and Asahi. In part 3, recent progress of plasmonic materials and devices are reviewed by Okamoto, Pileni, Yamada, Yoshikawa, Sun, and Ishihara. In addition, we had quite a few participants sharing the common interest in exciting nanophotonics science, although they were not able to contribute to this book. We would like to thank all the contributors and participants to the Handai Nanophotonics Book Series and Handai Nanophotonics Symposium 2. Satoshi Kawata and Hiroshi Masuhara at Handai, Suita, Japan
This Page is Intentionally Left Blank
vii Vll
Dedicated to the late professor Osamu Nakamura
Osamu Nakamura Professor of Applied Physics and Frontier Biosciences, March 23, 1962 to January 23,2004, who has ever loved the optical science and microscopy. Osamu Nakamura made a great contribution to computed-tomography microscopy, confocal laser microscopy, super-resolved nano-imaging theory, near-infrared bio-medical spectroscopy, and many other related nano-scale photon science and technologies. He has served the international community by organizing international conferences, inviting international scientists and students to Osaka, and fostering international research collaborations. He published a number of papers in nanophotonics and biophotonics, for imaging analysis, diagnosis, and fabrication. Professor Nakamura visited the conference site of the Handai Nanophotonics Symposium II in July 2004 in his wheel chair and exchanged friendship with his old friends. In his funeral, hundreds of his friends and students came to farewell him. We all miss him, and wish he will guide us.
viii Vlll
Organization of The International Nanophotonics Symposium Handai on Plasmonics: from fundamentals to applications Sponsored by Nanotechnology Researchers Network Center of Japan The Murata Science Foundation Handai Frontier Research Center, Osaka University Nanonet The Ministry of Education, Culture, Sports, Science and Technology has started Nanotechnology Support Project, the five year project, to strategically promote Japanese nanotechnology research collaborations among industry, aeademia, and government. The major roles of Nanotechnology Support Project are (i) providing opportunities to use Ultra-HV TEM, Nano Foundries, Synchrotron Radiation, and Molecular Synthesis and Analysis through Japanese top institutions attending the project, and (ii) providing information on both Japanese and International nanotechnology research activities. To perform these activities smoothly, "Nanotechnology Researchers Network Center of Japan (Nanonet) was launched in 2002. Chairpersons Satoshi Kawata (Department of Applied Physics, Osaka University; Nanophotonics Lab, RIKEN) Hiroshi Masuhara (Department of Applied Physics, Osaka University) Local Organizing Committee Osamu Nakamura (Department of Frontier Bioscience, Osaka University) Takayuki Okamoto (Nanophotonics Lab, RIKEN) Yasushi Inouye (Department of Frontier Bioscience, Osaka University) Tsuyoshi Asahi (Department of Applied Physics, Osaka University) Hong-Bo Sun (Department of Applied Physics, Osaka University) Katsumasa Fujita (Department of Frontier Bioscience, Osaka University) Satoru Shoji (Department of Applied Physics, Osaka University) Taro Ichimura (Department of Applied Physics, Osaka University)
ix IX
Introductory Remarks to the Handai Proceedings Since the beginning of recorded history light has been both a subject of natural curiosity and a tool for investigation of other phenomena. So closely is light linked to our understanding of the world that "I see" can mean the same as "I understand". Light brought the first information about the distant objects of our universe, and light revealed the first secrets of the microscopic world. Yet in recent times, despite its continuing importance in our lives, there are signs that light is losing its grip on the frontiers of technology. To 'see' the very small we turn to the electron microscope, or the scanning tunneling microscope. These tools are commonly deployed in the world of nanotechnology which is the focus of huge research investment and, through the semiconductor chip, has already revolutionised our lives. The photon with its scarcely sub-micron wavelength is a clumsy and myopic beast in this new world where the electron easily outclasses it in compactness. Electronics has very much led the field in the world of nanotechnology all the way from integrated circuits to quantum dots. Yet the photon's ability to move around so rapidly with minimal disruption of the medium is still prized: there is still work to be done by this ancient tool. Here plasmonics steps into the limelight. A synthesis between light and the collective motion of electrons, the plasmon can move almost as quickly as light, but can also be gathered into incredibly small dimensions to challenge the electron itself in compactness. It naturally inhabits the world of nanotechnology. In this book we have articles by the leaders in this new field. As yet the commercial applications are relatively modest, but the promise is huge and the rich variety of topics represented shows just how much potential is waiting to be unlocked by our researchers. J. B. Pendry Imperial College London July 2005
This Page is Intentionally Left Blank
xi XI
Participants List SusumuAruga Takahiro Asada Tsuyoshi Asahi Harry Atwater Kuo Pin Chiu Tai Chi Chu Xuan-Ming Duan Jing Feng Ulrich Fischer Yuan Hsing Fu Ayako Fujii Akiko Fujita Katsumasa Fujita Masuo Fukui Masayuki Futamata
Kazuyoshi Hakamata Keisaku Hamada Tomoya Harada Kazuhiro Hashimoto Mamoru Hashimoto Shinji Hayashi Norihiko Hayazawa Taro Ichimura Takashi Ihama Ryoichi Imanaka Akio Inoshita Yasushi Inouye Akito Ishida
SEIKO EPSON Corporation Department of Mechanical Science and Bioengineering, School of Engineering Science, Osaka University Department of Applied Physics, Osaka University Thomas J. Watson Laboratory of Applied Physics, California Institute of Technology Department of Physics, National Taiwan University Department of Physics, National Taiwan University Technical Institute of Physics and Chemistry (TIPC), Chinese Academy of Science (CAS) Nanophotonics Laboratory, RIKEN U.C. Fischer Physics Institute, University of Muenster Department of Physics, National Taiwan University Department of Human and Environmental Science, Kyoto Prefecture University Department of Frontier Biosciences, Osaka University Department of Applied Physics, Osaka University Department of Optical Science and Technology, Faculty of Engineering, The University of Tokushima Nanoarchitectonics Research Center (NARC), National Institute of Advanced Industrial Science and Technology (AIST) FDK Corporation Department of Frontier Biosciences, Osaka University FDK Corporation Department of Chemistry, School of Science and Technology, Kwansei-Gakuin University Department of Mechanical Science and Bioengineering, School of Engineering Science, Osaka University Department of Electrical and Engineering, Kobe University Nanophotonics Laboratory, RIKEN Department of Frontier Biosciences, Osaka University Department of Applied Physics, Osaka University Handai FRC, Osaka University Techno Search Department of Frontier Biosciences, Osaka University Department of Human and Environmental Science, Kyoto Prefecture University
xii Xll
Teruya Ishihara Hidekazu Ishitobi Syoji Ito Masayuki Ito Tamitake Itoh Takashi Iwamoto Shigeki Iwanaga Yuqiang Jiang
Takamasa Kai Kotaro Kajikawa Koshiro Kaneko Yosuke Kanki Jun-ichi Kato Kazuya Kawahara Kosuke Kawahara Satoshi Kawata Ryoichi Kitahara Minom Kobayashi Maximilian Kreiter Aaron Lewis
Xiangang Luo Hiroshi Masuhara Ryota Matsui Luis Martin Moreno Yuji Morimoto Yu Nabetani Osamu Nakamura Toshihiro Nakamura Sana Nakanishi
Participants List
Exciton Engineering Laboratory, Frontier Research System, RIKEN Handai FRC, Osaka University Division of Frontier Materials Science, Osaka University AISIN COSMOS R&D Corporation Department of Chemistry, School of Science and Technology, Kwansei-Gakuin University Shimadzu Corporation Department of Applied Physics, Osaka University State Key Laboratory of Quantum Optics and Quantum Optics Devices, College of Physics and Electronic Engineering, Shanxi University Department of Applied Physics, Osaka University Interdisciplinary Graduate School of Science and Engineering, Tokyo Institute of Technology Department of Applied Physics, Osaka University Graduate School of Science and Technology, Kobe University Nanophotonics Laboratory, RIKEN Department of Applied Physics, Osaka University NEC Machinery Corporation Department of Applied Physics, Osaka University Department of Applied Physics, Osaka University Department of Applied Physics, Osaka University Max-Planck-Institut fur Polymerforschung Department of Applied Physics and The Center for Neural Computation, The Hebrew University of Jerusalem Exciton Engineering Laboratory, Frontier Research System, RIKEN Department of Applied Physics, Osaka University Department of Applied Physics, Osaka University Departamento de Fisica de la Materia Condensada, ICMA-CSIC, University of Zaragoza Department of Medical Engineering, National Defense Medical College Department of Applied Physics, Osaka University Department of Frontier Biosciences, Osaka University Department of Electrical and Engineering, Kobe University Department of Applied Physics, Osaka University
Participants List
Takashi Nakano Yasuro Niidome Kimihiko Nishioka Hiroshi Noge Watara Nomura Toshihiko Ochi Isamu Oh Keishi Ohashi Takayuki Okamoto Kaoru Okamoto Kazunori Okihira Masatoshi Osawa Taisuke Ota Oskar Painter John Pendry Marie-Paule Pileni Yuika Saito Suguru Sangu Akihiro Sato Vladimir M. Shalaev Akiyoshi Shibuya Ayako Shinjo Koichiro Shirota Satora Shoji Michel Sliwa Nicholas Smith Takayoshi Suganuma Teruki Sugiyama Yung Doug Suh Fumika Sumiyama Hong-Bo Sun Qian Sun Toru Suwa
xiii xill
National Institute of Advanced Industrial Science and Technology (AIST) Department of Applied Chemistry, Kyushu University Olympus Corporation Matsushita Electric Works, Limited Department of Electronics Engineering, The University of Tokyo Enplas Laboratories, Inc. Department of Applied Physics, Osaka University NEC Corporation Nanophotonies Laboratory, RIKEN Canon Inc. Department of Electrical and Engineering, Kobe University Catalysis Research Center, Hokkaido University Department of Frontier Biosciences, Osaka University Thomas J. Watson, Sr. Laboratory of Applied Physics, California Institute of Technology The Blackett Lab,, Imperial College London Faculty of Science, University P & M Curie Nanophotonies Laboratory, RIKEN Ricoh Company, Limited Interdisciplinary Graduate School of Science and Engineering, Tokyo Institute of Technology School of Electrical and Computer Engineering, Purdue University Zeon Corporation Department of Human and Environmental Science, Kyoto Prefecture University Nanophotonies Laboratory, RIKEN Department of Applied Physics, Osaka University Department of Applied Physics, Osaka University Department of Frontier Biosciences, Osaka University Enplas Laboratories Inc. Department of Applied Physics & Handai FRC, Osaka University Korea Research Institute of Chemical Technology Department of Information and Physical Sciences, Osaka University Department of Applied Physics, Osaka University College of Physics, Nankai University Department of Applied Physics, Osaka University
xiv XIV
Takuji Tada Atsushi Taguchi Kenji Takada Junichi Takahara Kenji Takubo Mamoru Tanabe Kazuo Tanaka Hiroaki Tanaka Yoshito Tanaka Nao Terasaki Ryo Toyota Din Ping Tsai Tomoya Uchiyama Yasuo Ueda Arvind Vengurlekar Prabhat Verma Hiroyuki Watanabe Tadaaki Yabubayashi Sunao Yamada Yoshimiehi Yamada Kazuo Yamamoto Peilin Perry Yang Takaaki Yano Ryohei Yasukuni Hiroyuki Yoshikawa Yasuo Yoshikawa Masayuki Yuki Kenichi Yuyama Remo P. Zaccaria
Participants List
Department of Applied Physics, Osaka University Department of Frontier Biosciences, Osaka University Department of Applied Physics, Osaka University Graduate School of Engineering Science, Osaka University Shimadzu Corporation Department of Applied Physics, Osaka University Department of Electronics and Computer Engineering, Gifu University Murata Mfg Company Limited, Department of Applied Physics, Osaka University Photonics Research Institute, AIST Department of Applied Physics, Osaka University Department of Physics, National Taiwan University Department of Applied Physics, Osaka University Sumitomo Titanium Corporation Frontier Research System, RIKEN Department of Applied Physics, Osaka University Department of Applied Physics, Osaka University Sumitomo Precision Products Company Limited Department of Applied Chemistry, Kyushu University Department of Applied Physics, Osaka University Department of Applied Physics, Osaka University Department of Physics, National Taiwan University Department of Applied Physics, Osaka University Department of Applied Physics, Osaka University Department of Applied Physics, Osaka University Department of Chemistry, School of Science and Technology, Kwansei-Gakuin University International Reagents Corporation Department of Applied Physics, Osaka University Department of Applied Physics, Osaka University
July 26-28th, 2004 Icho-Kaikan in Suita Campus, Osaka University, Osaka, Japan
xv
This Page is Intentionally Left Blank
XV11 xvii
TABLE OF CONTENTS Preface Organization of the Symposium Introductory Remarks to the Handai Proceedings Participants List Group Photograph of the Symposium
v viii ix xi xv
PART I: THEORY OF NANOPLASMONICS Chapter 1: Magnetic plasmon resonance A. K. Sarychev, G Shvets, and V. M. Shalaev
3
Chapter 2: Theory of optical transmission through arrays of subwavelength apertures L. Martin-Moreno, J, Bravo-Abad, F. Lopez-Tejeira and F.J. Garcia-Vidal 15 Chapter 3: Linear and nonlinear optical response of concentric metallic nanoshells M. Fukui, T. Okamoto and M. Haraguchi Chapter 4:
31
Low-dimensional optical waveguides and wavenumber surface J. Takahara and T. Kobayashi 55
PART I I : PLASMON ENHANCED SPECTROSCOPY AND MOLECULAR DYNAMICS Chapter 5;
Specific Raman band shift caused by mechano-chemical effect in tip-enhanced near-field Raman Spectroscopy H, Watanabe, N. Hayazawa, Y. Inouye, and S. Kawata
81
Chapter 6:
Single molecule sensitivity in surface enhanced Raman scattering using surface plasmon M. Futamata and Y. Maruyama 101
Chapter 7:
Enhanced Raman scattering mediated by metallic surface-particle gap modes S. Hayashi 141
xviii
Table of Contents
Chapter 8:
Surface plasmon enhanced excitation of photofunctional molecules in nanospace towards molecular plasmonics A. Fujii and A. Ishida 153
Chapter 9:
Localized surface plasmon resonance enhanced second-harmonic generation K. Kajikawa, S. Abe, Y. Sotokawa, and K. Tsuboi 185
Chapter 10: Localized surface plasmon resonance-coupled photo-induced luminescence and surface enhanced Raman scattering from isolated single Ag nano-aggregates T. Itoh, K. Hashimoto, Y. Kikkawa, A. Ikehara, and Y, Ozaki 197 Chapter 11: Single particle spectroscopic study on surface plasmon resonance of ion-adsorbed gold nanoparticles T. Asahi, T. Uwada and H. Masuhara 219 PART III: MATERIALS AND DEVICES FOR NANOPLASMONICS Chapter 12: Enhancement of luminescence in plasmonic crystal devices T. Okamoto, F. H'Dhili, J. Feng, J. Simonen, and S. Kawata 231
Chapter 13: Intrinsic properties due to self-organization of 5nm silver nanoerystals M. P. Pileni
247
Chapter 14 : Gold nanorods: preparation, characterization, and applications to sensing and photonics S. Yamada and Y. Niidome 255 Chapter 15: Optical trapping and assembling of nanoparticles H. Yoshikawa, C. Hosokawa, and H. Masuhara
275
Chapter 16: Femtosecond laser fabrication of three-dimensional metallic micro-nanostructures H.-B. Sun, K. Kaneko, X.-M. Duan, and S. Kawata
289
Chapter 17: Nanophotolithography based on surface plasmon interference T. Ishihara and X. Luo
305
Author index
313
Subject index
315
PART I: THEORY OF NANOPLASMONICS
This Page is Intentionally Left Blank
Handai Nanophotonics, Volume 2 S. Kawata and H. Masuhara (Editors) © 2006 Elsevier B.V. All rights reserved.
3
Chapter 1
Magnetic plasmon resonance A. K. Sarychev\ G. Shvets\ and V. M. Shalaevc a
Ethertronics Inc., San Diego, CA 92121,
department of Physics, The University of Texas at Austin, Austin, TX 78712 c
School of Electrical and Computer Engineering, Purdue University, West Lafayette, IN 47907 The optical properties of nanostructured metamaterials have been intensively studied during the last decade. It has been proposed by Pendry, who further developed earlier studies on negative refraction [1,2] that a metamaterial with negative dielectric permittivity e and negative magnetic permeability |J. could be used for developing a super-lens providing a sub-wavelength resolution. According to Pendry, when the scattered light passes through a material with a negative refractive index (specifically, when « = - ^ / = - l and the two impedances are matched), the evanescent components of the scattered field grow exponentially, allowing the restoration of the scattered image with subwavelength resolution. Smith, Padilla, Vier, and Shultz [3] have demonstrated negative-refraction materials in the microwave range. These materials are also referred to as double-negative or left-handed materials (LHMs), because the electric field and magnetic field along with the wavevector form a left-handed system in this case. In addition to super resolution, the unusual and sometimes counter-intuitive properties of LHMs make them very promising for applications in resonators, waveguides and other microwave and optical elements (see [4] and [5-7]). Huge enhancement of the local em field, accompanying the subwavelength resolution, can be used to enhanced Raman and nonlinear spectroscopy of atoms and molecules distributed over the surface ofaLHM. In spite of large efforts LHMs have not been demonstrated yet in the optical range. To obtain a negative refraction in the optical range, one needs to have a metamaterial with optical magnetism, which is a challenging problem because magnetism is typically weak in the high-frequency range. Relaxation
4
A. G. Shvets and V.and M. Shalaev A.K. K.Sarychev, Sarychev, G. Shvets V. M. Shalaev
times of paramagnetic and ferromagnetic processes are long in comparison with the optical period and collective magnetic responses become small at high frequencies. With no collective effects, the magnetic susceptibility is very small since it is proportional to v2 Ic1 <= J32 «10", where v is the velocity of electron in atom, c is the speed of light, and fi = e /he = 1/137. [This is because the ratio v/c appears first with the magnetic field H in the interaction Hamiltonian and again in the magnetic moment M of atoms.] For microwave LHMs artificial magnetic elements such as split-ring resonators (SPRs) and Swiss roll structures have been proposed and experimentally implemented [3, 4, 8]. In the microwave part of the spectrum metals can be considered as perfect conductors because the skin depth is much smaller than the metallic feature size. In the optical part of the spectrum, however, thin (sub-wavelength) metal components behave very differently because their sizes become comparable to the skin depth. This is the physical reason preventing the transfer of the approaches used in design of microwave LHMs to the optical range. By proper accounting for the metal properties in the optical range (finite e < 0), we demonstrate that the artificial magnetism can exist in (sub-wavelength) plasmonic structures. Artificial magnetism is caused by the magnetic plasmon resonance (MPR), which is primely determined by the geometry and material properties of the structure and to a lesser degree, by the ratio of the structure size and radiation wavelength X. Previously we proposed optical LHMs based on half-wavelength-long metal rods so that a magnetic resonance in this case was directly related to the wavelength [4, 9]. Here we show that MPR can occur in structures much smaller than the wavelength. Moreover, there is a close analogy between the electrical surface plasmon resonance (SPR) and MPR. The electrical SPR occurs in the optical and infrared part of the spectrum and results from a collective electron oscillation in metal structures. Consider, for example, an elliptical metal particle that has the electrical dipole polarizability a% <* [l + y(em - 1 ) ]
, where em is
the metal permittivity and y < 1 is the depolarization factor, which depends on the aspect ratio. For "good" optical metals (Ag, Au, Al, etc.), the real part of em is negative and large while its imaginary part is relatively small in the optical range. The plasmon resonance corresponds to the condition Reem{(o) = l — l/y and it critically depends on metal properties and the shape of a metal nanoparticle. For particles much smaller than the wavelength, the SPR is sizeand wavelength-independent. Many important plasmon-enhanced optical phenomena and applications of metal nanocomposites are based on the electrical SPR (see, for example, [10]). Below we show that along with the electrical SPR, specially arranged metal nanoparticles can support a MPR, with the resonance frequency ffi^, independent
Magnetic plasmon resonance
5
of the size and A. Such structures act as optical nanoantennas by concentrating large electric and magnetic energies on the nanoscale at the optical frequencies. The magnetic response is characterized by the magnetic polarizability a^ with the resonant behavior similar to a^: its real part changes the sign near the resonance and becomes negative for co> cor, as required for LHMs. Similar to the electrical SPR where the optical cross section of a nanostructure with size a«X can be as large as A the MPR can also be characterized by a large optical cross-section.
Fig. 1. Currents in the two-wire line excited by external magnetic field H. The displacement currents, "closing" the circuit, are shown by dashed lines
We consider first a pair of parallel metal rods. The external magnetic field excites the electric current in the pair of the rods as shown in Figl, The magnetic moment associated with the circular current flowing in the rods results in a magnetic response of the system. Suppose that an external magnetic field H = HQ exp(-icat) is applied perpendicular to the plane of the pair. The circular current /(z) excited by the magnetic field flows in opposite directions in the nanowire pair, as shown in Fig. 1. The displacement currents flowing between the nanowires close the circuit. We introduce the electric potential U(z)=fil E d between the pair where the integration is along the line |a(z),Z>(z)j. To find the current /(z), we integrate the Maxwell equation curl E = /&(HQ + H ; M ) over the contour {a, b, c, d} in Fig. 1, where k = m/c and Hin = curl A is the magnetic field induced by the current. It is assumed that the nanowire length 2a is much larger than the distance d between the nanowires
6
A. G. Shvets and V.and M. Shalaev A.K. K.Sarychev, Sarychev, G. Shvets V. M. Shalaev
and the radius of a nanowire b«d. We also assume that M « 1 , Under these assumptions, the vector potential A is directed along the nanowires (z direction) and the integration of the Maxwell equation gives
(IR - ikAz + dUldz)A = ikHodAz,
(1)
where the pair impedance i?=2/(rorf>2) = 8i7(s&2) [ e=iAttalm is the metal permittivity] and +IR/2 are the electric fields on the surface of the nanowires. The electric potential U(z) between the pair is given by solution of Maxwell's equations that can be written for a» d» bm
r2
where q(z)
is the electric charge per unit length, rt - ^(z-zf
+ b1 ,
R2 = ^j(z-z^2 + h2 and the terms ~{bla) are neglected. We explicitly separate in Eq. (2) the first term, which has a singularity when b —> 0; it can be estimated as 4q(z)]n(d/b). The second term in Eq. 2 is regular for b—»0 and we can expand it over dla « 1. Thus we obtain a local relation between U and q:
U(z) = Cq(z) , where C = \4lag(d/b)-3(d/af+ (dkf(2log(2a/d)-l)/2\ The vector potential Az can be found following the same procedure which results in Az(z) = (L/c)I(z) , where the inductance is given by
L-4\n(d/b)-(d/af[3
+ 4iak + 6log(2a/d)]/6. We substitute U(z) and A2(z)
into Eq. (1), taking into account the charge conservation law kll dz = iajq{z), and obtain the second-order differential equation for the current,
dzl where
P)
c -a
= I(a)-0
g2 = k2 \LC-8c\(kb)2em~]
,
and
parameter
g
is
given
as
• The product LC can be estimated as LC - 1. We
consider here the "quasistatic" case |8C[(to) 2 E m ]" 1 |»l when parameter G = ag depends only on the metal permittivity and aspect ratio: G2 « -2(a I bf \n(d lb)lem
(4)
Magnetic plasmon resonance
7
The case of the strong skin effect (|(to) 2 e m |" 1 «l, g= k), when the pair of metal wires has a so-called antenna resonance at ha —nil, was numerically simulated in our previous papers [9,11] and in papers by Panina et al. [12]. We solve Eq.(3) for the current I(z) and calculate the magnetic moment m = (2e) J [rxj(r)]dr , where j(r) is the density of the current and the integration is over the two nanowires as well as over the space between them where the displacement currents are flowing. Thus we obtain m = -Hoai\n(d/b)(kdy 2
=— G3
(5)
The metal permittivity em has a large negative value in the optical range while its imaginary part is small; therefore, the magnetic moment m has a resonance at G^jcll when the moment m attains large values. The magnetic resonance frequency (0=0^ depends on geometry of the system and material properties. In analogy with the electric SPR in metal nanoparticles, we see that for the MPR, the size of the sticks can also be arbitrary small in comparison with the wavelength of the incident light. This is in a striking difference with the previously considered magnetic resonance at a = /l/4 [9]. For a lossless metal the magnetic polarizability 4x(m/HQ) goes to -°° at the resonance. Thus, the MPR opens the possibility for engineering efficient LHM in the optical range. For a typical metal, the permittivity em(o)) can be well approximated by the Drude formula for the red and infrared parts of the spectrum: £m(aJ)B-((a/G)p)
/(l-ie^/o)),
where cop is the plasma frequency and the
relaxation parameter is small,co T lca«\. Then the polarizability a^ normalized to the volume V = Aahd of the pair has the following form near the MPR: aM=
= ^z
y
\\-mlm
- i m I(7m.)
(6)
where the resonance frequency (Or = bm)p ^2 \og(d I b) /(4a) . The plasma frequency mp is typically in the ultraviolet part of the spectrum so that aT « mp and the pre-factor in Eq. (6) can be on the order of one, even for a nanowire length 2a much smaller than the wavelength A of the incident light, so that a strong MPR can be observed. We can also estimate the optical crosssection for the MPR, a^ - a^VI A, assuming that the logarithm factor is ~ 1 and that radiation losses dominate (so that mn~aWlk
); this gives
8
A. K.Sarychev, Sarychev, G. Shvets M. Shalaev A. K. G. Shvets and V.and M. V. Shalaev
a (d/b). Thus, the magnetic cross-section a^ can be very large and, in particular, comparable to X (as in the case of the SPR in spheroids, where (Tg ~ X despite the fact that all sizes involved are much smaller than the wavelength. Thus, by employing both resonances, SPR and MPR, one can accomplish a strong coupling of nanostructures to both components of light, electrical and magnetic. We now consider a metal nanoantenna that has a horseshoe\ shape, which is obtained from a pair of nanowires by shorting it at one of the ends (see Fig 2). When the quasistatic condition |8C[(te) 2 E m ]"'|»l holds, the electric current I(z) in a horseshoe nanoantenna can be obtained from Eq. (3) where the boundary condition changes to Iz=a - (dl I dz)z-Q = 0 and, as above, a » d » b. It is easy to check that the magnetic polarizability a^ is still given by Eq. (6), where a is now equal to the total length of the horseshoe nanoantenna. Therefore, the horseshoe nanoantenna provides the same magnetic polarizability a^ at twice shorter length. Consider now a magnetic permeability ji for a metamaterial where the horseshoe nanoantennas are oriented in one direction ("z" direction in Fig. 1) and are organized in the periodic square lattice. The tensor ju component, which is in the direction perpendicular to the plane of the sticks (H direction in Fig. 1), can be estimated from the Lorenz-Lorentz formula [14] (jue-l)/(jue+l)-pa^/3 , where p is the volume concentration of the nanoantennas. Results of our calculations of fie = fi\ + ip.j f° r silver horseshoe nanoantennas are shown in Fig. 2; the optical parameters for silver were taken from [10, 15], As one can see in the figure, the negative magnetism can be observed, for example, in the near-infrared part of the spectrum, including the telecommunication wavelength of 1.5 fan. By varying nanoantenna parameters, one can tune the position of the MPR for any frequency in the visible and infrared parts of the spectrum. For practical applications of the optical magnetism losses may play an important role. We estimate losses (given by jUj ) at the wavelength corresponding to the condition //j = - 1 as far = Xpco^^llogidIb) l[%(Bpadp\, where Xf. is the resonance wavelength. For metals at room temperatures losses are significant (for silver, far ~ 0-3 > s e e Kg- 2) but still they are relatively small. These losses can be much smaller at low temperatures and atomic quality of metal crystals. We also note that the radiative losses (which are given by the small imaginary part of the inductance L) are of no importance for nanoantennas arranged in a periodic array (i.e., in a plasmonic crystal); the radiation
9
Magnetic plasmon resonance
corrections, in this case, result in a change in the spatial dispersion rather than in an increase of /xj •
Fig. 2. Optical magnetic permeability [i = f&i + H2 (f^[ - continuous line, fij " dashed line) of the composite containing c shaped silver nanoantennas; volume concentration/? = 0.3; left curves: a = 200 nm, d = 50 ran, b = 13 nm; right curves a = 600nm, d=90 nm, b = 13 B/M.
MAGNETIC FIELD nm
-2
-6
inn
200
nm
Fig. 3. Magnetic plasmon resonance in silver nanoantenna, which is placed in a maximum of external field HQ directed perpendicular to the plane; the frequency corresponds to
10 10
A. K.Sarychev, Sarychev, G. Shvets M. Shalaev A. K. G. Shvets and V.and M. V. Shalaev
The above results can be easily extended to the two-dimensional case where the metal nanoantennas have horseshoe profile in x,y plane and extended to z = ±00. The quasistatic case corresponds to the condition \k2bdem\»l, where b is the thickness of the walls of the horseshoe nanoantenna and d is the distance between the opposite walls. Then the resonance frequency a>r is defined by the equation G = 2a J-21 {£mbd) = jt 12, where a is the nanoantenna length. In Fig. 3 we show the local magnetic field in the silver nanoantenna that resonates at wavelength A = 1,5urn. Near the resonance magnetic field inside the nanoantenna is large in magnitude and it is directed opposite to the external field HQ , which results in negative magnetic permeability. The size of the nanoantenna is much smaller than the wavelength yet the resonance magnetic field is not curl-free since it changes direction at the walls. To estimate the effective magnetic permeability in this case we use the approach developed in Ref. [16]. Thus we obtain \iz -\ + P{SHQ) J" (Hin-Ho)ds for a plasmonic crystal composed by the nanoantennas, where Hin is the magnetic field inside a horseshoe nanoantenna; the integration is over the area s = da, and p is a concentration of the nanoantennas organized in a square lattice. Near the resonance we obtain the following equation for fiz = (321Ji)a pK~ (jt/2-G) . For a good optical metal the magnetic permeability has a sharp resonance and can acquire large negative values for co> cor as shown in Fig. 4.
A (/fin) 2.2
Fig. 4. Optical magnetic permeability fi = y.\ + ^2 (Ml " continuous line, ^2 ~ dashed line) of the composite containing silver nanoantennas shown in Fig. 3 organized in square lattice; volume concentration/? = 0.4.
Magnetic plasmon plasmon resonance
11 11
LHM can be obtained from the horseshoe composite by adding, e.g., metal nanowires as it was done in the microwave case [3], The horseshoe metamaterial itself can show a left-handed behavior when the nanoantennas are closely packed. We design two-dimensional dense periodic structure consisting of alternative up and down horseshoe nanoantennas. One half of the elementary cell is shown in Fig. 5a. (The structure repeats itself in x and y directions; separation between antenna centers is 80 nm). Dispersion relation co(kx ) for the electromagnetic wave propagating through the periodic structure in x direction has been calculated by numerically solving the Maxwell's equation for magnetic field Hz . For computational simplicity, we have assumed a hypothetical lossless plasmonic material with the frequency-dependent dielectric permittivity e =l-a>p/co , where 2nc/o)p = 225nm. The frequency co and the wavevector k are normalized to a>Q=2jrc/?iQ and ICQ = 2JZ/AQ , respectively, where
OM 003
r. CLC1
IMS •cci AC4
Fig. 5. Plasmonic crystal composed from horseshoe metal nanoantennas; separation between antennas centers 80 nm. Magnetic (color and contours) and electric (arrows) fields inside a periodic array of horseshoe-shaped nanoantennas at the cutoff kx = 0 (b) Dispersion relation co v.s. kx for a left-handed electromagnetic wave.
Remarkably, one of the propagating modes (shown in Fig. 5b) exhibits lefthandedness; its group velocity vRr = dco/dk opposes its phase velocity. Fig. 5 a shows the magnetic field profile and the electric field inside the elementary cell for kx = 0 (magnetic cutoff condition corresponding to \x = 0). Magnetic field is concentrated inside the horseshoes, and has opposite signs in the adjacent horseshoes. The dominant field in the structure is Ex which does not contribute to the Poynting flux in the propagation direction. Electric field is primarily potential (i.e. can be derived from an electrostatic potential), but has a nonvanishing solenoidal component that produces the magnetic field. The fact that
12 12
A. K. G. Shvets and V.and M. Shalaev A. K.Sarychev, Sarychev, G. Shvets V. M. Shalaev
the dominant electric field Ex does not change the sign inside the cell indicates that the mode in question does not owe its negative dispersion to the bandfolding effect common in photonic crystals. The left-handed behavior occurs in the vicinity of k = l.BSjUm which is close to the MPR resonance. Negative index waves described in this letter occur in plastnonic nanostructures with large negative dielectric permittivity E m «-1 and, therefore, they are conceptually different from the negative index waves in perfectly conducting structures [4] and in the structures with em — 1 [17]. We considered here two types of nanoantennas that support the MPR in the optical range. Other possible designs could include, for example, nanosized metal spheres sectored into eight equal parts by thin dielectric slits and split-ring resonators (SRRs), The SRRs were successfully used earlier for the microwave LHMs [3], A subwavelength SRR can provide a large magnetic polarizability at the resonance, when the radius is as follows R -c^e^lln, with d being the thickness of the dielectric slit in the ring. However, it seems hard, if not impossible, to have the concentration of SRRs large enough to provide a reasonable negative magnetic permeability in the optical range. Our estimates show that for the optimal concentration, a negative magnetic response of a SRR metamaterial is significantly smaller than for the horseshoe metamaterial considered above. Yet we would like to stress out that SRR metamaterials can have a large paramagnetic response in the optical range (with large and positive H) with many interesting applications. In conclusion, we show that a specially designed metal nanoantenna, which is much smaller than the light wavelength, can have a magnetic plasmon resonance (MPR) with the resonant frequency depending on the shape and material properties of the nanoantenna rather than on the wavelength. In this sense, the MPR is similar to the surface plasmon resonance (SPR) in a metal nanoparticle. We show that composites comprising such non-magnetic nanoantennas may have a large magnetic response in the optical spectral range. Metamaterials based on plasmonic nanoantennas supporting both SPR and MPR can have a dielectric permittivity and magnetic permeability, which are simultaneously negative, and thus act as left-handed materials in the optical and infrared spectral ranges. ACKONWLEDGEMENTS The authors acknowledge useful contributions and discussions with D. Genov, and V. Podolskiy. This work was supported in part by NSF grants ECS-0210445 and DMR-0121814, and by the ARO MURIW911NF-04-01-0203.
Magnetic plasmon resonance
13 13
REFERENCES [I] V. G. Veselago, Soviet Physics Uspekhi, 10 (1968) 509. [2] J. B. Pendry, Phys. Rev. Lett, 85 (2000) 3966. [3] D. R. Smith, W. J. Padilla, D. C. Vier, S. C. Nemat-Nasser, and S. Shultz, Phys. Rev. Lett., 84 (2000) 4184. [4] For recent references see the special issue of Opt. Express, 11 (2003) No 7. [5] A. A. Houck, J. B. Brock, and I. L. Chuang, Phys. Rev. Lett., 90 (2003) 137401. [6] C. G. Parazzoli, R. B. Greegor, K. Li, B. E. C. Koltenbah, and M. Tanielian, Phys. Rev. Lett., 90 (2003) 107401. [7] A. Alu andN. Engheta, IEEE T. Microw. Theory, 52 (2004) 199. [8] J. B. Pendry, A. J. Holden, D. J. Robbins, and W. J. Stewart, IEEE T. Microw. Theory, 47 (1999) 2075; M. C. K. Wiltshire, J. V. Hajnal, J. B. Pendry, D. J. Edwards, and C. J. Stevens, Opt. Express, 11 (2003) 709. [9] V. A. Podolskiy, A. K. Sarychev, and V. M. Shalaev, J. Nonlin. Opt. Phys. Mat., 11 (2002) 65; Opt. Express, 11 (2003) 735; A. K. Sarychev, V. P. Drachev, H. K. Yuan, V. A. Podolskiy, and V. M. Shalaev, Proc. SPIE, 5219 (2003) 1. II1] A. K. Sarychev and V. M. Shalaev, Phys. Rep., 333 (2000) 275. [12] A. N. Lagarkov and A. K. Sarychev, Phys. Rev. B, 53 (1996) 6318. [13] L.V. Panina, A. N. Grigorenko, D. P. Makhnovskiy, Phys. Rev. B, 66 (2002) 155411. [14] D. Landau and E.M. Lifshitz, Electrodynamics of Continuous Media, 2nd ed. Pergamon, Oxford, 1984. [15] J. D. Jackson, Classical Electrodynamics, J. Wiley & Sons, Inc., 1999. [16] U. Kreibig and M. Volmer, Optical Properties of Metal Clusters, Springer-Verlag, Berlin, 1995. [17] A. K. Sarychev, V. M. Shalaev, R. C. McPhedran, Phys. Rev. B, 62 (2000) 8531. [18] G. Shvets, Phys. Rev. B, 67 (2003) 035109.
This Page is Intentionally Left Blank
Handai Nanophotonics, Volume 2 S. Kawata and H. Masuhara (Editors) © 2006 Elsevier B.V. All rights reserved.
15 15
Chapter 2
Theory of optical transmission through arrays of subwavelength apertures L. Martin-Moreno8, J. Bravo-Abadb, F. Ldpez-Tejeira" and F.J. GarciaVidal" a
Departamento de Fisica, de la Materia Condensada, ICMA-CSIC, Universidad de Zaragoza, E-50009 Zaragoza,
b
Departamento de Fisica Teorica de la Materia Condensada, Universidad Autonoma de Madrid, E-28049, Spain 1. INTRODUCTION Surface plasmons (SPs) have long been known to be able to guide light on the surface of a metal and to concentrate light in subwavelength volumes. But another functionality of SPs was added to the previous list in 1998, when Ebbesen and co-workers found that SPs could enhance the transmission of light passing through subwavelength holes [1]. That seminal paper reported that, when a metal film is perforated with subwavelength holes and these are arranged in a two-dimensional (2D) periodic array, the transmission of light is greatly enhanced at some particular wavelengths. The experimental spectral location of transmission peaks was found to be related to the dispersion relation of SPs modes running on the metal surface. Therefore, this first paper already established a close connection between the extraordinary optical transmission (EOT) and the excitation of SPs. Since 1998, several experimental and theoretical groups around the world have reproduced the main features present in the first set of experiments. Additionally, the influence of the metal forming the structure, as well as the dependence of EOT with the lattice symmetry (square or triangular), hole shape (circular, elliptical, square or rectangular) and frequency regime (optical, THz or microwave) have been thoroughly analyzed [2-13]. Interestingly, it was found [14] that the optical transmission though a single aperture (a hole or a slit) could be also be enhanced, if the aperture is flanked by periodic corrugations on the side the light is impinging on. Moreover, it was also
16 16
L. Martín-Moreno Martin-Moreno et al.
found that very strong directional emission (beaming) is possible through single subwavelength apertures if the corrugation is placed on the exit side. In this paper, we summarize our theoretical results on the EOT (in both 2D hole arrays and single apertures) and beaming from single apertures, concentrating on the basic physics behind these phenomena. 2. 2D SUBWAVELENGTH HOLE ARRAYS In this section we address the physical origin of EOT in subwavelength hole arrays. Except for the simplest (highly symmetrical) geometries, the calculation of optical properties of metals is a notoriously difficult computational problem. This is so due to the different length scales involved, ranging from the skin depth (a few tens of nanometres) to the system dimensions (say, some tens of microns). A representation of the electromagnetic fields in real threedimensional space requires a large number of mesh points. Similarly, in Fourier space, a very large number of plane waves are needed. So, for most of the problems it is necessary the use of simplified models which, although do not provide exact solutions, quite often show clearly the basics of the phenomena. Here, we present results for one of those simplified models able to take into account the existence of surface plasmons. Let us briefly summarize the basic ingredients of our theoretical formalism, which was already describe in Ref. 3 for the case of a 2D array of square holes and in Ref. 15 for arrays of circular holes., In order to go beyond the perfect conductor approximation (which considers the dielectric function in the metal to be minus infinity), the dielectric response of the metal is taken into account in our formalism by considering surface impedance boundary conditions [16] (SIBC) on the metal-interfaces defining the metal film. Despite this assumption, the metal is treated as a perfect conductor in the metal walls defining the hole. This allows the expression of the electromagnetic (EM) wave field inside the hole in terms of the eigenmodes of the hole, which are known analytically for simple hole shapes (such as rectangular or circular) [17]. Such an approximation therefore neglects absorption by the metal walls surrounding the hole. One can expect this not to be a serious shortcoming, as the area of the "horizontal" metaldielectric interfaces (in which absorption is properly taken into account) is larger than the "vertical" ones, for the geometrical parameters typically analyzed in the experiments. However, assuming perfect conductor walls also neglects the penetration of the EM fields in the "vertical" walls. This is an important limitation as, in the optical regime, EM fields penetrate into the metal up to a distance mainly controlled by the skin depth of the metal (of the order of 10-20 nm for noble metals). We circumvent this deficiency by considering an (wavelength dependent) effective hole radius such that the propagation constant inside the hole defined by perfect conductor walls is equal to the one extracted
Theory of optical transmission through arrays of subwavelength apertures
17 17
from an exact calculation (considering the actual dielectric constant of the metal). Within this scheme, the calculation of the transmission properties of 2D hole arrays amounts to expanding the EM fields in terms of the Bloch EM modes in each spatial region (plane waves in vacuum regions and hole waveguide modes inside the holes), and obtaining the expansion coefficients by just matching appropriately the parallel components of the E- and if-fields at the two metal-dielectric interfaces. •
1
•
1
•
1
•
0,5
Transmittance
0,4 CD
o
H 0,3 0,3
-
E CO
H 0,2 °'2 0,1
7v
nn 0,0 400
A 1 \
^
VI
500
600
700
800
900
Wavelength (nm) Fig. 1. Total transmittance calculated for a 2D hole array (period of the array, d =750 nm and diameter of the circular holes, a — 280 nm) perforated in a silver film of thickness h — 320 nm.
Figure 1 renders the transmission spectra obtained with the method previously described, for the geometrical parameters corresponding to the experiment reported in Fig. 1 of Ref. 3. In Ref. 3 we considered square holes, but here the calculation is presented for circular holes. Clearly, our model is capturing the main features of the experimental spectrum and the position of the highest peak (located at around 780nm) is in reasonable agreement with the experimental data. However, the experimental peak is lower and broader than the one obtained in the calculations. As our calculation is performed for an infinite array, this difference with the experimental data could be indicative of the presence of disorder and/or finite size effects. In order to gain physical insight into this phenomenon, it is convenient to look for the minimal model showing EOT. In Fig. 2 we compare the result of the fully converged calculation (black curve) displayed in Fig. 1 with the one (red curve) obtained by just considering one eigenmode inside the hole (the TEn
18 18
L. Martín-Moreno Martin-Moreno et al.
mode of the circular waveguide, which is the least decaying evanescent mode). As clearly seen in the figure, considering more evanescent modes inside the hole just produces a very small blue shift (2 nm) of the transmission peaks, without altering the overall picture of the spectrum. Moreover, by neglecting the absorption in the metal film (in our calculations this is readily done by taking the imaginary part of the dielectric constant of silver equal to zero), we find (see blue curve) that the main effect of the absorption in this range of parameters is the reduction of transmitted light, although EOT is still present. Added to that, the existence of two peaks is more clearly revealed in the calculation without absorption. Another interesting feature of the transmittance spectra is the presence of a deep transmittance minimum. The inset of figure 2 renders the transmittance in logarithmic scale in order to stress the existence of this minimum (the so-called Wood's anomaly, see Ref. 1), whose origin will be discussed later on. 1,0 Transmittance
1
Transmittance
0,8 O
CO
0,6
"E w
1E-3
1E-6
1E-9
1E-12
760
780
800
820
840
Wavelength (nm)
0,4
E
0,2
0,0 770
780
790
800
810
820
Wavelength (nm) Fig, 2. Zero-order transmittance for the structure in Fig.l, obtained from a fully converged calculation (black curve), from only considering the TEn mode inside the holes (red curve) and from the same calculation as the red curve but assuming that no absorption is present in the metal (blue curve); this case is also displayed in the inset but in a logarithmic scale.
From now on in this section we are going to analyze the results of this minimal model (considering only TE n and setting Im [£(($] - 0). In order to unveil the physical mechanism responsible for EOT and to clearly show that EOT depends on modes of the "horizontal" metal-dielectric interfaces, we present a multiple scattering formalism for the computation of the transmittance. In this framework, transmission amplitudes for crossing the whole system are obtained from the scattering amplitudes for crossing the two different individual metal-dielectric interfaces, and the propagation constant of the fundamental (TEn) mode inside the hole (see Fig. 3).
Theory of optical transmission through arrays of subwavelength apertures
19 19
1 fAiri II (Holes) (Air)
T l
* t23
Fig. 3. Schematic drawing of the different scattering magnitudes at interfaces I-1I and II-III. See text for a detailed explanation of the different terms.
The zero-order transmission amplitude (to) can be expressed then as:
where rl2 and % are the transmission amplitudes for crossing the I-II and the IIIII interfaces, respectively. kg - (ka2- (l,84/a)2)1/2, where ko is the EM wavenumber in vacuum and pR and pL are, respectively, the amplitudes for the TE a mode to be reflected back into the hole at the II-III and II-I interfaces. These reflection amplitudes coincide (pR = pL - p), when the dielectric constant in the regions of reflection and transmission are equal, as in the symmetric structure we are considering. Figure 4 renders the modulus of T/2 and % as a function of the wavelength for the case of a 2D square array of circular holes with lattice parameter d— 750 nm and nominal radius a - 280 nm. The spectral dependence of the modulus of p, for the same set of parameters, is presented in the upper panel of Fig. 5. There are several interesting features appearing in these scattering magnitudes. Firstly, the three quantities present a maximum at around 785 nm. Moreover, \p \ » 1 at this resonant location. The counterintuitive result of a reflection amplitude larger than unity is due to the fact that the fundamental eigenmode inside the hole is evanescent, for which current conservation only restricts Im [p] > 0, with no restrictions on the real part (or the modulus) of this scattering magnitude.
20
Martin-Moreno et al. L. Martín-Moreno 10
1
0,1
⏐τ12⏐
0,01
⏐τ23⏐ 1E-3 1E-3
1E-4 1E-4 740
760
780
800
820
840
Wavelength (nm)
Fig. 4. Modulus of Xn and T» as a function of the wavelength for an air-silver interface, where the metal is perforated with a 2D array of circular holes with diameter a = 280 nm. The lattice parameter is d- 750 nm.
The reflection amplitude p is a causal function, and as such, it satisfies the Kramers-Kronig relations. The strong peak in the modulus of p comes from a peak in its imaginary part (see [3]), which signals the existence of a surface resonance (or surface leaky mode) of the perforated metal surface. Its spectral width is related to the time the EM field spends at the surface before it is either radiated or absorbed. This large reflection amplitude opens up the possibility of resonant denominator in Eq. (1) even for metal thicknesses such that e J '* z '*«l, Figure 5 illustrates graphically that the peaks appearing in the zero-order transmittance occur at the wavelengths for which the distance between \p\ and e'*2 is minimal. This figure unambiguously shows that EOT in 2D hole arrays has a resonant nature and that the origin of this resonant behaviour is the existence of SPs decorating the metal-dielectric interfaces. For thin films (h = 100-400 nm in Fig.7-5), the two curves intersect at two different wavelengths giving rise to the appearance of two transmission peaks in the spectrum. It can be shown that these two peaks correspond to the symmetric and anti-symmetric combinations of the two SPs of the two interfaces that are coupled through the evanescent fields inside the holes.
Theory of of optical optical transmission transmission through through arrays arrays of of subwavelength apertures apertures Theory
21 21
35
(iii) 30
⏐ρ⏐
25 20 15
(ii) (iii)
10 5
(i)
Transmittance
3
2
1
(iii) h=500nm h = 500 nm (v)
(ii) h = 300 nm (iii) h=300nm
(i) h=100nm 0
740
760 760
780
800
820
840
(nm) Wavelength (nm)
Fig. 5. Upper panel: modulus of p and curves e'fe'* for different values of h (100, 300 and 500 nm) for the same geometrical parameters than in the previous figures. Bottom panel: zeroorder transmittance versus wavelength for the silver thicknesses considered in the upper panel.
For a range of metal thicknesses, these two coupled surface modes are able to transfer energy through the structure very efficiently, even 100% if no absorption were present in the system. When h is further increased, there is no crossing between the two curves and only one peak with associated transmittance less than 100% remains in the spectrum. This occurs because, for large h, the coupling of the surface modes through the evanescent field inside the hole is weaker and it consequently requires the EM field's spending more time inside the hole in order to build the resonance. When this time is larger than the typical radiation time (which can be extracted from the width of the reflection amplitude of a single interface), the EM field is radiated before it has time to "feel" the presence of two coupled modes. Even in that situation, there still exists a peak that reflects the EM field's longer stays at the surface (therefore enhancing the probability for tunnelling across the hole), but the mechanism now resembles more of sequential (although still coherent) tunnelling, where high orders in the multiple scattering mechanism inside the hole are not important. As commented above, the location of this peak coincides with the location of the SP at parallel momentum 2n/d of the silver surface perforated with a 2D array of holes. A detailed discussion of the formation of the coupled surface modes and the typical times in the transmission process can be found in Ref. 3.
22
L. Martín-Moreno Martin-Moreno et al.
An additional feature appearing in Fig. 4 is that both |Xi2| and fe | have a zero located at around 765 nm that moves into a minimum in the zero-order transmittance theoretical spectrum (see Fig. 2), the so-called Wood's anomaly. The reader is warned that the location of this zero does not coincide with the location of the Rayleigh minimum, which occurs when a propagating diffracted wave becomes evanescent (in this particular case this occurs at 750nm). On the contrary, it can be shown analytically that the location of the minima in both It^l and |x23| coincide with the location of the SP at parallel momentum 2n/d of the plain (without holes) silver surface. This has been the origin of some criticism [18] of the SPs as the origin of EOT. As we have stated here and already shown in Ref. 3, our calculations show that EOT is mediated by the SPs, but those corresponding to the structured metal surface. Once EOT is explained in the optical regime in terms of the excitation of SPS, the question of the transferability of EOT to other frequency regimes naturally arises. In Ref. 3 we showed that EOT phenomenon also appears even in a perfect conductor film perforated with a 2D array of holes. A more extensive theoretical analysis of the existence of EOT in perfect conductors can be found in Ref. 15. This result seems, at first sight, in contradiction to our previous claims, given that flat perfect-conductor interfaces do not posses SP modes. However, surface EM modes appear in corrugated perfect conductors and, in particular, in perfect conductors perforated with 2D hole arrays. Very recently, we have shown that these surface modes resemble those of a real metal, and are responsible for the existence of EOT in perfect conductors [19]. Therefore, EOT seems to be a more general phenomenon that will appear in any electromagnetic structure in which surface EM modes are present and can couple to radiative modes. This hypothesis has been verified for metals in the THz [12] and microwave regimes [13] and even for dielectric photonic crystal waveguides [20]. 3. EOT IN SINGLE APERTURES FLANKED BY CORRUGATIONS As discussed in the previous section, surface EM modes are at the origin of the EOT phenomenon. In hole arrays the main ingredients for observing EOT are: i) the existence of a surface EM mode and ii) the presence of a grating coupler that allows the incident light to interact with the surface mode. This suggested that perhaps it was possible to obtain EOT also in single apertures, if they were surrounded by a finite periodic array of indentations. This hypothesis was experimentally verified in Ref. 14 both for a ID slit surrounded by a finite array of grooves and for the bull's eye geometry (a 2D circular hole flanked by circular trenches). Here we present the theoretical foundation of this phenomenon for the ID case.
Theory of optical transmission through arrays of subwavelength apertures
23 23
Let us summarize the formalism used to calculate the transmission of light through a single slit of width a symmetrically flanked by a finite array (with period d) grooves. Although the formalism can deal with more general conditions, we restrict ourselves here to the case where there are 2N grooves placed symmetrically with respect to the slit, and of illumination by a normal incident p-polarized plane. The slit has width a, while the grooves have width a and depth w (see Fig. 6). The theoretical formalism we have developed is a nontrivial extension to finite structures of the framework previously used for analyzing 2D hole arrays. First we assume an artificial supercell with cell parameter L that includes the finite set of indentations we are considering. Then we express the EM-fields in different regions in terms of their mode expansion. In vacuum we expand the fields by a set of plane waves whereas for the grooves and central slit only the fundamental propagating eigenmode is considered. That is, inside indentation a, Ex is a linear combination of (/>a{x)e±kz, where k - 2n/A and (/>a(x) - dm inside the indentation and zero outside. The fields are matched appropriately on all interfaces (as in the case of 2D hole arrays, we apply SIBC in the horizontal interfaces while perfect metal boundary conditions are assumed in the vertical ones).
~WU u
U I
_«rL_n__n_l Fig. 6. Schematic drawing of the structure analyzed in this section: a single slit of width a, symmetrically flanked by 2N grooves of width a and depth w in either input and output surfaces (or both). The metal thickness is h, and a normal incident p-polarized plane wave is considered.
Finally, the limit L —> °° is taken analytically, eliminating the dependence on the artificial lattice parameter, which leads to a set of linear equations for the unknowns {Ea,E\}: [ G . -ea]Ea + YG^Efi-SaaGrEa
= Ia
where or and ^runs over all indentations (slit or grooves). The set {Ea} are the modal amplitudes of the x-component of the electric field right at the indentations in the input surface: Ex (x, z- 0+) = E a E a <pa(x) whereas the set
24
L. Martín-Moreno Martin-Moreno et al.
{E\} are the modal amplitudes of the jc-component of the electric field at the output surface; E% {x, z - h~) = Tiy"E\ ^(x). The different terms appearing in these equations have a clear physical interpretation (we present here its values for the perfect conductor case, its expressions when SIBC are applied are also analytical but slightly more involved). Ia derives from the direct initial illumination on indentation a, being essentially the overlap integral between the incident /3-polarized plane wave and wavefield $*, In this structure, as we are considering the metal as completely opaque, the two metal interfaces are only connected through the central slit by the term Gy = \lsin (kh). £& takes into account that the EM fields at one opening of a given indentation can bounce back (many times) at the other end of that indentation, and has the value: Ea=cot (kw) for grooves (or^O) and e$ = cot(kh) for the slit. The term Gap represents the EM coupling between indentations. It takes into account that each point in the indentation J3 emits radiation that can be collected by indentation or. Mathematically, G^ is the projection onto wavefields $a and $g of the Green's function G (r, r). It can be shown that this Green function contains the contribution of both diffraction modes and the SP channel. In general, it has to be computed numerically although in the case of perfect conductors its expression is known analytically to be G - (iTtfX)Hom(k\r-r% Hom being the 0order Hankel function of the first kind. Once the values for {Ea, £"Y} a r e calculated the normalized-to-area transmittance can be obtained from T=GvIm Figure 7 renders the calculated the dependence with number of grooves placed in the input side of the normalized-to-area transmittance T(X) for a single slit. The geometrical parameters chosen are typical experimental values used in the optical regime (a = 120 nm, d = 600 nm, w = 100 nm and h = 350 nm). The curve for JV=O (black curve) corresponds to the single slit case; in this frequency range, the spectrum presents two broad peaks that correspond to the excitation of slit waveguide modes inside the central slit [21]. As the number of indentations increases, a maximum in T(A) develops at ku- 760 nm. For this set of geometrical values and for the metal considered, maximum in T\A) saturates at about N= 5-10, when T is enhanced by a factor close to 5. With respect to the output corrugation, we have demonstrated in previous works (see Ref. 22) that it has little effect on the total transmittance. From the set of equations (2), it is possible to identify the different mechanisms that help to enhance the transmission of light through the central slit. Assuming that the slit is flanked by the grooves only at the input surface, the two equations governing {£Oi£"o} are:
25
Theory of optical transmission through arrays of subwavelength subwavelength apertures
Normalized transmittance
5
03
[0, 0] [1, 0] [5, 0] [10, 0] [15, 0]
4
o
£
3
2
-a N
2
1
0 400
600
800
1000
1200
Wavelength (nm) Fig. 7. Normalized to area transmittance spectra for a single slit of width a — 120 nm surrounded by 2N grooves (N ranging from 0 to 15) located symmetrically with respect to the central slit. The grooves are placed only in the input surface, while the output surface is not corrugated. The period of the array is d = 600 nm, the width of the grooves is also 120 nm and their depth is 100 nm. The calculation is done for a silver film, with thickness of 350 nm.
[Gw - et]Et + £ G t e £ e
= /„
(3)
[Gm-£0]Eo-GvEQ=0 In the single slit case, Eo- 2(Goo- £Q)ID and £"0 = 2GJD, where the denominator D = (GQO - £b)2-Gy2. Slit waveguide modes correspond to minima in D, leading to transmission resonances. Corrugating the input surface opens up the possibility of obtaining large Eo by having a large £«. The equation for Ea shows that its magnitude can be large if (Gaa- So)*3 0, which is the condition of the excitation of a groove cavity mode (as can be more clearly seen by analyzing the a —> 0 limit). However, an even larger Eo can be obtained if, additionally to Ea being large, the illumination coming from the different grooves reaches the central slit in phase. The phase in this re-illumination process is controlled by GOa- An estimation of when this in-phase re-illumination occurs can be done from asymptotic expression of Hom(x)*=e'kx. Therefore, it can be expected that all light re-emitted from the grooves will interfere constructively on the other grooves and the central slit for X ~ d. On the other hand, the existence of a phase
26
L. Martín-Moreno Martin-Moreno et al.
shift in the asymptotic expression of the Hankel function, and also presence of the SP channel in Goa modifies this condition (as can be seen in Fig. 7). Actually, due to these factors, it might be possible that the best transmission enhancement occurs for a non-uniform array, a point that deserves further investigation. The combination of the two mechanisms described above (groove cavity mode and in-phase groove re-emission) is responsible for the peak located at around 755nm visible in Fig. 7.
0
Wavelength (nm)
1200
3,0
6,0
9,0
12
(a)
0
4,5
9,0
13
18
(b)
1200
1000
1000
800
800
600
600
400
400 50
100
150 150
200
Depth of the grooves (nm)
250
300
50
100
150 150
200
250
300
Depth of the grooves (nm)
Fig. 8. Normalized-to-area transmittance versus both wavelength and depth of the grooves, for a single slit of width 120 nm symmetrically flanked by 20 grooves located in the input surface. The thickness of the metal film is 350 nm and the period of the array is 600 nm, as in the previous figure. Panel (a) shows the result for silver, assuming SIBC in the horizontal interfaces of the structure whereas panel (b) shows the results for a film of perfect conductor.
Figure 8a illustrates the presence of the three mechanisms previously described for enhancing the transmittance through a single slit. It renders T versus both X and depth of the grooves w, for the geometrical parameters a — 120 nm, h = 350 nm, d - 600 nm and N = 10. Figure 8a also shows that when two mechanisms coincide there is an additional boost in the transmittance. For small w, maximum transmittance appears close to the X = d condition. It can be shown that this line corresponds to the excitation of a surface EM mode, originated by the interplay between the groove cavity modes and the in-phase groove re-emission mechanisms. This surface mode has strong similarities with the one responsible for EOT in periodic apertures. In Fig. 8b we present the results for the same set of parameters, but obtained within the perfect conductor approximation. The similarities between the results obtained in these two cases
Theory of optical transmission through arrays of subwavelength apertures
27
reinforces the conclusion that the main ingredients of the EOT phenomenon in 2D hole arrays and in single apertures is already present in corrugated perfect conductor surfaces. As the results obtained within the perfect conductor approximation are scalable to other frequency regimes, our results also apply to the enhanced transmission through single apertures flanked by corrugations appearing in the microwave and millimeter regime [23,24]. 4. BEAMING OF LIGHT IN SINGLE APERTURES As previously stated, it was experimentally found [14] that the angular distribution of the transmitted radiation through single apertures in corrugated metal surfaces presents a very small angular divergence at some resonant wavelengths, and that this angular distribution is basically controlled by the output corrugation. With the theoretical framework described in the previous section it is possible to calculate the wavefield in all spatial regions and, therefore, the angular distribution of emitted light. 8 7
N=1 N=2 N=5 N = 10 N = 15
rSr (θ, N ) / rSr (θ, N= 0)
6
o
FF
ii
5 4 3
FF
•£-
2 1 0 = -80
-60
-40
-20
0 0
20 20
40
60
80
θ8 (deg)
Fig. 9. Radial component of the Poynting vector evaluated in the far field versus angle for a single slit of width c = 120nm surrounded symmetrically by 2JV" grooves (iV ranging; from 1 1 to 15) of width 120 nm and depth 100 nm. Wavelength of flie incident radiation is 760Inm. n x ig. z/, xvauicu vuixipuiiviii ui nil/ x ujutiLiK v w i u i bvaiumbu iix uib leu iiviu v u a u o ai
Figure 9 shows the calculated radial component of the Poynting vector ST (Q), in the far-field region and normalized to the total transmittance for a single slit surrounded symmetrically by 2N grooves in the output surface. Several
28
L. Martín-Moreno Martin-Moreno et al.
values of N are presented (from 1 to 15), for the resonant wavelength XM— 760 nm. Note that this resonant wavelength is the same as the one found for EOT in a single slit flanked by a finite array of grooves in the input surface for the same set of geometrical parameters. This fact clearly shows that the origin of the beaming effect is the same as the EOT in single apertures surrounded by periodic corrugations: the excitation of a surface EM mode in the output surface. The system behaves as a diffraction grating, illuminated from the central slit by the surface mode. The illumination of the grooves decays with the distance to the central slit, as the wave radiates as it propagates along the surface. Details about the formation of this surface mode and its relation with the radiation pattern mode can be found in Ref. 25, as well as the difference in the illumination of the grooves between the resonant (i.e. when a surface leaky mode is formed) and non-resonant cases.
REFERENCES [I] T. W. Ebbesen, H. J. Lezec, H. F. Ghaemi, T. Thio, and P. A. Wolff, Nature, 391 (1998) 667. [2] H. F. Ghaemi, T. Thio, D. E. Grapp, T. W, Ebbesen, and H.J. Lezec, Phys. Rev. B, 58 (1998) 6779. [3] L. Martin-Moreno, F. J. Garcia-Vidal, H. J. Lezec, K. M. Pellerin, T. Thio, J. B. Pendry, and T. W. Ebbesen, Phys. Rev. Lett., 86 (2001) 1114. [4] L. Salomon, F. D. Grillot, A. V. Zayats, and F. de Fomel, Phys. Rev. Lett., 86 (2001) 1110. [5] A. Krishnan, T. Thio, T. J. Kima, H. J. Lezec, T. W. Ebbesen, P. A. Wolff, J. Pendry, L. Martin-Moreno, and F. J. Garcia-Vidal, Opt. Commun., 200 (2001) 1. [6] A. Degiron, H. J. Lezec, W. L. Barnes, and T. W. Ebbesen, Appl. Phys. Lett., 81 (2002) 4327. [7] N. Bonod, S. Enoch, L. Li, E. Popov and M. Neviere, Opt. Express, 11 (2003) 482. [8] C. Genet, M. P. van Exter, and J.P. Woerdman, Opt. Comm., 225 (2003) 331. [9] W. L. Barnes, W. A. Murray, J. Ditinger, E. Devaux, and T.W. Ebbesen, Phys. Rev. Lett., 92(2004)107401. [10] R. Gordon, A. G. Brolo, A. McKinnon, A. Rajora, B. Leathern, K. L. Kavanagh, Phys. Rev. Lett., 92 (2004) 37401. II1] K. J. Klein Koerkamp, S. Enoch, F. B. Segerink, N. F. van Hulst, and L Kuipers, Phys. Rev. Lett., 92 (2004) 183901. [12] J. G6mez-Rivas, C. Schotsch, P. Haring Bolivar, and H. Kurz, Phys. Rev. B, 68 (2003), 201306. [13] M. Beruete, M. Sorolla, M. Campillo, J. S. Dolado, L. Martin-Moreno, J. Bravo Abad, and F. J. Garcfa-Vidal, Opt. Lett., 29 (2004) 2500. [14] H. J. Lezec, A. Degiron, E. Devaux, R. A. Linke, L. Martin-Moreno, F. J. Garcia Vidal, and T.W. Ebbesen, Science, 297 (2002) 820. [15] L. Martin-Moreno and F. J. Garcia-Vidal, Opt. Express, 12 (2004) 3619. [16] J. D. Jackson, Classical Electrodynamics, 2nd ed., Wiley, New York, 1975. [17] P.M. Morse and H. Feshbach, Methods of theoretical physics, McGraw-Hill, New York 1953.
Theory of optical transmission through arrays of subwavelength apertures
29 29
[18] Q. Cao and P. Lalanne, Phys. Rev. Lett., 88 (2002) 57403. [19] J. B. Pendry, L. Martin-Moreno, and FJ. Garcia-Vidal, Science, 305 (2004) 847. [20] E. Moreno, F. J. Garcia-Vidal, and L. Martin-Moreno, Phys. Rev. B, 69 (2004) 121402. [21] J. A. Porto, F. J. Garcia-Vidal, and J. B. Pendry, Phys. Rev. Lett., 83 (1999) 2845. [22] F. J. Garcia-Vidal, H. J. Lezec, T. W. Ebbesen, and L. Martin-Moreno, Phys. Rev. Lett., 90(2003)213901. [23] M. J. Lockyear, A. P. Hibbins, J. R. Sambles, and C. R. Lawrence, Appl. Phys. Lett., 84 (2004) 2040. [24] S. S. Akarca-Biyikli, I. Buhl, and E. Ozbay, Appl. Phys. Lett., 85 (2004) 1098. [25] L. Martin-Moreno, F. J. Garcia-Vidal, H. J. Lezec, A. Degiron, and T. W. Ebbesen, Phys. Rev. Lett., 90 (2003)167401.
This Page is Intentionally Left Blank
Handai Nanophotonics, Volume 2 S. Kawata and H. Masuhara (Editors) © 2006 Elsevier B.V. All rights reserved.
31 31
Chapter 3
Linear and nonlinear optical response of concentric metallic nanoshells M. Fukui, T. Okamoto and M. Haraguchi The University of Tokushima, Faculty of Engineering, Department of Optical Science and Technology, 2-1, Minamijosanjima-cho, Tokushima 770-8506, Japan 1. INTRODUCTION Surface plasmon polaritons are electromagnetic modes with a locally enhanced electric field. These modes are expected to become the key for the development of photonics of the 21st century and thus the applications of surface plasmon polaritons have become a worldwide target to be studied. In particular, localized surface plasmons (LSP's) have been widely studied as a key electromagnetic mode to develop nano-photonic technology [1-12]. One of the benefits of mastering the physical properties of the LSP comes from its application to nonlinear optics. We have however little knowledge of nonlinear optical response of the LSP, although it is quite significant to develop optical devices, e.g. an optical switch, based on nonlinear optical phenomena due to the LSP excitation. We have succeeded to present the phenomena of optical switching and optical bistability by using a modified nonlinear Mie theory for a Ag nano-sphere coated with a CdS film [13]. This simulation takes a long time and is achieved along a radius axis. It is therefore not easy to understand circumstances over the whole of the concentric shell. To overcome this disadvantage, the nonlinear optical response of a Ag sphere coated with CdS films is demonstrated using finite-difference time domain (FDTD) simulations. In this chapter, we will firstly introduce numerical results of the nonlinear optical response of a Ag sphere coated with CdS films. Secondly, the fabrication process of a Ag sphere coated with CdS films will be described. Thirdly, the experimental results of the linear and the nonlinear optical responses of samples fabricated will be given.
32
M. Fukui, T. Okamoto and M. Haraguchi
2. NUMERICAL METHOD FOR A LINEAR OPTICAL RESPONSE: MIE THEORY TREATMENT By employing the spherical coordinates shown in Fig. 1, for the plane wave incidence, the electromagnetic fields inside and outside of a sphere, E and H, are expressed as [14]
Fig. 1. Spherical coordinates.
(1) (2)
Here, "T
i
,
.
,
sin
zd(kJR)P*(coaff) cos sm
,
. i
02-zd(kJR)kJRdR
(3)
sin
sm (4)
cos Here, M=mexp(-jcof) and N=nexp(-/Gtf) are the vector spherical wave functions which are the solution of the Maxwell equations in Mie theory. The superscripts/, which are (0), (1) and (2), denotes the core sphere, the film coating the sphere and the surrounding medium, respectively. The superscripts d - i and d = b imply the ingoing (from the outside to the inside of the sphere) and outgoing waves, respectively. Note that the electromagnetic fields in the respective media are expressed by overlapping the fields given by Eqs.l and 2, and kj denotes the wave number of waves in thejth layer, n (=1,2,3,•••) is the mode number. The
Linear and nonlinear optical response of concentric metallic nanoshells
33
functions znd(kR) are the spherical Bessel functions for d ~ f and the spherical Hankel functions for d = b, respectively. The function P,,1 is the associated Legendre function. We consider a«(2) and 6n(2)f to be one and EQ is the amplitude of the electric field of the incident light. Note that a«(0)f and 6B(0)f are zero. ajd and bjd are determined from the Maxwell boundary conditions. The normalized scattering cross section Ws is expressed as
E (5)
where a is the radius of the sphere and e and ju have the respective usual meanings. 2.1. Linear optical responses Figure 2 shows the geometry for calculations. In order to clearly investigate the physical properties of LSP's, we employ SiO2 as a coating film because it is an optical loss-free material in the visible range. We adopted Ag as a metal sphere with a radius of a. (a) lncident light
Air or Dielectric Fig. 2. Geometry of calculations: (a) single metal sphere, (b) single metal sphere coated by a dielectric or a Kerr-nonlinear film, a: radius of metal sphere, h: thickness of coating film. A and A' denote the observation points of the light intensity.
As is well known, the dielectric function of metal spheres depends upon their sizes when the sphere size becomes smaller than the electron mean free path. We employed the expression developed by Kreibig and Vollmer [15] as
34
M. Fukui, T. Okamoto and M. Haraguchi
1 2
co +T{a)2 0)Z
T(a)
of+T{af where F«, = vp/X« and F(a) = vF/a, Here,
(6) ((Op: plasma angular
frequency) = 9.52 eV, vF (Fermi velocity) = 1.39x106 m s"1 , and X«, (electron mean free path) = 52 nm. The dielectric constant obtained from Johnson and Christy [16] was adopted as the bulk dielectric constant £"buik(co) of Ag. It is natural that the imaginary part of e{(a,a) is more strongly affected by the size of Ag spheres, compared with the real part of e(to,a). The dielectric function of SiO2 was evaluated from the dispersion equation of a fused silica presented by Malitson [17]. 2.1.1. Physical properties of LSP's excited in a single Ag sphere Before discussing LSP's in concentric metallic shells, it may be instructive to introduce significant physical properties of LSP's in a Ag single sphere.
3
O
•f <1>
3
g ( X
hay
500)
( eV )
Fig. 3. Normalized scattering cross section of a single Ag sphere in air. n is the mode number of LSP's.
Linear and nonlinear optical response of concentric metallic nanoshells
35
Figure 3 shows the spectra of the cross section of scattering for a single Ag sphere in air. The peaks in Fig. 3 are produced by the excitation of LSP's. Note that the respective peaks give a strong electric field in the proximity of Ag sphere surfaces. From Fig. 3, it is confirmed that the peaks move to the lower energy side with increasing a and the two peaks appear in the radius range of more than 50 nm. The spatial profiles of electric fields, |2?R|, \E£ and \E£, projected onto the x-z plane are given for the peak at a = 20 nm (/zoo = 3.435 eV, n-l), as shown in Fig. 4. Here the x-polarized incident light propagates toward the z-direction. We assumed the amplitude of the incident plane wave to be one. The results clearly indicate that |£R| at point A takes a maximum value. As a consequence, we focus on a normalized light intensity enhancement factor ]iS'A|2/j-£*o|2s which is equal to |£A| 2 because £0 was assumed to be one, at peaks of the spectra of the cross section of scattering in subsequent discussions. The dependence of lisAp/lisol2 on a is shown for Ag spheres in air and Ag spheres in SiC>2 in Fig. 5. The maximum values of |.£'A|2/|-£'O|2 are 287 for a — 20 nm (Ag/air) and 730 for a — 12 nm (Ag/SiOa). The reason for such maximum values is due to the fact that the effect of light confinement becomes stronger, while the optical loss becomes larger with decreasing a. Namely, the position of the peak should be determined from the mediation between the two effects, the light confinement and the optical loss. The maximum value of |.EA|2/|-EO|Z for Ag/SiC>2 is approximately 2.5 times larger than that for Ag/air. This magnification originates from the difference in the respective dielectric constants of air and SiO2.
E Rl
'0
Fig. 4. Spatial distribution of the electric field of a single Ag sphere, (a), (b) and (c) represent |£R|, \EQ\ and \EV\, respectively, a - 20 nm, ftco = 3.435 eV.
36
M. Fukui, T. T. Okamoto and and M. Haraguchi M.
800
20 30 a ( ntn) Fig. 5. Relative light intensity of a single Ag sphere, |£A|2/|£O|2, as a function of the sphere radius a. The solid and dashed lines indicate the single Ag sphere in air and the single Ag sphere coated with SiC>2, respectively. The incident photon energy is at the peak of the cross section of scattering.
10
20
30
a(am) Fig. 6. Relative light intensity of a single Ag sphere coated with the SiC>2 film, \EA,xf/\Eof, as a function of the Ag sphere radius a. h is fixed at 20nm. The solid and open circles indicate |£A|2/|£o|2 at point A on the surface of the SiO2 film and |i?A-l2/|£o|2 at point A' on the Ag-SiO2 interface, respectively. The incident photon energy is at the peak of the cross section of scattering.
2.1.2. Light intensity enhancement of a Ag sphere coated with a SiO2film Consider the Ag sphere coated with S1O2. The dependence of \EA\ /\E0\2 and |i?A>|2/|iio|2 on a in the case of h = 20 nm, where h is the thickness of SiC>2 film, is shown in Fig. 6. \EA\2/\Eof and lisA'p/l-Eof take maxima due to the origin presented in 2-1-1. The values of a for these maxima are larger than that given in Fig. 5. The reason may be as follows. In the case of [£rA*|2/]-fi'o]25 the Ag sphere is not only surrounded by SiOz but also by air. |ZsA|2/|iso|2 increases with
Linear and and nonlinear optical response of concentric metallic nanoshells
37
increasing in a, as indicated in §3-2. \EA'\ /|2?o| takes a maximum at a = 15 nm. For 0 = 1 5 nm, we evaluated |i?A|2/|.Zro|a and \E^\2/\E0\2 with varying h, as shown in Fig. 7. l^p/l^ol 2 gradually increases and approaches a certain value with increasing h. This may be because the configuration is changed from Ag/air to Ag/SiO 2 with increasing h. Since the light intensity in the SiO2 film decays with distance from the surface of the Ag sphere, |2?A| l\Eaf decreases with increasing in h. 800 Ag-SiO; film interface (A1)
600
Ei
200
10
20 h ( nm)
30
40
Fig. 7. Relative light intensity of a single Ag sphere coated with the SiC>2 film, as a function of the SiCh film thickness h. a is fixed at 15nm. The solid and open circles indicate |-EA|2/|£O|2 at point A on the surface of the SiO2 film and |£A'|24£b|2 at point A' on the Ag-SiC>2 interface, respectively. The incident photon energy is at the peak of the cross section of scattering.
2.2. Nonlinear optical responses Figure 2(b) shows the geometry for calculations. We adopted Ag as a metal and CdS as a Kerr-nonlinear film. The geometry is the same as that employed in [13,18]. The calculation procedure was also carried out along the line presented in [13,18]. The optical Kerr effect is given by the relationship (7) Here, &o is the wave number of light in vacuum, EK the linear-specific dielectric constant, a the nonlinear coefficient and \E\ the amplitude of the electric field. The values of E\ determined by Gottesman and Ferguson were used as the dielectric constant of CdS [19]. The nonlinear coefficient,, a in [20], was assumed to be 10"15 m2 V"2.
38
M. Fukui, T. Okamoto and M. Haraguchi
2.1
2.2
2.3
2.4
2.5
fit!) ( CV )
Fig. 8. Normalized scattering cross section of a single Ag sphere coated with the CdS fihn. a - 20 nm, h — 20 nm. The peak corresponds to the « = 1 mode of the LSP.
^ 10
2.206eV
2.191eV
10 I, (kW/mm 2 )
20
Fig. 9. Nonlinear optical response of a single Ag sphere coated with the CdS fikn. a = 20 nm, h = 20 nm. The solid, dash-dotted and dashed lines indicate ha> = 2.191 eV, 2.206 eV and 2.222 eV, respectively.
The normalized scattering cross section for the linear response is shown as a function of the incident photon energy in Fig. 8. The peak at hat = 2.305 eV is due to the excitation of the LSP with n = 1. The light intensity is strongest at point A outside the sphere in Fig. 2(b). In Fig. 9, the normalized light intensity |^A]2/|£"o|2» where EAi$ the sum of the electric field of the ingoing and outgoing waves at point A outside the sphere, is shown as a function of the incident light intensity I\ which is proportional to |£0|2- At ft© = 2.191 eV that is smaller than 2.305 eV corresponding to the excitation of the LSP with n - 1, an optical bistability phenomenon occurs. As ftco is increased, the incident light
Linear and and nonlinear optical response of concentric metallic nanoshells
39
intensity at which the optical bistability starts is decreased, but the incident light intensity region where the optical bistability occurs becomes narrower. Finally, at Tito = 2.206 eV, the optical bistability turns into optical switching. We define the incident light intensity required for the optical switching as the critical intensity Ic, Ie being approximately 13.5 kW mm"2 in this case. Moreover, the normalized switching intensity, /sw> is defined to be the difference between the maximum values of \EA\2/\E0\2 for the non-linear response and |Z?A|2/|£o|2 for the linear response. In the present case, /sw = HA. 2.2.1. Dependence of optical switching characteristics on the radii of Ag spheres The thickness of CdS film, h, is fixed at 20 nm. Figure 10 shows |£'A|2/|-E'O|2 VS. I-, for various radii of Ag spheres. Note that h(Q was chosen such that it was equal to Ic for various values of a because h(& corresponding to the excitation of LSP's varies with changing a. At a = 20 nm, Ic becomes minimum. /Sw increases with increasing a. In order to interpret these two points, we evaluated \EAyA>\2/\Ea\2 vs. a for the linear response, as shown in Fig. 11. The maximum of |J£'A»J2/|JE'ol2 arises from the mediation of Ag size-dependent optical confinement and Ag sizedependent optical loss, as already described. As shown in Fig. 12, the electric field is decreased with increasing R. The variation of the refractive index due to the Kerr effect should, therefore, be largest at point A'. As a consequence, the larger the |isA'|2/|£o|2 becomes, the larger the nonlinear effect also becomes. Finally, since |i?A'|2/|£o|2 takes a maximum value at a = 20 nm, / c becomes minimum there.
10" Ii (kW/mm 2 ) Fig. 10. The a-dependence of the nonlinear optical response of a single Ag sphere coated with the CdS film, h is fixed at 20nm. The incident photon energy is set at the value required for the optical switching.
40
M. Fukui, T. Okamoto and M. Haraguchi
100
so v Ag-CdS
film interface (A1)
1 | 60 CdS film surface (A)
5 40 20 0 0
Fig. 11. Relative light intensity of a single Ag sphere coated with the CdS film, as a function of the Ag sphere radius a. The incident photon energy is at the peak of the cross section of scattering. The solid and open circles indicate |£A|2/|£O| a* point A on the surface of the CdS film and \Exf/\Eo\2 at point A' on the Ag-CdS interface, respectively.
A' i
100 -
Ag
CdS
Air
80
1 ^40
-
20 0, 0
20
40 R (nm)
60
Fig. 12. Spatial distribution of light intensity along the radius axis of a single Ag sphere coated with the CdS film. Note that the figure shows the spatial distribution along the radius axis passing points A and A' through the center of the sphere, a - 20 nm, h — 20 nm and tim = 2.305eV.
The reason why /Sw increases with increasing a is presented as follows. The optical switching occurs at h(£> less than that for the peak of the normalized scattering cross section, so that Ifi'Ap/l^'ol2 in Fig- 10 is smaller than those in Fig, 11. Note, however, that the characteristic of I-EAP/I-EOI2 VS. a in Fig. 10 is almost the same as that in Fig. 11. Secondly, the maximum of \EA\ /|2?o]2 vs. a in Fig. 10 is also almost the same as that of I^AI 2 /!^! 2 VS. a in Fig. 11. With increasing a, Iu which gives the
Linear and and nonlinear optical response of of concentric metallic nanoshells
41
maximum of |£rA|2/|JE'o|2* is increased. Consequently, an optical self-focusing effect appears, and thus |£ A | 2 /|£o| 2 in Fig. 10 becomes larger than that in Fig. 11.
102 1, (kW/min ) Fig. 13. The A-dependence of the nonlinear optical response of a single Ag sphere coated with the CdS film, a is fixed at 20 nm. The incident photon energy is set at the value required for the optical switching.
200 Ag-CdS film interface (A1)
.5-100
30 40 A (run) Fig. 14. Relative light intensity of a single Ag sphere coated with the CdS film, |£A,A-|2/]£O|2> as a function of CdS film thickness h. The incident photon energy is at the peak of the cross section of scattering. The solid and open circles indicate |£A| 2 /|£OI at point A on the surface of the CdS film and \EA'\2f\Eo\2 at point A' on the Ag-CdS interface, respectively.
2.2.2. Dependence of optical switching characteristics on the thickness of a CdS coatingfilm The radius of Ag sphere, a, is fixed at 20 nm, giving the minimum value of/c
42
M. Fukui, T. Okamoto and M. Haraguchi
in Fig, 10. Figure 13 shows \EA\2/\E0\2 VS. 7J for various thicknesses of CdS films. It is apparent that 7C and 7SW decrease with increasing h. Definite optical switching disappears in the h- range beyond 20 nm. In order to interpret such results, we evaluated 'o| vs. h for the linear response at points A and A', as shown in Fig. 14. |£A.| /\EO\ increases with increasing h, so that 7G decreases. On the other hand, |7jA|2/|7io|2 decreases with increasing h and, in addition, the value of 7j which gives a maximum of \EA\2/\Eo\2 decreases. As a consequence, 7SW decreases. 2.2.3. Mechanism of the optical switching explored from the spatial distribution of a light intensity In order to explore the mechanism of the optical switching, we have evaluated the spatial distribution of the light intensity before and after the switching. Setting the incident light intensity before the switching, as shown by mark (1) in Fig, 15, we obtain a usual profile of the distribution, as indicated in Fig.l6(a). For the incident light intensity just before the switching, as shown by mark (2) in Fig. 15, the localization of light becomes stronger because of a self-focussing effect due to the third-order nonlinearity of the CdS film, as shown in Fig.l6(b). For just after the switching, see mark (3) in Fig. 15, light confinement around a central line becomes extremely intense, as shown in Fig.l6(c). As is well known, the LSP with n=l is a Frolich mode in a linear optical region, so that the spatial distribution of the light intensity should be independent of the radius. 30 j
to 2 206eV 20 H
\
'
j
—-~-
\
1
(2)
\
is.
\
(4)
10
n
i
10 ( kW/tnm 2 )
20
Fig. 15. Nonlinear optical switching of a single Ag sphere coated with the CdS film, a — 20 nm, h - 20 nm, hco = 2.206 eV.
Linear and nonlinear optical response of concentric metallic nanoshells
(a)
(b)
(c)
43
(d)
Fig. 16. Spatial distributions of \E 12/|£0|2
The nonlinear effect of CdS, however, leads to a large deformation of the spatial distribution, as indicated in Fig.l6(c). After the switching (mark(4)), the area of strong light intensity is of a fan-type, as shown in Fig.l6(d). Such an area is further expanded and thus light confinement becomes weaker with increasing incident light intensity, so that light intensity along the central line is decreased. This is the mechanism of the occurrence of the switching understood from the spatial distribution of the light intensity. 3. NONNLINEAR OPTICAL RESPONCES EVALUATED BY A FDTD METHOD As known, the Finite-Difference Time-Domain (FDTD) method [21] is well suited for simulations of LSP resonances on nano metal particles. The FDTD method can solve Maxwell's equations for complex geometries containing nonlinear materials and give a transient solution. In LSP resonance simulations, however, nonphysical artifacts often appear for resonant light fields around the surfaces of the particles. Such artifacts may come from the difference between the numerical and the real surfaces of the particle which arises from the employment of cubic cells in FDTD calculations. In such conventional FDTD simulations, it is therefore not possible to exactly reproduce the curvature of surfaces, hi simulations for nonlinear phenomena, the artifact will cause serious numerical errors because the nonlinear effect is quite sensitive to the intensity of the light field. In order to avoid nonphysical artifacts, we can employ another type of grid instead of the orthogonal grid, e.g., an irregular unstructured grid using general space filling polyhedral cells, a nonorthogonal grid expressed by spherical (or
44
M. Fukui, T. Okamoto and M. Haraguchi
cylindrical) coordinates and a fine grid using a local subcell with contour-path modeling [21]. The FDTD method with the unstructured grid can be applied to simulations for complex geometries and is expected to give a drastic improvement for numerical error caused by the nonphysical artifacts discussed above. However, using the unstructured gird, we need a complex algorithm for FDTD calculations of 3D structures and extensive computational resources. The FDTD algorithm with spherical coordinates can be applied to geometries containing a single spherical particle and is expected to give the smallest numerical error among the different grids. It is, however, difficult to apply for geometries except a single particle. For the fine grid, the FDTD algorithm is simple and is applicable for geometries containing several particles. It may not be, however, easy to decrease numerical error although it will be improved compared with the case of the FDTD method with the orthogonal grid. Our final purpose is to investigate the nonlinear optical response of a metal sphere, coated with a Kerr material, related to LSP resonances by using FDTD simulations. As the first step towards that, we aim to develop the two-dimensional FDTD program to evaluate the nonlinear optical response of the metal cylinder coated with a Kerr material. 3.1. Procedure of calculations We employed the two-dimensional FDTD method taking into account the nonlinear dispersive optical response given by Taflove [21]. Figure 17(a) shows the simulation structure: a metal cylinder coated with a Kerr material is located in vacuum surrounded by the PML absorbing boundaries. In order to compare the accuracy of the field calculation, three approaches were employed. The first is to use the cylindrical coordinates. The second is to employ the local subcell as the FDTD cells at the boundary between the metal and the Kerr material. The third is to employ the standard squared cell coordinates. The calculation geometry for both, i.e. the standard square cell and the local subcell was the same except around the boundary region between the metal and the Kerr material, as shown in Fig.l7(b). The TM-polarized incident light was launched from the left side of the cylinder. The diameter of the cylinder and the thickness of the coated layer were set at 40 nm and 20 nm, respectively. As shown in Fig. 17, we focus our concern on the observation points 1 and 2 positioned at 2 nm away from the surface of the metal cylinder and at 2 nm away from the surface of the Kerr material. For the standard square cell and for the local subcell, size of the calculation area and size of a single cell were 300 nm x 300 nm and 1 nm x 1 nm, respectively. For cylindrical coordinates, the form of the cell is of fan-type and the circle diameter of the calculation area, the size of the cell along the radius and the angle dividing along the rotational directions are 240nm, 1 nm and 1 degree, respectively. The time step was 1.4 x 10'3 fs for all simulations.
Linear and nonlinear optical response of concentric metallic nanoshells
PML 8 layers PML 16 layers \ light sourceligs^observotion point 2 light puree line \ ^observation point 1 /
/
J
y
45
observation point 2 / / observation point 2 // ^ ^ x ^
\s< Ken material
300 am-
(b)
(a) Fig. 17. Numerical configuration for FDTD calculations.
We assumed that the dielectric constant of the metal was expressed by the Drude model. The plasma and the collision frequencies are 1.965 x 1015 rad s"1 and 1.433 x 10° rad s'1, respectively. These parameters give us the same LSP resonance frequency as the experimental one of the nano Ag sphere with a diameter of 40 nm. As a Kerr material, we employed a material with the dielectric constant expressed by a single Lorentz function, where the parameters were evaluated at a resonance frequency of 3.09 x 1015 rad s"1, a collision frequency of 2.24 x 1013 s"1, a static dielectric constant of 4.69 and a high frequency dielectric constant of 4.47. These parameters correspond to the dielectric constant of CdS in the wavelength range from 400nm to 600nm. The expression of the Kerr nonlinearity, as given by Taflove [21], was employed in the algorithm. The third order nonlinear part of the electric polarization ¥SL(xy,f) responding to the electric field E(x,y,t) is expressed as P NL (x,j,O = -
-T)[E(x,y,T)]dT
(9)
(3) where £Q, XO (3) ^ d g(t) are the dielectric constant of vacuum, the third order nonlinear susceptibility and the response function. Considering the resonant and nonresonant responses of ¥HL(xy,t), g(f) can be written as
(10) where ft and SJ) are the weight parameter and the Dirac delta function, respectively. gR(?) is given by the function with the step function U{t),
46
M. Fukui, T. Okamoto and M. Haraguchi
where l/T\ and 1/% are the resonant frequency and the bandwidth of a single Lorentzian response for the Kerr nonlinearity. We set nonlinear optical parameters at j ^ 3 ) = 1.0 x 1(T15 m2 V 2 , fi= 0.7, tx = 1.62 fs and t2 = 139 fs.
3.2. Numerical results Figure 18 shows the normalized electric field observed at the observation point 1 (see Fig. 17) as a function of wavelength. Thin solid, dashed and thick solid lines are for the square cells, the local subcells and the fan cells, respectively. There is a resonant peak in each spectrum, which is caused by the LSP excitation. For the square cells, i.e. the thin solid line, the peak makes 10 nm-red shift, compared with the other two cells. The peak intensity for the thick solid line, i.e. the fan cell, is highest of all. The difference between the highest and the lowest peak intensities is about 15 %, and this difference cannot be ignored because it will give a large influence on nonlinear simulation results. In order to explore the origin of such a difference, we calculate the electric field intensity distribution at the respective resonance. Figure 19 shows the electric field intensity distribution temporally averaged over one time-period of the incident light at the resonance wavelength, A = 400 nm in a linear response region, (a), (b) and (c) are for the standard squared cell, for the local subcells and for the fan cells, respectively. The outer and inner white dashed circles in (a) and (b) mean the vacuum-Kerr material interface and the Kerr material-metal interface, respectively. The white and black colors mean the maximum and zero intensities, respectively. In order to clearly observe the field intensity pattern, the maximum of the intensity scale in Fig. 19 is set at 75% of the maximum intensity calculated. Outside the metal cylinder, there should be two regions with a higher intensity, induced by the LSP excitation, and such regions would be localized in proximity of the cylinder, as predicted from results derived for metal spheres [15]. In Fig. 19 (a) and (b), however, we find higher intensity regions and many bright spots along the metal surface. Increasing the size of cells, then the intensity at the spots becomes stronger. The spot size in Fig. 19 (b) is smaller than that in Fig. 19 (a). In Fig. 19 (b), note that the intensity in the metal region is larger than that in the surrounding dielectric region. On the other hand, there are no such spots in Fig. 19 (c). The intensity in the metal cylinder is large and the gray area with a moderate intensity exists in the vacuum.
Linear and nonlinear optical response of concentric metallic nanoshells
Al 47
-4
300
400 500 wavelength [nm]
600
Fig. 18. Normalized electric field intensity observed at the observation point 1. £b is the electric field of the incident light. Thin solid, dashed and thick solid lines are for the square cells and for the local subcells and for fan cells, respectively.
(a)
(b)
(c)
Fig. 19. Spatial profiles of the electric field intensity at X = 400 nm under the linear response condition, (a), (b) and (c) are for the square cells and for the local subcells and for the fan cells, respectively.
Judging from the result that the spot intensity increases with the increase in the cell size, the spots in Figs. 19 (a) and (b) may be expected to arise from the employment of a square cell. The electric field associated with LSP's is localized around metal surfaces. The numerical error may be, therefore, enhanced by such a square cell. Especially, around the corners of the square cells located at the boundary between the metal and the dielectric material the light intensity tends to be spuriously enhanced. From Fig. 19 (a) and (b), spurious distributions of a light intensity at the LSP excitation can be clearly improved with the local subcell method, but not enough to simulate nonlinear responses.
48
M. Fukui, T. Okamoto and M. Haraguchi
In the following, we have calculated the dependence of the electric field intensity on the incident light intensity at the observation point 2 at X - 460 nm, as shown in Fig. 20. The dashed and solid lines are for the local subcell and for the fan cells. We can confirm that the solid line shows an optical bistability in the incident intensity range from 4.05 W Jim"2 to 5.95 W um"2. On the other hand, the dashed line indicates no bistability but only the nonlinear response in the intensity region from 4.5 W |xm"2 to 5.6 W |im"2. Moreover, in the intensity region beyond 5.6 W |lrn"2 in the case of the subcell, we were not able to obtain any stable numerical results. As shown in Fig. 19(b), four spurious spots may not give any accurate nonlinear responses in calculating numerically. 1
1
'
1
•i _ /f
y
/
^
\ -
1
4
5 Incident intensity [W/jtLm'J
6
Fig.20. Electric field intensity at the observation point 2 indicated in Fig. 17 as a function of the incident light intensity at k — 46Qnm.
Fig. 21. Spatial profiles of the electric field intensity at X = 460 nm under the bistability condition obtained for the fan cells. The incident intensity is 5.7W |im' 2 . (a) and (b) are for the lower and the higher intensity states marked by the thick arrows in Fig. 4, respectively.
Linear and nonlinear optical response of concentric metallic nanoshells
49
In Fig. 20, the on/off ratio of the bistability is about 2.5. This value is too small. This may come from a cylindrical structure. Confirmed from Fig. 18, the field enhancement due to the excitation of LSP is small. This is because the electric field associated with LSPs in circular cylindrical metals is not confined along the central axis. To overcome this drawback, we assumed %® to be quite large. This implies that a macroscopic nonlinear polarization exists over the whole of the Kerr-material even when the intensity of the incident light is small. Namely, it may not be good to conclude that the optical bistability presented in Fig. 20 is due to the excitation of LSPs. On the other hand, the field enhancement by a LSP in a sphere is considerably larger than that for a cylinder. The story of optical bistability should be then changed dramatically. Figure 21 shows the images of the electric field intensity in the bistable state obtain by using the fan cells at an incident intensity of 5.7 W urn"2 at X = 460 m a (a) and (b) are for the lower and the higher intensity states marked by the thick arrows in Fig. 20, respectively. The scale of the intensity is indicated by the black (zero intensity )-white (maximum intensity) color. The maximum of the intensity scale in Fig. 21 was again set at 75 % of the maximum intensity calculated. In Fig. 21 (a) and (b),the position of the region with a high intensity differs from each other, i.e. outside of the Kerr material in Fig. 21 (a) and inside of the metal cylinder in Fig. 21(b). In the lower intensity state at 460 nm, the dielectric constant of the Kerr material doesn't satisfy the LSP resonance condition. In consequence, no bright regions existed inside the Kerr material and the metal cylinder, as shown in Fig. 21 (a). On the other hand, for Fig. 21(b), two bright regions exist in the metal cylinder and the intensity of the electric field in the Kerr material is much lower than those in the bright spots. The bright regions may suggest that a kind of the resonance occurs. The origin of the resonance, however, is not clear, as mentioned already. The dielectric constant of the Kerr material is increased with increasing the incident light. In consequence, the intensity of the electric field in the Kerr material decreases because of its enhanced dielectric constant. As the first step, we succeeded to develop the two-dimensional FDTD method to be capable of analyzing a Kerr-nonlinear effect of metallic nanoshells, Through this treatment, we can have information about circumstances over the whole of nanoshells in a nonlinear optical region and transient phenomena of nonlinear optical responses. This is quite useful in developing optical nano-devices required. The next direction is to develop a three-dimensional FDTD technique to analyze nonlinear optical responses due to LSP excitation.
50
M. Fukui, T. Okamoto and M. Haraguchi
4. FABRICATION AND LINEAR OPTICAL RESPONSE OF AG PARTICLES COATED WITH CDS FILMS In the following, we shift our discussions to experiments on Ag particles coated with CdS. Ag particles coated with CdS have not, to our knowledge, been fabricated yet. Samples have been fabricated here by employing the so-called reversed micelle technique. The reversed micelle method is a technique to synthesize a particle in a hydrophilic solvent surrounded by a surfactant dispersed in a hydrophobic solvent (reversed micelle constructions) [22]. A typical recipe is summarized as follows. Firstly, the microemulsion solution with Ag particles is prepared by mixing 1.43ml non-ionic surfactant Igepal CO-520, 3.57ml cyclohexane and an amount of 0.30 m£ of 10"2 mol X'1 AgNOa solution. Ag particles have been synthesized in a water-in-oil microemulsion system of AgNO3 solution / Igepal / cyclohexane. After that, as a coating process, 10"2 mol X1 Cd(NO3)2 and 10"2 mol XT1 Na^S solutions were added to the microemulsion sequentially. We finally obtain a microemulsion containing Ag particles coated with CdS. Figure 22 shows the extinction spectra for the microemulsion containing only Ag particles (dashed line) and that containing Ag particles coated with CdS (solid lines). The coating process was repeated three times. The peak at 417nm for the dashed line is due to the LSP excitation. This peak shifts toward a longer wavelength side as increasing the thickness of CdS films, i.e. increasing the step-number of the coating process, and the extinction coefficient decreases prominently. Note that the increase in the extinction coefficient in a short wavelength region may come from pure CdS particles themselves. We have compared experimental extinction spectra with theoretical ones obtained from the Mie theory. In comparison, the size distribution of Ag particles has been assumed to be of a Gaussian type. We employed the size-dependent dielectric constant of Ag particles and the wavelength-dependent dielectric constant of the CdS film in calculations. Finally, we have evaluated the size of Ag particles to be (40±16)nm and the thickness of CdS film to be (l±4)nm. In the case of the thickness of CdS film, the standard deviation is large and thus Ag particles coated with CdS having a thickness of ~10nm should be contained. We may assure this prediction from an increase in the extinction coefficient in the wavelength range from 500nm to 600nm, as shown in Fig. 22. Nevertheless, since the best thickness of CdS film required for an optical switching is around 20nm, it is necessary to fabricate thicker CdS. To know the constituents of Ag particles coated with CdS, we have observed a SEM image and an energy dispersive X-ray (EDX) spectrum. Figure 23 shows the SEM image for a sample which was heated for one hour at 300 degree to resolve a surfactant after dropping the microemulsion having Ag particles coated with CdS onto an Al substrate. White dots correspond to Ag particles coated with CdS. It is confirmed from this SEM image that the average diameter of Ag particles coated with CdS is roughly 40nm. From an EDX analysis, particles fabricated here mainly consist of Ag and the thickness of CdS
Linear and nonlinear optical response of concentric metallic nanoshells
51
is expected to be too thin to confirm the existence of CdS. This supports the result obtained from the extinction spectrum that the thickness of CdS film is (l±4)nm.
1
Before
/
-
\ \
X)
*. Attei t;oaung process •
o0.5 .
O 1 si slop A 2nd sijcp X 3rdslcp
\ ^
W n
V<
)0
i
1
400
500
600
X (nm)
Fig.22. Extinction spectra of the microemulsion.
Fig.23. SEM image of the Ag particles coated with CdS.
5. OPTICAL NONLINEAR RESPONSE OF AG PARTICLES COATED WITH CDS Figure 24 shows the experimental setup for observing a nonlinear optical response. The two lights emitted from a Q-switched YAG laser are brought through the two paths. Light with a wavelength of 355nm is focussed in the air, so that a charge dissociation occurs, leading to the white light emission. This light is employed as the probe light. The other light having a wavelength of 532nm, a pulse width of 7ns and a repetition rate of lOHz is used as the
52
M. Fukui, T. Okamoto and M. Haraguchi
pumping light. In this system, tuning the timing of illumination between the pumping light and the probe one, we can have readily information about an extinction spectrum for nonlinear optical responses. For the Ag microemulsion, the illumination of the pumping light did not lead to any distortion of the extinction spectrum, implying that no nonlinear effect occurs. On the other hand, for the microemulsion of Ag particles coated with CdS, the extinction spectrum did not change before and after the pumping whose intensity was 57MW mm'2. However, during the pumping, the extinction signal prominently decreases, as shown in Fig. 25. This implies that this decrease arises from a change in the refractive index of CdS. Note that a rapid decrease around 532nm in Fig. 25 is due to a stray light of the pumping light. To explore the origin of the change in the refractive index of CdS, we calculated the dependence of the extinction spectra on the refractive index of CdS by employing the Mie theory, as shown in Fig. 26. The size distribution of Ag particles and the distribution of the thickness of CdS have been assumed to be of a Gaussian type. From the extinction spectrum without the pumping light, the size of Ag particles and the thickness of CdS were determined. The refractive index of CdS under the illumination of the pumping light was assumed to be «o(>l) + O.lno(A) where n0 is the linear refractive index depending upon the wavelength,/!* The expression of 0.1 «o(^) corresponds to the nonlinear refractive index due to the third-order nonlinear effect. With the increase in the refractive index, the peak value is decreased and the full width at half maximum
Q=swNd:YAG laser Light emission due to the discharge of the ait* Probe light (White light)
Sample Ag/CdS. or A.g
"Pump light (3i=532 nm)
\ I
r\ Multi-channel \ i Spectrometer Fig. 24. Experimental setup for observation of nonlinear optical responses.
Linear and nonlinear optical response of concentric metallic nanoshells
^^
53
, unr
Ho I ore On Pumping
% c
g
§0.5 \
aX
i
w i
ii
i
400
i
i
,
500 A (nm)
•
i
600
Fig. 25. Nonlinear extinction spectra of the microemulsion. The solid line denotes the extinction coefficient during the illumination of the pumping light. The dashed line indicates the extinction coefficient before and after the pumping light.
400 A
500 (nm)
600
Fig. 26. The calculated extinction spectra of the microemulsion. The refractive index of Ag was evaluated from the expression developed by Kreibig and Vollmer [15]. Johnson and Christy [16] was adopted as the bulk refractive index of Ag. The values of no determined by Cardona and Greenaway were used as the refractive index of CdS [23].
is increased in the extinction spectrum. Judging from this result, we can definitely conclude that the refractive index of CdS coating Ag particles has been increased through an optical Kerr-effect produced by illuminating the pumping light. Since we employed the pumping light with a pulse width of 7ns, it may not be possible to avoid the occurrence of a thermal effect, i.e. a decrease in the refractive index generated from a temperature rise. Namely, the optical Kerr-effect and the thermal effect cancel each other. The result presented in Fig. 25 should be, therefore carefully interpreted. In order to observe a prominent
54
M. Fukui, T. Okamoto and M. Haraguchi
optical switching phenomenon, we need CdS films having a thickness of more than lOnm. REFERENCES [1] R, Hillenbrand, F. Keilmann, P. Hanarp, D. S. Sutherland, and J. Aizpurua, Appl. Phys. Lett., 83 (2003) 368. [2] W. Rechberger, A. Hohenau, A. Leitner, J. R. Krenn, B. Lamprecht, and F. R. Aussenegg, Opt. Commun., 220 (2003) 137. [3] E. Prodan, C. Radloff, N. J. Halas, and P. Nordlander, Science, 302 (2003) 419. [4] H. Kuwata, H. Tamara, K. Esumi, and K. Miyano, Appl. Phys. Lett., 83 (2003) 4625. [5] K. Tanaka and M. Tanaka, J. Appl. Phys., 95 (2004) 3765. [6] H. Tamara, H. Kuwata, H. T. Miyazaki, and K. Miyano, Appl. Phys. Lett., 80 (2002) 1826. [7] Y. Niidome, H. Takahashi, T. Kawasawa, A. Hori, and S. Yamada, Jpn. J. Appl. Phys., 42 (2003) 7640. [8] S. A. Maier, P. G. Kik, and H. A. Atwater, Phys. Rev., B67 (2003) 205402. [9] N. Felidj, J. Abhard, G. Levi, J. R. Krenn, A. Hohenau, G. Schider, B. A. Leitner, and F. R. Aussenegg, Appl. Phys. Lett., 82 (2003) 3095. [10]T. Okamoto, I. Yamaguchi, and T. Kobayashi, Opt. Lett., 25 (2000) 372. [11]K. Tanaka, H. Hosaka, K. Itao, M. Oumi, T. Niwa, T. Miyatani, Y. Mitsuoka, K. Nakajima, and T. Ohkubo, Appl. Phys. Lett., 83 (2003) 1083. [12] S. Yamada, T. Tasaki, T. Akiyama, N. Terasaki, and S. Nitahara, Thin Solid Film, 438 (2003) 70. [13]T. Okamoto, M. Haraguchi, and M. Fukui, Jpn. J. Appl. Phys., 43 (2004) 6507. [14] J. A. Stratton, Electromagnetic Theory, McGraw-Hill, New York, 1941, p. 392. [15]U. KreibigandM. Vollmer, Optical Properties of Metal Crusters, Springer, Berlin, 1995. [16]P. B. Johnson and R. W. Christy, Phys. Rev. B, 6 (1972) 4370. [17] I. H. Malitson, J. Opt. Soc. Am., 55 (1965) 1205. [18]T. Okamoto, M. Haraguchi, and M. Fukui, J. Microsc, 210 (2003) 193. [19] J. Gottesman and W. F. C. Ferguson, J. Opt. Soc. Am., 44 (1954) 368. [20]T. Okamoto, M. Haraguchi, and M. Fukui: J, Microsc, 210 (2003) 193. [21] A. Taflov, Computational Electrodynamics, Artech House, Boston, 1995. [22] J. H.Adair, T. Li, T. Kido, K,Havey, J. Moon, J. Mecholsky, A. Morrone, D. R. Talham, M. H. Ludwig and L. Wang, Mater. Sci. Engr., R23 (1998) 139. [23] M. Cardona and D. L. Greenaway, Phys. Rev., 131 (1963) 98.
Nanophotonics, Volume 2 Handai Nanophotonics, S. Kawata and H. Masuhara (Editors) © 2006 Elsevier B.V. All rights reserved. reserved.
55 55
Chapter 4
Low-dimensional optical waveguides and wavenumber surface J, Takahara" and T. Kobayashi* "Graduate School of Engineering Science, Osaka University, Toyonaka, Osaka 560-8531, Japan 1. INTRODUCTION Diffraction is a natural phenomenon exhibited by optical waves. Diffraction limits the resolving power of optical instruments or minimum spot size of focused optical beams. Diffraction also limits the size of optical beams guided along dielectric optical waveguides. We cannot obtain nanometer-sized optical beam beyond the diffraction limit by using dielectric optical waveguides. Metal waveguides are widely known and has been extensively studied for a long time. Metal strip lines and rectangular metal waveguides have been used for microwave transmission. In a current terminology, metal waveguides at optical frequency are those known as negative dielectric waveguides. Today, metallic nanostructures with negative dielectric are considered to be key devices in subwavelength optical technology [1]. In negative dielectric waveguides, surface plasmon polariton (SPP) plays important roles. The physics of SPP has been investigated for planar metal interfaces [2—4] and for cylindrical metal interfaces [5,6] since 1960's. Recent progress of SNOM has attracted much interest in these systems again [7]. Such research area is today called "plasmonics", which is growing rapidly as a crucial part of nanophotonics. Technical potentials of plasmonics are not only limited to microscopy, but are also finding use in a wide range of applications from novel light emitting devices to nanometer-sized optical circuits. High-density optical switching devices constructed by nano-optical circuits are needed to satisfy the increasing demands of optical communication [8]. New optical waveguides using photonic crystals or high refractive index materials have been proposed for decreasing the size of dielectric optical waveguides. The size of these new waveguides, however, is still restricted by the diffraction limit. Hence, we need a new approach to achieve optical waveguides of nanometer scale (1 to lOnm).
56
J. Takahara and T. Kobayashi
In negative dielectric waveguides, spatial field of optical waves decays exponentially unlike conventional dielectric waveguides. We can model this type of waves using the concept of low-dimensional optical waves [9,10]. We have studied propagation modes in one-dimensional and two-dimensional optical waveguides and shown that the beam size of TM mode can be squeezed to nanometer-order in negative dielectric rod [9,10], hole [10-12], tube and coaxial line [10,12] and gap structure [12,13]. In order to reveal the physical meaning of mode calculations, we shall introduce wavenumber surface and discuss the diffraction limit in geometrical way. In this chapter we shall describe nano-optical waveguides breaking through the diffraction limit of light. First, the concept of low-dimensional optical waves and wavenumber surface are introduced in wavenumber space. Then, the diffraction limit is discussed from the point of view of wavenumber surface. Mode analyses of two-dimensional and one-dimensional optical waveguides are presented and the confinements of optical beams in nanostructures are discussed. Propagation loss and excitation methods of low-dimensional optical waves are also discussed for applications to nano-optical devices. 2. LOW-DIMENSIONAL OPTICAL WAVES AND WAVENUMBER SURFACE In this section, we define low-dimensional optical wave, n-k space and wavenumber surface. We describe the diffraction limit by using geometry of the wavenumber surfaces in «-k space. 2.1. Dimensions of optical waves Let us consider a monochromatic scalar field £/(?) in lossless homogeneous media. The spatial distribution of the field satisfies the Helmholtz equation; V 2 t/(r)+P[/(r) = 0
(1)
The solution of Eq. (1) is expressed in the form of a plane wave as e'kT, where k is the wavenumber vector. Because any optical waves can be expressed as a superposition of plane waves, an optical beam can be synthesized by a superposition of plane waves. Let us define the dimension of optical waves. We define the dimension of optical waves as the number of real number elements in k. An optical wave having three real number elements in k is defined as a three-dimensional (3D) optical wave. Optical waves having 2 and 1 real number elements in |-(the other elements are pure imaginary numbers) are defined as a two-dimensional (2D) and one-dimensional (ID) optical wave, respectively. An optical wave that
Low-dimensional optical waveguides and wavenumber surface
57
has no real number elements in k (all the elements are pure imaginary numbers) is defined as a zero-dimensional (OD) optical wave. We define that 2D, ID, and OD optical waves are low-dimensional optical waves. We summarize the dimension of optical waves in Table 1, where i (i? = -1) is the imaginary unit. Table 1 The dimension of optical waves and the elements of wavenumber vector. n-k space dimension elements of k 3 3-k space (k
2
(kt,ky,iKt)
k
k "S
(ft,,*,,*,)
2-k space
1
1-k space
0
0-k space
kj, Kj (/=x,y,z) are real numbers.
For example, a light propagating in vacuum or homogeneous dielectric is 3D optical waves. The dimension of optical waves defined above is not always the same with geometrical shapes of media: a light confined in a core of a dielectric optical fiber is not ID but 3D optical waves in spite of the ID shape of the fiber. A light in a dielectric sphere known as a whispering gallery mode is not OD but 3D optical waves although the shape of a sphere is OD. The typical example of 2D optical waves is evanescent waves generated by total internal reflection (TIR) on a planar dielectric interface. 2.2. Wavenumber surface of 3D optical waves in dielectric We consider a plane wave of angular frequency co in a homogeneous dielectric medium with refractive index n. The dispersion relation of light is
\krn7
(2)
where | k \ — nko is the wavenumber in the medium, k® — m/c is wavenumber in vacuum and c is the speed of light in vacuum. The elements of wavenumber vector in Cartesian coordinates, k= (£» ky, kz), must satisfy the dispersion relation (Eq. (2)) as follows:
k2 = k] + kl + k2z = (nkB)2 =
(3)
58
J. Takahara and T. Kobayashi
where n = 4e-J~ji, £ and fi are the relative permittivity and the permeability of the medium, respectively. Since we have considered a monochromatic field, we can define a normalized wavenumber vector k = k I k0 = (kx, kv, k2) by fixing ko as: ~kx=kjka~ky=kylka,kz=kjk<)
(4)
Substituting Eq. (4) into Eq. (3), one can obtain the normalized dispersion relation; .2
.2
_2
(5)
kx+ky+kz = ~ ky
~ ky
1
~y κ
-4 4
-1 1
~ κz -4
-1
-4 -4
-4
~ kx
~ kx
4
-4 4
~ κz
~ kz -1
4
~ kx
4
4
1
(a)
(b)
(c)
Fig. 1. Wavenumber surfaces in dielectric: (a) 3D optical wave {kx+ky+kz =1) (b) 2Doptical _2
-2
_2
-2
_2
_2
wave (£*+&,.-K"z = l ) ( c ) ID optical wave ( k X—Ky—Kz =1).
In dielectric materials(£> 0 and ju> 0), Eq. (5) represents a sphere of radius of n = -Jefi in the normalized wavenumber space as shown in Fig. 2.2(a). Mathematically speaking, the surface of a sphere is 2-dimensional manifold in 3-dimensional Euclidean space i?3. In this case k moves on "the surface" of the sphere in i?3 given by the normalized wavenumber elements k*, kv and k2. Here we introduce the term "3-k space" for simplicity instead of 3-dimensional Euclidean space given by the normalized wavenumber elements. Such surface of the sphere is the geometrical representation of the dispersion relation infixedCO. Here we name such surface defined in 3-k space "wavenumber surface". 2.3. Wavenumber surface of low-dimensional optical waves in dielectric Next we consider low-dimensional optical waves in dielectric (D). If only one element of k (e.g. kz) is an imaginary number and the other elements are
Low-dimensional optical waveguides and wavenumber surface
59 59
real numbers, the dimension of such optical wave is 2D according to the _2
_2
_2
_
definition. Such optical wave must satisfy kI+ky>kz. Taking kz =iKz in Eq. (5), the normalized dispersion relation of 2D optical waves is: kx+ky-Kz =en, eju>0
(6)
where KZ -Kjk0 and KZ is an attenuation coefficient in D. Here we define new normalized wavenumber space instead of 3-k space. On treating wavenumber surfaces of 2D optical waves, it is natural to introduce the normalized wavenumber space of (kx,ky,Ks) named "2-k space". Eq. (6) represents a one-sheeted circular hyperboloid (CH) in 2-k space. This is shown in Fig. 2.2(b). Furthermore, if two elements of k (e.g. ky and kz) are imaginary numbers, the dimension of such optical wave is ID. Such optical wave must satisfy kx>ky+kz. Taking ky=iKy, kz=iKz in Eq. (5), the normalized dispersion relation of ID optical wave is, ~2
_2
_2
kt-Ky-K^siLi,
e/i>0
(7)
where K> =Kylkf) and Ky is an attenuation coefficient in D. Similarly, on treating wavenumber surfaces of ID optical waves, it is natural to introduce normalized wavenumber space of (kx,icr,icz) named "1-k space". Eq. (7) represents a two-sheeted CH in 1-k space. This is shown in Fig. 2.2(c). These wavenumber surfaces shown in Fig. 2.2(b) and (c) are the variation of quadratic surfaces. 2.4. Wavenumber surface in negative dielectric or negative permeability Let us consider optical waves in homogeneous materials with negative permittivity (e < 0) or negative permeability (fl < 0). Here we call materials with e < 0 and fi > 0 "negative dielectric (ND)", and materials with e > 0 and fi< 0 "negative permeability media (NP)". The dimension of the optical waves is low-dimensional in ND or NP, hence the arguments in section 2.2 and 2.3 are changed. We shall discuss the wavenumber surfaces of 2D,1D and 0D optical waves below. If £ju < 0, at lease one element of k must be imaginary numbers from Eq. (5). Taking k* =ix2 in Eq. (5), the normalized dispersion relation of 2D optical waves is,
60 60
J.Takahara Takahara and Kobayashi J. and T. T. Kobayashi
~ ky
4
~ κy
-4 4
~ κy
4 -1
-4 4
~ κz
1
~z κ
~ κz -4
-1
-4 -4
-4 -4
~ kx
~ kkxx
1
4
-1
~ κx 1
4
(c) (c) (b) (b) Fig. 2. Wavenumber surfaces in ND {£ < 0, // > 0) or NP (£> 0, ju <0): (a) 2D optical (a) (a)
~2
^2
_2
_2
_2
_2
wave( kx+ky-Kz = -1 ) (b) ID optical wave (^-X"y-/fz = - 1 ) (c) OD optical wave _2
_2
_2
As shown in Fig. 2,4(a), the wavenumber surface of 2D optical waves in ND or NP is a two-sheeted CH in 2-k space from Eq. (8). Taking ky =itcy, k2 =/r 3 in Eq. (5), the normalized dispersion relation of ID optical wave is, £fl
(9)
From Eq. (9), the wavenumber surface of ID optical waves in ND or NP is a one-sheeted CH in 1-k space as shown in Fig. 2.4(b). Furthermore, one can obtain the normalized dispersion relation of OD optical waves by taking kx=iKx, ky = iKy, k: = iKz inEq. (5) as: (10) In treating the dispersion relation of OD optical waves, it is natural to introduce normalized wavenumber space of (Kx,icy,icz) named "0-k space". From Eq. (10), the wavenumber surface of OD optical waves in 0-k space is a sphere of radius of J-qi under the condition of e/i < 0. This is shown in Fig. 2.4(c). We shall not discuss OD optical waves here, because the physical meaning is still unclear. 2.5. The diffraction limit and wavenumber surface Let us discuss the origin of the diffraction limit of a monochromatic optical
61
Low-dimensional optical waveguides and wavenumber surface
beam. An optical beam can be synthesized by many plane waves propagating in different directions, i.e. 3D optical waves having different k. As shown in Fig. 2.2(a), the wavenumber surface of 3D optical waves is a sphere of radius n in 3-k space. Since all the elements of k are real numbers, each element kj (/=x,y,z) takes values as -k<,kj<,k . This means that the extent of the wavenumber space Ak is 2| k | ( M = 2«&b). According to the uncertainty relation of Fourier transforms, Ak and the extent of space Ar satisfy ArM £ %. Therefore, Ar satisfies the following inequality: Ar>^ = ^ M An
(11) '
K
where A® = 2?ifk0 is wavelength in vacuum. Hence, Ar takes a minimum given as:
*•-.«=£•
(12)
An
Eq. (11) means that the minimum size of a beam synthesized by 3D optical waves is limited to the order of AQ. We stress that this is the origin of the diffraction limit and is inevitable as long as the dimension of optical waves is 3D. It is worth noting that the dimension of optical waves is 3D in conventional dielectric waveguides or even in PC waveguides or high refractive index waveguides. Thus, the size of optical beams in these waveguides is limited to
Armin3D. As we mentioned in the previous subsections, the wavenumber surfaces change from a sphere to CHs in the case of ID or 2D optical waves. The geometrical difference of a CH from a sphere is that the surface of a CH is open: the surface spreads infinitely at k -» •». Because there is no limitation in Ak in CH, we have now a possibility to increase Ak larger than 2nko. If Ak is larger than 2j ft | (that is, M > 2nka), we can take Ar < ArminiD. Thus, we can use ID or 2D optical waves in order to break through the diffraction limit of 3D optical waves. We point out the analogy between wavenumber surface and Fermi surface that is familiar in solid state physics. In Fermiology, it is known that the topology of Fermi surface, e.g. open or closed trajectory in k space, plays important roles in electric conductivity in magnetic field. In similar sense, we stress that the topology of wavenumber surface plays important roles in the diffraction limit. Although homogeneous media themselves allow 2D and ID optical waves to exist, we note that 2D and ID optical waves are physically meaningful in a
62
J. Takahara and T. Kobayashi
half-space or in a limited region. This is because the field intensity of low-dimensional optical waves diverges in considering the whole space. This is different from 3D optical waves that are meaningful in a whole space. Since there is no special direction in homogeneous media, we need to introduce a boundary along which ID and 2D optical waves propagate. Otherwise, in the whole space of homogeneous media, ID and 2D optical waves are prohibited to propagate due to the divergence of the field intensity. 3. TWO-DIMENSIONAL OPTICAL WAVEGUIDES In this section, we describe 2D optical waves in planar D/D and D/ND interface (in this chapter, D/D represents the interface of two Ds, and D/ND represents the interface between D and ND). These planar interfaces are 2D optical waveguides. We describe propagation properties of 2D optical waveguides in simple manner by using the geometry of wavenumber surface. 3.1. A planar D/D interface 2D optical waves are physically meaningful in a half-space as described in section 2.5. A planar dielectric interface is the actual system where 2D optical wave is meaningful. We consider a planar D/D interface as shown in Fig. 3(a): an interface (xy-plane, z = 0) between D half-spaces, R\ with % n- \{z < 0) and R2 with &i, fi — 1 (z > 0). At the interface, we can excite 2D optical wave as an evanescent wave by TIR of 3D optical wave. Wavenumber surfaces in Rr and R% are a sphere and a one-sheeted CH, respectively. Fig. 3 (b) shows these surfaces in the same coordinate system: 3D optical waves in 3-k space and 2D optical waves in 2-k space. Because the projection of the wavenumber vector to the interface must conserve between R\ and Rj due to the law of the conservation of wavenumber, ks on the wavenumber surfaces must coincide just on the interface. This is equivalent to the law of the conservation of momentum, which is derived from translational invariance of space. In k space, the law of the conservation of wavenumber means that the projection of fe to ^^-plane coincide each other. The projection and matching in fc^-plane can be expressed as a cylindrical surface in Fig. 3 (b). The intersections of wavenumber surfaces and the cylindrical surface are circles, which we name "a wavenumber circle". The radius of the sphere limits the maximum radius of the wavenumber circle. We stress that this is the geometrical representation of the diffraction limit of 2D optical waves excited by TIR.
63
Low-dimensional optical waveguides and wavenumber surface
~ ky
2 D ε £2 D
R22
x y
3D
D D
ε£1 1 (ε 1>ε2>0) (e1>e2>0)
(a) (a)
-3 3
z
2D
3
~ kz -3 -3
R Ri1
κ~z sphere in 3-k space
~k ^ ^ ^ ^ J / ^ V 1-sheeted 1-sheetedCH CH kx in space in 2-k 2-k space 3
(b)
Fig. 3. 2D optical waves in a planar D/D interface : (a) evanescent waves generated by TIR, (b) wavenumber surfaces: a sphere (kx+ky+kz =4) in 3-k space and a one-sheeted CH (kx+ky—Kz =1) in 2-k space. A cylindrical surface means the wavenumber matching condition. Wire frames are used for visibility.
3.2. A planar D/ND interface A planar D/ND interface is another physical system of 2D optical waves. We consider a planar D/ND interface as shown in Fig. 4 (a): an interface (xy-plane, z =0) between D half-space RD with e ( > 0), fi =1 (z < 0) and ND half space i?ND with Em{< 0), // =1 (z > 0). At the interface, there are two kinds of 2D optical waves as shown in Fig. 4 (a): an evanescent wave excited by reflection and SPP. The dimensions of an evanescent wave are 3D in RD and 2D in /?ND according to the definition of low-dimensional optical waves. On the other hand, the dimensions of SPP are 2D in both i?D and Rm. Fig. 4 (b) shows two wavenumber surfaces of an evanescent wave that is 3D optical wave in RD and 2D optical waves in R^D- Wavenumber surface in ND is a two-sheeted CH in 2-k space, while the surface in D is a sphere in 3-k space. The law of the conservation of wavenumber is expressed as a cylindrical surface. Intersections of these two surfaces and the cylindrical surface are wavenumber circles. 2D optical waves generated by reflection have the diffraction limit, because the radius of the sphere limits the maximum radius of the wavenumber circle. Fig. 4 (c) is wavenumber surfaces of SPP that is 2D optical wave in both /?D and i?M> Wavenumber surfaces in D and ND are one-sheeted and two sheeted CH, respectively. In contrast to Fig. 4 (b), there is no sphere in k space. This suggests that SPP have a potential to overcome the diffraction limit of 3D optical waves.
64
Kobayashi J. Takahara and T. Kobayashi
RND
ND
εND
z
2D
2D
x
2D
3D
y
ε
D
RD
reflection
SPP
(a) 2-sheeted 2 -sheeted CH in 2-k space
sphere 3-k space space inn 3-k
1-sheeted CH 1-sheeted in 2-k space
_ 4
~ kkzz ~ Kzz κ
W
4
~z κ
KZ 1
4
1 /W&SS& ^ I / 4
4 -4 -4
~ kkxx
/
IDHml
-4
~ kyy
-4
/
~ k^ xx
4 -4
(b)
(c)
4
~ ~kkyy
44 -4
Fig, 4, 2D optical waves in a planar D/ND interface : (a) schematic field distribution of evanescent waves and SPP. Magnetic field distribution Hx(z) is plot for SPP. (b) wavenumber ~2
-2
-2
surfaces of evanescent waves: a sphere (kx+ky+kz ~2
~2
=1) in 3-k space and a two-sheeted CH
_2
(&*+&,.—x"z =—4) in 2-k space, (c) wavenumber surfaces of SPP: a one-sheeted CH _2
_2
_2
^2
_2
_2
(^x+ij,— K"z =1) and a two-sheeted CH (kx+ky~Kz = - 4 ) in 2-k space. A cylindrical surface means the wavenumber matching condition. A wire frame is used for visibility.
Here we summarize well known properties of SPP. SPP is a coupled mode of a light and a surface mode of the collective excitation of a free-electron system (surface plasmon) [14]. We can derive electromagnetic field and wavenumber of SPP by solving Maxwell equations under boundary conditions. Electromagnetic field of SPP is a TM (Transverse Magnetic field) mode and localized at the interface as shown in Fig. 4(a). The wavenumber of SPP &SPP along a D/ND interface is, (13)
Low-dimensional optical waveguides and wavenumber surface
65 65
From Eq. (13), one can obtain the condition for SPP propagation as: £m<-e<0
\em\>£
(14)
We can derive attenuation coefficients in ND and D by using Eq. (6) and Eq. (8) as follows: _2
_2
_2
.2
(15) (16) From Eq. (13), one can obtain the normalized attenuation coefficients as: (17)
Note that HK^D and 1/KVJ are the penetration depths in ND and D, respectively. Since \£m\ > £, penetration depth in ND is smaller than that in D, i.e. 1/Ksw
It is worth noting that k$pp is slightly larger than ko for metals at optical frequency. A typical numerical example is kSpp = 1.03 ko for e = 1 and eND = - 1 9 in lossless silver at AQ = 633 nm (lossless values are selected for the sake of simplicity and used again below). 33. Two planar D/ND interfaces As we mentioned in the previous subsection, SPP in D/ND interface have a potential to increase Ak to infinity beyond the diffraction limit because wavenumber surfaces are not a sphere. We can synthesize 2D optical beam from many plane 2D optical waves of different £. In single D/ND interface, however, the minimum beam size of SPP is still limited to subwavelength order as follows: &rmm?n=
=—/
— < AT j - , n
(18)
^"mitoo is so to speak "the diffraction limit of 2D optical waves". Although kfrnirOD is smaller than ArmMD (n - 1) under the condition of Eq. (14), it is fixed by dielectric constants of materials. If we can somehow increase Ak instead of
66
J. Takahara and T. Kobayashi
changing the material constants, we can achieve &rmm2D
K h
H Hyy
D
t2D
ND
Hy
1
D
β
z
even
ND
odd x y
(a)
D
even
odd
(b)
Fig. 5. Two planar D/ND interfaces: (a) an ND gap (b) an ND film. Magnetic field distribution Hy(z) is plot schematically.
Let us consider two D/ND interfaces placed closely each other. Fig. 3.3(a) shows an ND gap structure: two ND half-spaces are separated by D region at the distance of/?. Fig. 3.3(b) shows an ND film structure: two D half-spaces are separated by ND region at the distance of h. As we decrease h to the order comparable to the penetration depth of SPP, SPP become to couple and form a new propagation mode. Schematic field profiles of the coupled mode are shown in Fig. 3.3. These modes are known as the Fano mode and these are also 2D optical waves [2,3]. Let us assume that the coupled mode propagates along the interface with wavenumber of P (propagation constant). We can calculate ft by solving the characteristic equation derived from Maxwell equations under boundary conditions. The characteristic equation of TM mode of 2D optical waves in ND gap or film is as follows: —)+$&- = o
(even mode)
(19)
tanh( K - ) + ^ - = 0
(odd mode),
(20)
2
2
£2yt
Wi
where £\ and e% is relative permittivity in core and clad layer, respectively. jj(j = 1, 2) is defined as:
Low-dimensional optical waveguides and wavenumber surface
67
We take £]-£, £2=£ND in the ND gap, and £\= £m, Sf=-e in the ND film. Note that even and odd modes are defined by the field distribution of Hy{z) as shown in Fig. 3.3. The characteristic equations of TM mode of 3D optical waves are as follows: 1
- ) - ^ - =0 £r
(even mode)
(22)
-)+^-=o
(odd mode)
(23)
2
2
W
The characteristic equations of TE (Transverse Electric field) mode of 3D optical waves are as follows: h
y
tanCf,—)-—=0 2 Y\ tan( Yx - ) + — = o 2 7%
(even mode)
(24)
(odd mode)
(25)
Figure 6 shows normalized /? with respect to h in ND gap and film for e^D = -19 and £= 1. The dimension of optical waves is 2D above horizontal dotted lines (J3/ka>l), and the dimension is 3D below the lines (0 <]3/ka
'M-A+-
(26)
68 68
J. J. Takahara Takahara and and T. T. Kobayashi Kobayashi
where r i s an attenuation coefficient in the clad layer. This means that t%D goes to zero when h approaches zero because ralso goes to infinity from Eq. (26). This indicates that the 2D optical waves of the upper branches can be confined into very thin («/lo) structures. Such unique properties of 2D optical waves are useful for making extremely narrow optical beams as described next. 3.4, An ND gap having a dielectric core By using the ND gap, we can decrease the vertical width of 2D optical wave to zero by decreasing h. However, in order to confine an optical wave in a nanometer-sized region, we have to confine it not only in vertical but also in horizontal direction. Hence, we have proposed an ND gap having a dielectric core as new 2D optical waveguides [12]. Figure 7 shows a schematic view of the 2D optical waveguides: a dielectric core is embedded in an ND gap where relative dielectric constant of the core £^ore is larger than the clad £dad. In this waveguide, we can confine and guide 2D optical waves along the dielectric core. 3
3
odd
even
2D
2
2
β/k0
β/k0
2D
odd
1
1
even 3D 3D
3D 3D
0
0 0
0.5
1 h/λ0
1.5
0
0.05 h/λ0
0.1
(b) (a) (a) Fig. 6. ft with respect to h: (a) an ND gap (b) an ND film. Solid and dotted lines show TM and TE mode, respectively. The dimension of optical wave is 2D above horizontal dotted-lines, and it is 3D below the lines.
In order to estimate the beam width confined in the core, we have performed approximate calculations using the effective index method [12]. Since even coupled TM mode of SPP can be confined in the dielectric core, we can define the beam width W as FWHM of the intensity profile. We have calculated W of the even TM mode confined in the core for eND = -19, %«* = 2 and E(;Ore — 4. The beam width has been calculated with respect to the width of the dielectric core d for various gap distances h. We have shown that W decreases linearly and takes a minimum value as d decreases. We stress that the
Low-dimensional optical waveguides and wavenumber surface
69
minimum value of W decreases as h decreases. Thus, it is possible to make and propagate an optical beam with nanometer-order diameter by using 2D optical waves in the ND gap having a dielectric core. The size of an optical beam synthesized by 2D optical waves is limited by the diffraction limit of 2D optical waves (Arm(>,2o). The important point is that we can control the value of Armln2D by changing the geometrical parameter of h. We are able to decrease Armin2D to zero by decreasing h to zero, because f3 of the even coupled TM mode increases infinity as h decreasing to zero as shown in Fig. 6(a). Recently, we have performed a numerical calculation by using Finite-Difference Time-Domain (FDTD) method for this system [13]. We have calculated the field distribution inside the ND gap and confirmed that 2D optical waves can be guided along the dielectric core.
co "Q D core D clad Dclad \
ε£clad clad
\
ε£core core
£ND ND ND ND clad clad ε
\
K
h<<λ0
\
\ \
1
1—L_ d<<λ0
t
ε£clad clad < ε core £core
Fig. 7. The cross sectional view of an ND gap having a dielectric core.
4. ONE-DIMENSIONAL OPTICAL WAVEGUIDES In this section, we expand the arguments about wavenumber surfaces of 2D optical waves to ID optical waves. Cylindrical interfaces are ID optical waveguides. We describe propagation properties of ID optical waves in cylindrical D/ND interfaces. 4,1. Wavenumber surfaces In D/D and D/ND interfaces ID optical waves are physically meaningful in many kinds of structures having D/D or D/ND interfaces as described below. Here we shall not consider a specific structure as ID optical waveguides and let us only assume ID optical waves propagate along x-axis. Fig. 8 shows three possible combinations of wavenumber surfaces in D/D and D/ND interfaces. Fig. 8 (a) is two wavenumber surfaces at D/D: in D the wavenumber surface of ID optical waves is a two-sheeted CH while the surface of 3D optical waves is a sphere. Fig. 8 (b) is wavenumber surfaces at D/ND: in ND the wavenumber surface
70
J. Takahara and T. Kobayashi
of ID optical waves is a one-sheeted CH while the surface of 3D optical waves is a sphere in D. kxs on each wavenumber surfaces must coincide at the interface by the law of the conservation of wavenumber. Because the wavenumber matching condition means the intersections of wavenumber surfaces and a plane (kx= const), kx is limited by the radius of the sphere. This is the geometrical representation of the diffraction limit. Thus, the beam radius is limited by the diffraction limit as long as 3D optical waves are considered. sphere in 3-k space
~ ky ~ κy
4
-4 4
~ ky ~y κ
4
2-sheeted CH in 1-k 1-k space in
κ~y
-4
~ kz ~z κ
-4 4
~ kz κ~z
4
~z κ
-4 -4
-4 -4
~ kx
-4 -4
~ kx 4
4
4
2-sheeted CH in 1-k space
(a)
1-sheeted CH in 1-k space
~ kx 4
(c)
(b)
Fig. 8. Wavenumber surfaces in ID optical waveguides : (a) D/D interface: a sphere ~2
~J
~2
_2
.2
_2
(kx+ky+kz =4) in 3-k space and a two-sheeted CH (kx-tcy—iCi =1) in 1-k space, (b) _2
_2
_2
D/NDinterface: a sphere ( kI+ky+kz=l ~2
~2
~2
) in 3-k space and a one-sheeted CH _2
_2
_2
{kx-Ky-Kz = - 4 ) in 1-k space, (c) D/ND interface: a one-sheeted CH (kx-/cy-iCz = - 4 ) ~2
_2
_2
and a two-sheeted CH (fc^-xv-xv =1) in 1-k space. A plane of fc*= const means the wavenumber matching condition.
Fig, 8(c) shows another combination of wavenumber surfaces at D/ND; the dimensions of optical waves are ID in both D and ND. Wavenumber surfaces of ID optical waves in ND and D are one-sheeted and two-sheeted CH, respectively. In contrast to Fig. 8 (a) and (b), there is no sphere in k space. Because CHs are open in k space, we do not have limitation of kx for wavenumber matching. Therefore, by using such kind of ID optical wave, we can form nano-sized beams beyond the diffraction limit. 4.2. A cylindrical D/ND interface The above general arguments suggest that there is no limitation in k space by using ID optical waves. Let us consider concrete waveguide structures next. Fig. 9 shows ID optical waveguides having a cylindrical D/ND interface: a ND rod and ND hole.
Low-dimensional optical waveguides and and wavenumber surface
(a)
71
(b)
Fig. 9. Cylindrical D/ND interfaces: (a) anND rod, (b) an ND hole. The characteristic equation of ID optical waves of i^th order hybrid mode for lossless ND rod and hole of radius a is as follows:
f2 JU where J3i$ propagation constant along the cylindrical axis, Iv and Kv are the V'-th order modified Bessel functions, and fi = Yia> & — Y2 <*• £\ arid Si is the relative permittivity of the core and the clad, respectively, fjij— 1, 2) is defined as:
(28)
By substituting v= 0 and using /'o(x) = I\{x) and ^T'o(x) = -K\{x), we can obtain two characteristic equations which correspond to TM and TE mode as: £iMiL+£iKMJ
Atf)
^(fe)
=0
0
(TMmode)
(29)
(TEmode)
(30)
The details of the analysis are described in the references [9,11-12]. Figure 10(a) shows the calculated /?with respect to a in an ND rod. In the calculations, the relative permittivity of the ND core and dielectric cladding are taken to be Em = - 1 9 and e-l. The TM mode and hybrid mode are propagation modes in the ID optical wave, while the TE mode is not a propagation mode in
72
J. Takahara and T. Kobayashi
the ID optical wave. Here, three modes are shown: the TM mode, and hybrid modes of order v=l and v-2. As shown in Fig, 10(a), fi of the TM mode diverges as a decreases to zero. This property is similar to the ND gap or ND film denoted above. We stress that such a dispersion relation of the TM mode is important for the confinement of an optical beam at nanometer-scale, p of the v=l hybrid modes asymptotically approaches ko and the cutoff effect is never observed, whereas the v —1 hybrid modes have a cutoff at a/Ao = 0.7. 1.2
0.25
TM
TM
rH /λ0
β/k0
0.2
1.1
0.15
0.1
hybrid ν=1
1 0
0.5
kSPP
ν=2 1
0.05
0 0
0.1
a/λ0
a/λ0
(a)
(b)
0.2
Fig. 10. Propagation modes of ID optical wave in the ND rod with Q
Let us consider beam radius of the TM mode. The electromagnetic field in the ND rod is localized at the D/ND interface and decays rapidly in the clad. We define the beam radius rH of the TM mode by the magnetic field: the beam radius is defined as the distance from the center to the point where magnetic field amplitude decreases to l/e from the value at the interface. This definition gives, (31) rH as a function of a can be calculated by solving Eq, (28) and Eq. (31) numerically. Figure 10(b) shows r# of the TM mode with respect to a. r# decreases to zero as a decreases beyond the diffraction limit. Therefore, it is possible to make and propagate an optical beam with nanometer-order diameter by using the TM mode of the ND rod.
Low-dimensional Low-dimensional optical waveguides and wavenumber surface
73
An ND hole is also interesting for nanophotonics. Many studies have been done since the observation of the extraordinary light transmission through a hole array [15]. In contrast to the ND rod, the dispersion curves in the ND hole are complicated and sensitive to the permittivity. TM mode curves in the ND hole for different £)VD are shown in Fig. 11. The dimension of optical waves is ID above a horizontal dashed line (/%> > 1) and 3D below the line. Under the condition of SPP propagation (i.e. | £Vo| >£)» the propagation mode shows the same type of cutoff behavior as conventional metal waveguides: /? shows a cutoff at a/Ao~O-35 for £ND=-19. The mode curves are changed in the case of I £ND\ < £ ft diverges as the core radius approaches zero and does not have a cutoff [11,12]. In addition, the group velocity of this mode is negative. 3 -0.5
ε=1
-1
εND
β/k0
2
1D
o
oil
-4 1
-
3D
εND=-19
-0.1 0 0
0.2
0.4
0.6
0.8
1
a/λ0 Fig. 11. TM modes in the ND hole: /?of TM mode versus a for e= 1 and £5MCF— 19 (solid), -4 (dashed), -1 (dotted), -0.5 (dash-dotted) and -0.1 (solid), ft is normalized to &o, a are normalized to ^o=633nm. The dimension of optical wave is ID above the horizontal dashed-line, and it is 3D below the line.
We can fulfill such condition (\eND]<e) by using high refractive index materials such as silicon (Si). Figure 12 shows typical two examples in the ND hole using Si as a dielectric: ft of TM and TE modes under the condition of I^WDI5* £and \£ND] < £• At io=500nm, /? diverges as the core radius approaches zero and does not have a cutoff. Thus, we can make nano-sized optical beams in the ND hole only under the condition | em |< e.
74 74
J.Takahara Takahara and Kobayashi J. and T. T. Kobayashi =15 10
10 10
1D 1D
=-19
0n
β/k0
β/k0
εND
j
5
I
0
'' f '
!.
I
0.2 0.2
A
a/Xo0 a/λ (a) (a)
IL
-kspp kSPP
15 ( £ε==15 )
5
3D 3D
/ 0
0.4 0.4
0
ε8=18 =18 • - ^ 8 NεD ND
1D 1D
=-8 =-8
-^ i / 0.2 0.2
3D 3D /
•'''"
0.4
a/λ a/Xo0 (b)
(b)
Fig. 12. TM and TE modes in the ND hole with Si core: /? versus a for (a) £ND=-19, £=15 (\£ND I > £) -?
4.3, Other linear structures There are a lot of candidates of ID optical waveguides in linear structures having D/ND interfaces. As for the cylindrical structures, trivial examples are two cylindrical interface system such as an ND tube or an ND coaxial hole. We have reported the propagation properties of ID optical waves in these waveguides [10,12]. By using the ND tube or the ND coaxial hole, we can squeeze the beam diameter of ID optical waves to nanometer order as an ND rod. Metallic slab waveguides have been proposed theoretically [16] and demonstrated [17-19]. These linear structures are also considered to be ID optical waveguides. Furthermore, the another interesting structures of ID optical waveguides are wedge structures, where edge SPP modes can propagate through [20,21]. A wedge structure is also considered to be a ID optical waveguide. Further investigations are needed about these linear structures from the viewpoint of ID optical waves. 4.4. Propagation loss and applications Metals at optical frequency are real materials as ND, because the imaginary part of eis much smaller than the real part (|Im[£]| « | R e [ f ] | ) . Hence, we treated metals as an ideal lossless ND in the previous sections. In actual metals, however, £is a complex number having a small imaginary part, e.g. ^ © = - 1 9 0.53/ at >?o=633nm in silver [3], The imaginary part is attributed to Ohmic loss, which causes some transmission loss.
Low-dimensional optical waveguides and wavenumber surface
75 75
In order to estimate propagation length L of ID optical waves, we have calculated propagation constant (fi= J3R - ifk) for lossy ND [9]. L is defined as l/e decay length of the power and can be calculated as L = 1/2$. We have calculated L of the TM mode with respect to XQ for various diameters in the ND rod. We have shown that the order of L is from lOOnm to lOjim and L gradually decreases as A$ decreases until the rapid decrease at UV region [12]. For a silver-core rod of lOnm diameter in air (e=l), the beam diameter is 22 nm and L = 0.5p,m at ^=633 nm. Furthermore, as we have shown in previous papers, $ increases as o decreases. This means that a ID optical beam with smaller diameter has higher loss. We stress that the transmission loss is not a very serious problem for short range transmission (<^o) as long as applications to optical devices of nanometer size are considered. One of the important applications of ID optical waveguides in the near future is a high-efficiency optical probe with high spatial resolution. We could use such a device for SNOM, optical recording, and nano- manipulation. As for the applications of small apertures, the ND coaxial hole is more sufficient rather than the ND hole, because ID optical wave can propagate through the ND coaxial hole without a cutoff. There are many other attractive applications too, such as nano-optical circuits and devices. Low-dimensional optical waveguides may play a crucial role in constructing nano-optical integrated circuits. 4.5. Excitation In order to use nano-sized optical beams, we need to excite the TM mode in 2D or ID optical waveguides. The FDTD simulations have confirmed that 2D optical waves can be excited actually by an end-fire method in ND gap with gap distance of only lOnm [13]. Because k of nano-optical beams is much larger than k of conventional optical beams propagating in air, optical fibers or dielectric waveguides, the coherent coupling between different k vectors are difficult in principle. In addition, the field distributions between 3D and 2D optical waves are different. As a result, a coupling efficiency between 3D and 2D optical waves is low. It is necessary to couple efficiently from 3D optical waves to the TM mode of 2D or ID optical waves for applications to nano-optical devices. In order to excite the TM mode efficiently, we have proposed lowdimensional optical wave sources and couplers as shown in Fig. 13: an ND gap and an ND tube including fluorescent materials [22]. In ND gaps and ND tubes, there are no propagation modes except TM mode of low-dimensional optical waves. By using these devices, we can excite 2D or ID optical waves electrically by an electric current. These structures also can be used as a coupler or a mode converter from 3D optical waves to 2D or ID optical waves. These are so called "incoherent" coupling devices: irradiated 3D optical waves to the edge of the waveguide penetrate into the waveguide as evanescent waves and
76
J. Takahara and T. Kobayashi
are absorbed by fluorescent materials, then excite the TM modes incoherently through energy transfer from the fluorescent materials to low-dimensional optical waves. In our experiment, we have observed TM polarized light emissions from ND gap containing organic electroluminescent materials such as Alq3 and ff-NPD [23]. We have attributed the polarized light emissions to electric excitation of 2D optical waves. near near field field excitation excitation
ND gap fluorescent fluorescent media \ media
fluorescent \ media
\
X
near field excitation
ND coaxial line ND \ \ ^ -
\
TV "-...-I; .1
\
A
r
current current excitation
2D Optical Waves
excitation
(a)
^k
•1 ^\
L H | - V\/\/w current
excitation
1D Optical Waves
(b)
Fig. 13. 2D and ID optical source: (a) an ND gap with fluorescent media, (b) an ND tube with fluorescent media. TM mode can be excited electrically or optically.
5. SUMMARY We have described negative dielectric waveguides by introducing concepts of low-dimensional optical waves and wavenumber surface. The propagation and optical beam properties of low-dimensional optical waveguides have been presented theoretically. A lot of propagation modes in planar and cylindrical negative dielectric waveguides have been classified from the point of view of nanophotonics. Physical meanings of the formation of nano-sized optical beam have been discussed by wavenumber surface. Applications to nano-optical devices and excitations have been discussed. Low-dimensional optical waves are the useful concept in treating optical waves in nanostructures. REFERENCES [1] W. L. Barnes, A. Dereux, and T. W. Ebbesen, Nature, 424 (2003) 824 and references therein. [2] E. N. Economou, Phys. Rev., 2 (1969) 539. [3] J. J. Burke, G. I. Stegeman, and T. Tamir, Phys. Rev. B, 33 (1986) 5186. [4] M. Fukui, V-C.Y. So and R. Normandin, Phys. Status Solidi B, 91 (1979) K61. [5] J. C. Ashley and L. C. Emerson, Surface Science, 41 (1974) 615. [6] C.A. Pfeiffer, E.N. Economou, and K.L. Ngai, Phys. Rev. B, 10 (1974) 3038. [7] L. Novotny and C. Harrier, Phys. Rev. E, 50 (1994) 4094.
Low-dimensional optical waveguides and wavenumber surface
11 77
[8] M. Ohtsu, K. Kobayashi, T. Kawazoe, S. Sangu, and T. Yatsui, IEEE J. Select. Top. Quantum Electron., 8 (2002) 839. [9] J. Takahara, S. YamagisM, H. Taki, A. Morimoto, and T. Kobayashi, Opt. Lett., 22 (1997)475. [10] J. Takahara and T. Kobayashi, Opt. Photon. News, 15, No.10 (2004) 54. [11] J. Takahara, in Springer Series in Optical Sciences 84, S. Kawata, M. Ohtsu, and M. Me (eds.), Nano-Optics, Springer-Verlag, 2002, pp. 126-133. [12] J. Takahara and T. Kobayashi, Proc. SPIE, 5604 (2004) 158. [13] F. Kusunoki, J. Takahara, and T. Kobayashi, Appl. Phys. Lett., 86 (2005) 211101. [14] H. Raether, Surface plasmons on smooth and rough surfaces and on gratings, Springer, Berlin, 1988. [15] T. W. Ebbesen, H. J. Lezec, H. F. Ghaemi, T. Thio, and P. A. Wolff, Nature, 391 (1998) 667. [16] P. Berini, Opt. Lett., 24 (1999) 1011. [17] J. C. Weeber, J. R. Krenn, A. Dereux, B. Lamprecht, Y. Lacroute, and J. P. Goudonnet, Phys. Rev. B, 64 (2001) 045411. [18] T. Yatsui, M. Kourigi, and M. Ohtsu, Appl. Phys. Lett, 79 (2001) 4583. [19] T. Onuki, T. Tokizaki, Y. Watanabe, T. Tsuchiya, and T. Tani, Appl. Phys. Lett., 80 (2001) 4629. [20] I. V. Novikov and A. A. Maradudin, Phys. Rev. B, 66 (2002) 035403. [21] D. K. Gramotnev and D. F. P. Pile, Appl. Phys. Lett., 85 (2004) 6323. [22] J. Takahara, A. Eda, K. Nakamura, M. Yokoyama, and T. Kobayashi, Technical Digest of 7th Int'l Conference on Near Field Optics and Related Techniques (nfo-7) (2002) 232. [23] J. Takahara, Y. Fukasawa, and T. Kobayashi, J. Kor. Phys. Soc. (2005) in press.
This Page is Intentionally Left Blank
PART I I : PLASMON ENHANCED SPECTROSCOPYAND MOLECULAR DYNAMICS
This Page is Intentionally Left Blank
Handai Nanophotonics, Volume 2 S. Kawata and H. Masuhara (Editors) © 2006 Elsevier B.V. All rights reserved.
81 81
Chapter 5
Specific Raman band shift caused by mechano-chemical effect in tip-enhanced near-field Raman spectroscopy H. Watanabea>\ N. Hayazawa', Y. Inouyed, and S. Kawata*1* a
Graduate School of Engineering, Osaka University, 2-1 Yamadaoka, Suita, Osaka 565-0871, Japan b
Analysis Technology Center, Advanced Core Technology Laboratories, Fuji Photo Film Co. Ltd., 210 Nakanuma, Minami-ashigara, Kanagawa 250-0193, Japan c
Riken, 2-1 Hirosawa, Wako, Saitama 332-0012, Japan
d
Graduate School of Frontier Biosciences, Osaka University, 2-1 Yamadaoka, Suita, Osaka 565-0871, Japan 1. INTRODUCTION Tip-enhanced near-field Raman scattering (TERS) spectroscopy [1-4] using an apertureless metallic probe tip [5-9] realizes nanoscale observation of molecular vibrations in practical use. The tip provides two enhancement effects; one is caused by an enhanced electric field due to the localized surface plasmon polariton excitations [10-14], and the other is generated as a result of a chemical interaction between samples and the tip [15-17]. These effects are analogous to the electromagnetic mechanism and the charge transfer mechanism of surface enhanced Raman scattering (SERS), respectively. We reported on TERS spectra of adenine (Fig. 1) [18], a DNA base, where the observed Raman bands exhibits similar frequency shifts to the conventional SERS. In addition, we found that the vibrational frequencies of several Raman bands of the TERS unambiguous shift to the values of the corresponding bands in SERS. In SERS measurement, an adenine molecule has been reported to adsorb chemically on silver substrates through a specific nitrogen atom of adenine molecule [19-22], which afford chemical interaction. The interaction is well known as a cause of the charge transfer (CT) or chemical mechanism [23,24] that is one of the enhancement mechanisms of SERS. Quantum chemical studies
82
H. Watanabe et al.
[25,26] for the Raman band shifts of pyridine molecules adsorbed on various metal surfaces have been reported using density functional theory (DFT) calculations. These studies were made to interpret the SERS spectral features on the basis of the CT mechanism, where a simple metal-pyridine (1:1) complex representing an appropriate adsorption model of pyridine molecules on the metals was discussed. In the TERS spectroscopy, the CT complexes of metals, which are located at the apex of the metallic tip, might be formed with the adsorbed species. However, the unambiguous shift of specific Raman band in TERS spectrum suggests that another interaction exists in addition to the effect on the CT mechanism. In this chapter we describe the experimental results of the TERS spectroscopy of a single nanocrystal of DNA-base adenine molecules using a silver-coated apertureless probe of an atomic force microscope. Then we analyze the detected TERS bands in comparison with the bands obtained in the conventional SERS spectrum by performing high-level DFT calculations of the silver-adenine (1:1) complexes [27],
(2)
Adenine Fig. 1. Molecular structure and atomic numbering of adenine.
Specific Raman band shift caused by mechano-chemical effect in tip-enhanced near-field Raman spectroscopy
83 83
2. EXPERIMENTAL AND COMPUTATIONAL PROCEDURES Figure 2 shows the configuration of the experimental setup of the TERS spectroscopy [4]. The light source for excitation of Raman scattering was a frequency- doubled Nd:YVC>4 laser, with an average power of about 2.5 mW, operating at a wavelength of 532 nm, at which adenine has negligible absorption. The probing laser beam was focused at the sample by an oil-immersion objective lens having a large numerical aperture (NA) equal to 1.4 and a magnification factor of 60. A portion of the incident beam with focusing angles equivalent to NA less than 1.0 was truncated by a circular absorbing mask inserted at the light path to form an evanescent field light at the focused spot [28]. The Raman NSOM probe was an atomic force microscope (AFM) silicon cantilever coated with a 40-nm-thick silver layer by a thermal evaporation process. The distance between the sample and the metallic cantilever was controlled by the contact mode of standard AFM operation. The scattered light was collected by the same objective lens, and directed to a spectrophotometer (the focal length is 300 mm, 1200 lines/mm) equipped with a liquid-nitrogencooled charge-coupled device (CCD) detector (1340 x 400 channels) for Raman spectra measurements. The excitation light and the scattered Rayleigh signal were sufficiently reduced using a holographic notch filter [the center wavelength is 532 nm, the full width at half maximum (FWHM) is 6 nm, and the optical density (OD) is greater than 6.0 at 532 nm]. Adenine was purchased from the Extrasynthese Co. and was not treated with any further purification. Adenine nanocrystals were prepared by casting an ethanol solution of adenine molecules (0.1 mmol/1) on a cover-glass slip. The size of the nanocrystals was about 7 x 20 nm in the lateral direction and 15 nm thick. Quantum chemical calculations for analysis and identification of the vibrationai frequencies and Raman intensities for free adenine molecule were carried out using the B3LYP functional [29,30] with the basis set of 6311+G(4/0- The vibrationai properties of the adenine-silver complexes were calculated using the UB3LYP functional in combination with the basis set of 6311+G(4/0 f° r carbon, nitrogen or hydrogen atom, and the Stuttgart/Dresden relativistic effective core potentials combined with the associated basis functions of the valence electrons (SDD) [31-33] for silver atom. The calculations were performed using the GAUSSIAN98 Revision A.9 program package [34]. Optimized geometries and the calculated vibrationai modes were visualized using the MOLDEN program (QCPE 619) [35] and the VLX program [36,37], respectively.
84
H. Watanabe et al. PZT scanner (AFM head)
XY-PZT stage Objective lens(NA:1.4)
Controller
Laser Nd:YVO4 laser (532 nm)
PC Mask
Controller^— Controller Notch filter
CCD camera
N22 cooled CCD | Liquid N
\*—
Fig. 2. Experimental setup of TERS spectroscopy.
3. RESULTS AND DISCUSSION 3.1. TERS spectra of adenine nanocrystals Figure 3 shows Raman spectra measured using our system. The TERS spectrum [Fig. 3 (a)] of a single nanocrystal of adenine molecules was measured when the silver-coated tip was in contact with the sample surface. Eight Raman bands, including two intense scattered bands at 739 or 1328 cm~l, were detected. The peak at 924 cm - 1 was due to the Raman scattering of the cover-glass slip. The far-field Raman spectrum [Fig. 3(b)] of the same sample using the same microscope, without the silver coated tip, was shown to have almost no Raman scattering from the adenine molecule except the faint band at 723 cm."1. The measurement conditions were exactly the same for both experiments except for the tip-sample distance. Exposure time was set to 1 min for both measurements.
1328
85 85
(a) 1396
1232 1268
1021
965
798
Raman intensity [arb. units]
739
Specific Raman band shift caused by mechano-chemical effect in tip-enhanced near-field Raman spectroscopy
3
600 600
800 800
1000
1200
1400
w c
(b) (b) 723
c
CO
(0 (0
l V '\WVi^^^^^J^ v v w ^ ^ - l ^ ^ 600 600
800 800
1000
1200
1400
1
[cm-–1] Raman shift [cm Fig. 3. (a) Tip-enhanced near-field Raman and (b) far-field Raman spectra of adenine nanocrystals. The peak at 923 cm~l is derived from the glass substrate.
Figure 4(a) shows a spectral mapping of the TERS spectra for nanocrystals of adenine molecules obtained with the silver-coated tip at 30-nm intervals. Contact mode was used for AFM operation. The exposure time of the image is set to 10 s per line scan. Figure 4(b) exhibits intensity distributions of the two predominant Raman bands at 739 and 1328 cm" 1 . The intensity distribution at 850 cm" 1 , where no Raman band exists, shows no particular optical response signal in the same figure. The peak response of the intensity profile at 60 nm of X axis as well as the edge response around 120nmofXaxisinthe Fig. 4(b)
86
Raman intensity
H. H. Watanabe Watanabeetetal. al.
450 420 390 360 330 300 270 240 210 180 150 120 90 60 30 0
(a)
c a> c
-i
nm ]
c
Xa xis [
(0 (0
500 500
1000 1000
1500 1500 1
[cm-–1] Raman shift [cm
c 3
1328 cm "
1
(b)
(0 4-1
739 cm"
c
1
(0
100
200
300
400
500
X axis [nm] Fig. 4. (a) Near-field Raman spectral mapping of adenine nanocrystals at 30 nm intervals. Each spectrum is the subtraction of the corresponding tip-enhanced near-field Raman spectrum from the far-field Raman, (b) Raman intensity distributions of typical Raman bands at 739 and 1328 cm"1 against the tip displacement within the focused spot. The intensity distribution at 850 cm"1 is also given.
Specific Raman band shift caused by mechano-chemical effect in tip-enhanced near-field Raman spectroscopy
87
suggests that the smallest observable feature is 30 nm. Assuming that the diameter of the enhanced electric field is 30 nm corresponding to the lateral resolution and that the diameter of the focused light spot is 400 nm, the enhancement factor for the ring-breathing mode of the whole molecule at 739 cm - 1 is 2.7 x 103. However, it is difficult to calculate the enhancement factor for the band at 1328 cm" 1 because the far-field Raman signal without the tip is too weak to be detected. An enhancement factor of 10^ ~ 10^ observed in the silvercoated tip is reasonable compared to the SERS spectra of adenine molecules adsorbed on silver spheres having the same radii as the width of the probe tip apex (the excitation frequency is also the same) [38]. 3,2. Vibrational analysis of adenine molecule The vibrational frequencies and the Raman intensities have been calculated using DFT calculations. The geometrical structure of the adenine molecule was optimized at the B3LYP functional using 6-311+G(rf,p) basis set, which was then followed by calculations of the vibrational properties at the same level of theory using the same basis set. The calculated geometry was a planar shape (C$ symmetry) having no imaginary frequencies, representing an energy minimum. The predicted bond lengths and bond angles agree well with the crystallographic structure of adenine [19,39,40] within the error of 1%. Multiplying the calculated vibrational frequencies by a single scale factor of 0.9942 provides a good fit between the calculated and the observed frequencies. Table 1 shows the normal modes of the eight Raman bands appearing in the TERS spectrum of adenine nanocrystals and the calculated frequencies with the uniform scaling. The peak positions observed in the normal Raman spectrum of polycrystalline adenine [22] and in the SERS spectrum of adenine molecules adsorbed in colloidal silver [22] are listed together in the same table. Vibrational patterns of all the normal modes listed in Table 1 are also depicted in Fig. 5. The near-field Raman band at 739 cm"1 (Band H) is assigned to a ring-breathing mode of whole adenine molecules. The band at 1328 cm" 1 (Band B) is assigned to a summation of two different vibrational modes. The first mode involves the C5-N7 and N1-C2 symmetric stretching coupled with the C2-H and Cg-H symmetric in-plane bend. The second mode involves the C2-H, Cg-H and N9-H in-plane bend with a minor contribution from the Ce-Ni, Cg-Ng and N3-C4 stretches. The six other bands are also assigned to the corresponding internal vibrational modes of the adenine molecule. It is important to note that some of the eight bands detected in the silver-tip-enhanced near-field Raman spectrum for adenine nanocrystal apparently differ in the frequencies from the corresponding bands in the SERS for adenine adsorbed on colloidal silver. The band due to the ring-breathing mode at 739 em- 1 in the near-field Raman is 16 cm" 1 higher than the Raman band of the ring-breathing mode in normal Raman.
88
H. H. Watanabe Watanabe et et al. al.
Table 1 Calculated vibrational frequencies at B3LYP functional using &-'5\\+G(d,p) basis set and vibrational frequencies of the tip-enhanced Raman (TERS) of adenine nanocrystals. The results of the SERS and the normal Raman (NR) of adenine from McNaughton et al. (Ref. 22) are listed. Frequencies are given in cnr 1 .
This study B3LYP 1406 1356 1259 1235 1003 939 804 721
Band
A B C D
E F G H
TERS 1396 1328 1268 1232 1021 965 798 739
a McNaughton et al. SERS NR 1397 1419 1336 1333 1248 1268 1234 1244 1025 1029 942 961 790 797 733 723
"Reference 22.
νA
νB
νC
νD
νE
νF
νG
νH
Fig. 5. Vibrational patterns of the normal modes for adenine molecule. The mode numbers are given to each modes corresponding to the bands in Table 1.
The same band in SERS spectrum is observed at 733 cm- 1 , of which wavenumber is only 10 cm- 1 higher than that in normal Raman spectra. These interesting phenomena suggest that the shifts of Raman frequencies caused by dynamic contact of the silver-coated tip on the samples may differ from the shifts attributed to the surface enhancement effects of conventional SERS caused by thermally stable adsorption of the samples onto the silver surfaces.
Specific Raman band shift caused by mechano-chemical effect in tip-enhanced near-field Raman spectroscopy
89
3.3. Vibrational calculation of silver-adenine complexes Four possible silver-adenine complex isomers were examined in which every silver atom is adjacent to any position of nitrogen atoms at Ni, N3, N7, and N10, respectively. Complex isomers in which a silver atom is adjacent to N9 nitrogen of N7-H adenine tautomer were omitted because of instability of the N7-H tautomer [19]. These four isomers are the simplest models having the interaction of adenine with the silver metal surface or the silver-coated tip at the apex. The structures of the Ad-Nl, Ad-N3, Ad-N7, and Ad-NIO isomers are illustrated in Fig. 6. The geometries of these isomers were fully optimized with the UB3LYP functional using the same basis set for adenine moiety and the SDD pseudopotentials for silver. The vibrational properties were then calculated using the same level of the DFT calculations and basis set conditions. The spin states for these isomers are doublet and the charges are neutral for all. The calculated geometrical parameters are listed in Table 2, which also contains the calculated relative binding energies (AE) corrected with the zero point energies (ZPE) based on scaled harmonic normal-mode frequency with a single factor of 0.989 [41]. The optimized geometries of the adenine molecule and four complex isomers are demonstrated in Fig. 7, which represent energy minima due to the absence of imaginary frequencies. In our calculation level, the most stable isomer is the Ad-N3, where the silver atom is located along the plane of the purine ring and two hydrogen atoms that are slightly above the plane (Cj symmetry). The calculated binding energies of the other isomers were nearly equal or slightly higher by only a few kilocalories per mole, with or without the ZPE correction. This means that these four silver-adenine complex isomers are probable candidates for the interaction models as far as the binding energies are concerned. Both the Ad-Nl and the Ad-N7 exhibit geometrical characteristics similar to the Ad-N3. These three isomers demonstrate a 'side-on' adsorption to silver surfaces. In the geometry of the Ad-Nl 0, the silver atom is located above the plane in contrast with the other isomers. This configuration represents a simplest model of a 'flat-on' adsorption onto silver surfaces.
90
H. Watanabe et al.
Table 2 Optimized bond length (in A), bond angles (in degrees), dihedral angles (in degrees), and binding energies (in kilocalories per mole) of adenine at the B3LYP/6-31l+G(drp) level and its silver complex isomers at the UB3LYP/6-31 l+G(d,p)(for C, N, H)/SDD(for Ag) level. Adenine Ad-Nl Ad-N3 Ad-N7 Ad-NIO B3LYP Crystal" UB3LYP 2.524 2.502 2.502 2.946 0.0 0.1 -112.2 0.7 1.341 1.338 1.347 1.335 1.339 1.342 N1-C2 C 2 -N 3 1.334 1.332 1.338 1.335 1.334 1.328 1.342 1.334 N 3 -C 4 1.336 1.338 1.340 1.334 1.382 1.396 1.399 1.397 1.396 1.395 Q-C5 1.410 1.410 1.413 1.405 C 5 -C 6 1.409 1.409 C 5 -N 7 1.385 1.385 1.382 1.383 1.384 1.387 1.308 1.312 N 7 -C 8 1.308 1.312 1.308 1.308 C 8 -N 9 1.380 1.380 1.381 1.380 1.367 1.371 1.353 1.337 1.346 1.349 1.349 1.371 C6-N10 N,-C 2 -N 3 128.5 129.0 128.0 127.9 128.4 128.2 111.5 110.8 112.4 111.4 C 2 -N 3 -C 4 111.7 111.7 126.9 126.7 126.7 125.9 126.9 126.6 N 3 -C 4 -C 5 C4-C5-C6 116.0 116.9 116.4 116.3 116.1 115.8 C 4 -C 5 -N 7 111.3 111.3 111.0 110,4 111.2 110.7 C 5 -N 7 -C g 104.2 104.2 103.9 104.2 104.3 105.0 112.6 N7-C8-N9 113.2 113.8 113.2 113.2 113.3 122.4 123.1 122.3 C 5 -C 6 -N, o 122.3 123.4 122.6 Ni-Cs-Nio-H 0.0 -0.6 -19.4 -1.3 -0.5 3.9 3.7 3.5 0.5 M 3.2 3.4 3.0 -0.3 A(£+ZPEd) "Atom numbering as in Fig. 1. "References 19,39, and 40. c The angle between the plane of the purine ring and the bond axis of the silver atom and the adjacent nitrogen is given. d The zero point energy (ZPE) is corrected by the scaled frequencies with a single factor of 0.989 (Ref. 41). Coordinate3 N-Ag Ad-N-Agc
Specific Raman band shift caused by mechano-chemical effect in tip-enhanced near-field Raman spectroscopy
91 91
Ad-N1 M j
Ad-N7
Ad-N10
Fig. 6. Molecular structure of four silver-adenine (1:1) complex isomers.
Ad-N1
Ad-N3
Ad-N7
Ad-N1
Fig. 7. Optimized geometry of silver-adenine (1:1) complex isomers.
The calculated frequencies of these isomers were uniformly scaled by the same factor of 0.9942 as employed in a free adenine molecule. The corrected vibrational frequencies vs. the Raman intensities of an adenine molecule and its isomers are listed in Table 3. This does not include the results relating to the CH stretching modes having frequencies higher than 3000 cm~l. Figure 8 shows the SERS spectrum of adenine adsorbed on colloidal silver, the TERS spectrum of adenine nanocrystal, and the predicted Raman spectra, which are plotted versus the corresponding frequencies in the region of 1500 to 600 cm" 1 .
92
H. Watanabe et al.
Table 3 Calculated vibrational frequencies, Raman intensities of adenine at the B3LYP/6-311+G(4p) level and its silver complex isomers at the UB3LYP/6-311+G(4p)(for C, N, H)/SDD(for Ag) level. Vibrational frequencies scaled by a single factor of 0.9942 are given. Raman activities are given in units of A 4 amu"'.
Adenine
Ad-Nl
Mode
V/cnT1
Intensity
v/cnT1
Intensity v/cnf1
Ad-N3
Intensity v/cm"1
Intensity
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37
1647 1625 1599 1505 1493 1424 1406 1356 1348 1319 1259 1235 1136 1073 1003 970 939 895 846 804 721 681 663 615 572 536 529 517 514 297 275 215 163 51 -
9.1 21.9 4.6 77.1 13.0 0.6 27.3 39.1 36.3 17.9 16.8 15.5 2.5 8.5 5.0 0.1 4.2 1.7 1.0 0.6 25.3 0.2 0.2 7.2 0.7 0.0 3.2 2.8 0.9 0.0 3.0 0.0 0.2 0.0 -
1654 1621 1598 1504 1497 1429 1406 1362 1347 1331 1258 1238 1140 1073 1007 968 941 901 850 799 721 679 662 623 572 545 532 522 523 296 283 214 158 222 79 67 30
18.6 24.4 19.6 134.1 0.6 10.2 41.4 57.4 35.0 39.0 22.5 23.2 2.6 10.3 23.4 1.4 6.9 3.6 1.4 0.1 30.5 0.2 1.5 15.5 0.9 5.3 3.3 3.0 0.3 0.6 4.0 5.1 0.2 5.7 2.9 0.7 5.5
27.6 19.8 5.5 52.4 29.1 11.0 36.0 32.7 54.0 23.7 58.0 25.6 6.7 22.1 5.2 0.3 4.4 0.6 0.5 0.5 49.6 0.1 0.3 4.6 4.2 0.1 10.7 2.4 4.1 0.3 2.4 0.2 0.1 0.3 4.4 0.8 5.3
13.3 18.0 15.5 86.4 35.0 4.2 33.4 21.0 33.4 13.1 32.9 1.8 3.8 3.0 3.2 0.1 4.8 0.9 0.0 0.1 36.8 0.2 0.6 5.0 1.0 6.4 4.3 4.8 1.3 0.2 5.5 1.5 0.4 11.1 4.9 0.0 4.7
1651 1624 1603 1506 1492 1433 1412 1357 1359 1318 1259 1237 1141 1072 1005 968 942 903 853 799 724 679 664 615 574 547 535 522 535 301 277 224 188 157 85 51 40
Ad-NIO
Ad-N7 1652 1625 1601 1510 1493 1425 1411 1358 1352 1317 1259 1232 1136 1092 1009 975 951 895 849 799 721 681 660 616 572 546 529 521 524 299 301 202 158 248 90 66 37
Intensity
1638 1629 1601 1505 1488 1419 1406 1352 1344 1315 1259 1246 1133 1073 1032 970 938 894 854 807 721 681 662 619 574 537 555 523 515 300 496 217 172 273 47 28 25
7.6 27.8 15.6 95.3 19.8 5.0 29.0 99.6 37.7 68.3 14.5 24.6 4.9 7.9 10.5 1.1 6.0 3.0 3.3 8.9 23.7 25.7 1.5 7,5 3.4 7.7 8.8 4.1 3.4 2.4 1.0 0.2 12.7 2.5 27.2 3.4 1.0
Specific Raman band shift caused by mechano-chemical effect in tip-enhanced near-field Raman spectroscopy
93 93
(a) (a)
Raman intensity [arb.units]
f
J (b)
c
3 .Q (0
0)
c a> c
(c)
4 1 |iO m u -–1 10AÅ4aamu
,. , , . II
'l f
-
\
I
i
f
|
i
(d)
i
(0 (0
-
| (e)
I . . I . I
DC -
I I
L.i.
600
(f) (f)
800
h , 1000
(g)
-
II
1200
1400
1
Raman shift [cm [cm-–1] Fig. 8, (a) The SERS spectrum of adenine molecules adsorbed on colloidal silver, (b) The TERS spectrum of adenine nanocrystals. The predicted Raman spectra of (c) adenine and its silver complex isomers: (d) Ad-Nl, (e) Ad-N3, (f) Ad-N7, and (g) Ad-NIO, as synthesized from the calculated Raman intensities against the corresponding frequencies.
94
H. Watanabe et al.
3.4. Comparison between the calculated and measured Raman spectra Comparing with the SERS spectrum of Fig. 8(a), there are some spurious Raman bands in the calculated Raman spectra of both the Ad-Nl and the AdN10. The calculated Raman spectrum of the Ad-Nl [Fig. 8(d)] involves a strong band at 1007 cm" 1 due to the Nio-H in-plane rocking mode (Vis in Fig. 9), while no significant Raman peaks are observed in this frequency region of the SERS spectrum. The calculated spectrum of the Ad-NIO in Fig. 8(g) exhibits two strong bands at 807 cm~l and 681 enr~l due to the out-of-plane ring-deforming modes (V20 and V22 respectively in Fig. 9), while no corresponding peaks exist in the frequency regions of the SERS spectrum. Judging from the results, these two isomers do not seem to be active species of SERS. The calculated Raman spectra of both the Ad-N3 [Fig. 8(e)] and the Ad-N7 [Fig. 8(f)] agree well with the SERS spectrum. The calculated Raman band due to the ring-breathing mode (V21) of the Ad-N3 shows a higher frequency shift by 3 cm - 1 than the calculated band of free adenine [shown in Fig. 8(c)], which coincides with the experimental data observed in SERS. In contrast, the same bands of the Ad-N7 show no frequency shift. These results show that the Raman active species observed in SERS have the structure of adenine molecules adsorbed on the silver surfaces via N3 nitrogen (Ad-N3). Therefore the Ad-N3 model represents the most probable structure; however, it is not quite easy to differentiate between the two isomers (Ad-N3 and Ad-N7) from such a small frequency shift.
Ag
v 15 (Ad-N1)
Ag
v 20 (Ad-N10)
v 22 (Ad-N10)
Fig. 9. Vibrational patterns of some normal modes for silver-adenine complex isomers. The mode number is given to each mode of the isomers.
Specific Raman band shift caused by mechano-chemical effect in tip-enhanced near-field Raman spectroscopy
95
Whole spectral patterns of TERS spectrum are similar to that of the calculated Raman spectra of both the Ad-N3 and the Ad-N7. However, the large frequency shift of TERS band towards high frequency due to the ring-breathing mode at 739 e n r 1 does not appear in the calculated spectra of either the Ad-N3 or the Ad-N7. The fact means that either of Ad-N3 and Ad-N7 isomers is not exactly active species in TERS spectroscopy. 3.5. Deformation of silver-adenine complex due to the atomic force One of the major differences between the TERS and the SERS experiment is sample preparation of adenine molecules: a unimolecular form in the SERS spectroscopy and a nanocrystalline form in the TERS spectroscopy were used respectively. However, this difference is not the dominant factor of a change in the frequency because there are no significant frequency shifts observed among IR frequencies of matrix isolated adenine [21], and IR and Raman frequencies of crystalline adenine [22]. The other difference in experimental condition is the interaction mechanism of adenine with silver surfaces. In SERS, adenine is adsorbed in equilibrium onto silver surfaces, whereas in the TERS, nanocrystalline adenine is pressed by the silver probe tip with a constant atomic force. Assuming that the atomic force is applied only to the contraction of the bond between the silver atom of the tip and the adjacent nitrogen of the adenine molecule that are formed into complex, the bond distance would be expected to shrink and the vibrational frequencies may then shift. In our experiment, we used a cantilever with a spring constant of 0.03 N/m and a diameter of a silvercoated tip apex of this cantilever was 5 - 1 0 nm. The atomic force was kept constant at 0.3 nN by the feedback loop. Deduced from the unit cell parameters of single-crystal 9-methyladenine [39, 40], a couple of adenine molecules exists in a rectangle approximately 0.77 nm lateral, by 0.85 nm wide. If we assume that the force is equally applied to all molecules which are adjacent to the tip apex, the adenine molecules are subjected to a pressure of ~ 1 - 5 pN/molecule by the silver atom attached on the surface of the tip. For the further understanding of the TERS active species of adenine molecules, we investigated the transition states of both the Ad-N3 and the Ad-N7 isomers by changing the bond distance (in the model) between the nitrogen of adenine molecule and the silver atom. Assuming that the bond distances for the calculations are 2.502 (equilibrium), 2.75 (10% elongation), 2.25 (10% contraction), and 2.0 angstroms (20% contraction), the binding energies (AE), the vibrational frequencies of free adenine molecules and the two complex isomers of three states are calculated as shown in Table 4. The results relating to the CH stretching modes having frequencies higher than 3000 cm~l are omitted.
96
H. Watanabe et al.
Table 4 Calculated relative binding energies (A£), and vibrational frequencies of adenine at the B3LYP/6-311+G(fi?,/>) level and its silver complex isomers at the UB3LYP/6-311+G(4/?)(for C, N, H)/SDD(for Ag) level against the bonding distances of Ag-N (in A). Energy units are kilocalories per mode. Frequencies scaled by a single factor of 0.9942 are given. Raman activities are given in units of A4 amu"1.
Ag-N
AE Mode 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37
Adenint5 Ad-N3 2.75 2.50 0 3.4 3.9 Wavenumbers /cm'1 1650 1651 1647 1624 1625 1624 1602 1599 1603 1505 1505 1506 1493 1492 1492 1424 1430 1433 1412 1406 1409 1356 1357 1357 1356 1348 1359 1319 1319 1318 1259 1259 1259 1235 1236 1237 1136 1139 1142 1073 1070 1072 1003 1005 1005 970 968 968 939 941 942 895 899 903 846 853 852 804 800 799 721 722 724 681 680 679 663 664 664 615 615 615 572 572 574 544 542 536 529 531 535 517 520 522 514 529 535 300 301 297 275 276 277 215 220 224 163 175 188 51 149 157 43 85 41 51 35 40
2.25 2.2
2.00 -7.8
Ad-N7 2.75 2.50 3.5 2.7
2.25 1.8
2.00 -7.3
1653 1623 1605 1507 1491 1435 1415 1357 1364 1314 1258 1236 1146 1071 1006 967 943 915 855 798 729 679 665 615 575 550 543 524 539 298 279 234 199 158 149 66 54/
1656 1619 1607 1509 1488 1437 1419 1357 1370 1303 1256 1233 1158 1071 1006 964 943 952 857 796 741 677 665 617 577 552 563 526 541 294 285 250 201 231 157 47 103/
1650 1624 1601 1508 1493 1425 1409 1358 1350 1318 1259 1232 1135 1086 1008 974 945 895 847 800 721 681 660 615 571 542 528 519 520 298 293 156 229 193 54 35 28
1652 1625 1598 1512 1491 1427 1414 1358 1352 1315 1259 1230 1137 1099 1010 977 964 895 850 797 721 681 657 619 572 548 529 522 527 300 302 159 271 205 154 70 35i
1652 1624 1595 1516 1490 1429 1419 1358 1351 1310 1256 1232 1143 1106 1019 978 995 895 848 796 725 681 654 627 573 552 533 524 530 303 307 257 287 205 145 51 97/
1652 1625 1601 1510 1492 1425 1411 1358 1352 1317 1259 1232 1136 1092 1009 975 951 895 849 799 721 681 660 616 572 546 529 521 524 299 301 158 249 202 90 66 37
Specific Raman band shift caused by mechano-chemical effect in tip-enhanced near-field Raman spectroscopy
97
The partially optimized geometries have imaginary frequencies, indicating transition states for the Ad-N3 and the Ad-N7. The calculated frequency shifts of the Raman bands having the highest intensities (Vs and V21 in Table 4) in the TERS spectroscopy and the calculated potential curves are plotted as a function of the bond distance in Fig. 10. As for the rest of the six bands observed in the TERS, it is difficult to analyze the frequency shifts quantitatively because the signal-to-noise ratio of these bands is not sufficient for the analysis.
Raman shift [cm –1]
1380 i oou
E o
(0
-
13701370
(a) (a)
1360 7 i-
a. - -&
_
13401340 1330, 1330 750 750, -
740 *
CO
730730 ( 720 720-
CO
-
13501350
c
E
ν88 v a
ν221 V 1
710710 •
Binding energy [kcal/mol]
700700
0) £
.—
o
-8
(b)
-4
—o--.
•
Ad-N3 Ad-N7
0 4 2
2.5
3
Bond distance [Å] [A]
∞
Fig. 10. (a) The calculated frequency shifts of two Raman bands (vg and V21) of the Ad-N3 and the Ad-N7 and (b) the calculated binding energies as a function of the bond distance for the Ag-N linkage.
As the metallic tip of the cantilever approaches the surface of adenine nanocrystal, the tip is at first subject to van der Waals attractive force and after passing through the equilibrium point, the tip receives a repulsive force. In our experiment, the atomic force which is balanced with the repulsive force is set at 1 ~ 5 pN/molecule as described before. When the bond distance of the Ag-N linkage is reduced by 10%, the repulsive force of 7 pN/molecule is derived from a harmonic oscillation of the displacement by 0.025 nm and the energy
98
H. Watanabe et al.
difference by 1.7 kcal/mol in the case of the Ad-N3, The repulsive force coincides with the atomic force set in our experiment. The ring breathing mode V21 of the Ad-N3 shows a significant shift towards a higher frequency as a function of the contracted bond distance between the silver atom and the N3 nitrogen of adenine. The frequency of the V21 mode is shifted upwards by 5 cm"1 when the bond distance is reduced by 10%. The frequency is shifted upwards by 17 cm~l when the bond distance is reduced by 20%. In contrast, the calculated frequency shift of the other band (Vg in Fig. 10) of the Ad-N3 is quite small. The frequency shifts of these two bands agree with those of the bands obtained by the TERS spectroscopy. On the other hand, the Ad-N7 shows only small frequency shifts for both of these two bands in Fig. 10. These differences in frequency shift suggest that interaction between an adenine molecule and the silver tip (e.g. the Ad-N3) may be one of the possible reasons why we see large enhancement effects in the TERS spectroscopy. In addition to the higher frequency shift, Raman band broadening is also observed in the TERS spectra. The line broadening as well as Raman frequency shift has been well reported in the high-pressure induced Raman spectroscopy [42-44]. Our TERS spectra caused by the pressurizing tip could be thought to observe the phenomenon which was similar to high-pressure Raman study. The line broadening occurs not only as a result of surface interaction, but also as a result of pressures caused by the silver tip. While the quantum chemical vibrational calculation can provide quantitative estimates of frequency shifts, its accuracy is dependent on the physical model. Here, we treated the silver tips as a simple atom, leading to some discrepancies between the calculation and the experiment. More practical models are required for getting a precise picture. For example, it has been reported that a cluster of four silver spheres provides a gigantic magnification of the electromagnetic field [45]. Silver ions have also shown a significant shift to higher frequency in the ring-breathing mode of pyridine [24]. 4. CONCLUSION Near-field Raman spectra of adenine nanocrystals have been measured using the local field-enhancement effect at the metallic tip. The smallest observable feature reaches 30 nm that exceeds the diffraction limit of the light. Intensity of Raman signals is amplified by a factor of more than 2.7 xlO-* times compared with the far-field Raman intensity. These TERS spectra of adenine, as a first approximation, exhibit spectral patterns analogous to that of the conventional SERS spectra of adenine adsorbed on silver nanoparticle. These spectral patterns involving anomalous enhancement and large frequency shifts of some specific Raman bands are attributed to chemical interactions between the molecules and the metallic tip.
Specific Raman band shift caused by mechano-chemical effect in tip-enhanced near-field Raman spectroscopy
99
In addition to the chemical interaction effect as well as the electromagnetic field enhancement effect, we found the third effect, a "mechanical" pressure effect, which can give rise to the spectral changes in the TERS spectroscopy. The TERS spectra of adenine, in detail, partially differ from the conventional SERS in vibrational frequencies. By means of normal mode analysis using the DFT calculations, we obtained vibrational frequency shifts for the transient states of the adenine-silver metal complexes. The calculated Raman spectra of the complexes could be made to agree with the TERS spectra of adenine by reducing the bond distance. Repulsive forces calculated from reduction of the bond distance between adenine and the silver atom were equal to the atomic force applied to the adenine molecule in our TERS experiment. All results support mat the active Raman shift occurs owing to the deformation of adenine molecules by the silver atoms of the metallic tip. In particular, the silver atoms presumably pressed against the N3 nitrogen of the adenine molecules. The phenomenon of shifting near-field Raman spectra caused by pushing molecules with AFM is used for a novel spectroscopic instrumentation of molecular analysis and identification at nanoscale. Spatial resolution of our proposed method should be given by AFM which perturbs individual molecules. Hence the technique has a possibility of achieving molecular resolution in vibrational spectroscopy, such as short fragments of DNA sequencing lying flat on a surface.
REFERENCES [I] Y. Inouye, N. Hayazawa, K. Hayashi, Z. Sekkat, and S. Kawata, Proc. SPIE., 3791 (1999) 40. [2] N. Hayazawa, Y. Inouye, Z. Sekkat, and S. Kawata, Opt. Commun., 183 (2000) 333. [3] R. M. Stockle, Y. D. Suh, V. Deckert, and R. Zenobi, Chem. Phys. Lett., 318 (2000) 131. [4] M. S. Anderson, Appl. Phys. Lett., 76 (2000) 3130. [5] Y. Inouye and S. Kawata, Opt. Lett., 19 (1994) 159. [6] R. Bachelot, P. Gleyzes, and A. C. Boccara, Opt. Lett., 20 (1995) 1924. [7] J. Koglin, U. Ch. Fischer, and H. Fuchs, Phys. Rev. B, 55 (1997) 7977. [8] B. Knoll and F. Keilmann, Nature, 399 (1999) 134. [9] S. Kawata (ed.), Near-field Optics and Surface Plasmon Polariton, Springrer-Verlag, Berlin Heidelberg, 2001. [10] U. Ch. Fischer and D. W. Pohl, Phys. Rev. Lett, 62 (1989) 458. II1] H. Furukawa and S. Kawata, Opt. Commun., 148 (1998) 221. [12] L. Novotny, E. J. Sanchez, andX. S. Xie, Ultramicroscopy, 71 (1998) 21. [13] J. Jersh, F. Demming, L. J. Hildenhagen, and K. Dickmann, Appl. Phys. A: Mater. Sci. Process., 66 (1998) 29. [14] R. K. Chang and T. E. Furtak, Surface Enhancement Raman Scattering, Plenum, New York, 1982. [15] N. Hayazawa, Y. Inouye, Z. Sekkat, and S. Kawata, Chem. Phys. Lett., 335 (2001) 369. [16] N. Hayazawa, Y. Inouye, Z. Sekkat, and S. Kawata, J. Chem. Phys., 117 (2002) 1296. [17] H. Watanabe, N. Hayazawa, Y. Inouye, and S. Kawata, J. Phys. Chem. B, 109 (2005) 5012.
100 100
H. Watanabe et al.
[18] N. Hayazawa, T. Yano, H. Watanabe, Y. Inouye, and S. Kawata, Chem. Phys. Lett., 376 (2003) 174. [19] J. Wiorkiewicz-Kuczera and M. Karplus, J. Am. Chem. Soc, 112 (1990) 5324. [20] M. Majoube, Ph. Millie, P. Lagant, and G. Vergoten, J. Raman Spectrosc, 25 (1994) 821. [21] M. J. Nowak, L. Lapinski, J. S. Kwiatkowski, and J. Leszczynski, J. Phys. Chem., 100 (1996) 3527. [22] B. Giese and D. J. McNaughton, J. Phys. Chem. B, 106 (2002) 101. [23] A. Otto, J. Timper, J. Billmann, G. Kovacs, and I. Pockrand, Surf. ScL, 92 (1980) L55. [24] A. Otto, J. Bilhnann, J. Eickmans, U. Ertuerk, and C. Pettenkofer, Surf. Sci., 138 (1984) 319. [25] D. Y. Wu, B. Ren, Y. X. Jiang, X. Xu, and Z. Q. Tian, J. Phys. Chem. A, 106 (2002) 9042 [26] A. Vivoni, R. L. Birke, R. Foucault, and J. R. Lombardi, J. Phys. Chem. B, 107 (1999) 5547. [27] H. Watanabe, Y. Ishida, N. Hayazawa, Y. Inouye, and S. Kawata, Phys. Rev. B, 69 (2004) 155418. [28] N. Hayazawa, Y. Inouye, and S. Kawata, J. Microscopy, 194 (1999) 472. [29] A. D. Becke, J. Chem. Phys., 98 (1993) 5648. [30] C. Lee, W. Yang, and R. G. Parr, Phys. Rev. B, 37 (1988) 785. [31] P. Fuentealba, H. Preuss, H. Stoll, and L. v. Szentpaly, Chem. Phys. Lett., 89 (1982) 418. [32] M. Kaupp, P. v. R. Schleyer, H. Stoll, and H. Preuss, J. Chem. Phys., 94 (1991) 1360. [33] T. Leininger, A. Nicklass, H. Stoll, M. Dolg, and P. Schwerdtfeger, J. Chem. Phys., 105 (1996) 1052. [34] M. J. Frisch, G. W. Trucks, H. B. Schlegel, G. E. Scuseria, M. A. Robb, J. R. Cheesemann, V. G. Zakrzewski, J. A. Montgomery Jr., R. E. Stratmann, J. C. Burant, S. Dapprich, J. M. Millam, A. D. Daniels, K. N. Kudin, M. C. Strain, O. Farkas, J. Tomasi, V. Barone, M. Cossi, R. Cammi, B. Mennucci, C. Pomelli, C. Adamo, S. Clifford, J. OchtersM, G. A. Petersson, P. Y. Ayala, Q. Cui, K. Morokuma, D, K. Malick, A. D. Rabuck, K. Raghavachari, J. B. Foresman, J. Cioslowski, J. V. Ortiz, A. G. Baboul, B. B. Stefanov, G. Liu, A. Liashenko, P. Piskorz, I. Komaromi, R. Gomperts, R. L. Martin, D. J. Fox, T. Keith, M. A. Al-Laham, C. Y. Peng, A. Nanayakkara, M. Challacombe, P. M. W. Gill, B. Johnson, W. Chen, M. W. Wong, J. L. Andres, C. Gonzalez, M. Head-Gordon, E. S. Replogle, and J. A. Pople, GAUSSIAN98 Revision A.9, Gaussian Inc., Pittsburgh, PA, 1998. [35] G. Schaftenaar, MOLDEN, The University of Nijmegen, Nijmegen, Netherlands, 1991; G. Schaftenaar and J. H. Noordik, J. Comput. -Aided Mol. Des., 14 (2000) 123. [36] Y. Yamakita, LXVIEW, The University of Tokyo, Tokyo, Japan, 1995; Y. Yamakita and M. Tasumi, J. Phys. Chem., 99 (1995) 8524. [37] Y. Okamoto, VLX, Fuji Photo Film. Co., Ltd, Kanagawa, Japan, 2001; H. Watanabe, Y. Okamoto, K. Furaya, A. Sakamoto, and M. Tasumi, J. Phys. Chem. A, 106 (2002) 3318. [38] M. Kerker, D. S. Wang, and H. Chew, Appl. Opt, 19 (1980) 4159. [39] K. Hoogsten, Aeta Crystallogr., 12 (1959) 822. [40] R. F. Stewart and L. H. Jensen, J. Chem. Phys., 40 (1964) 2071. [41] C. W. Bauschlicher Jr. and H. Partridge, J. Chem. Phys., 103 (1995) 1788. [42] J. R. Ferraro, Vibrational Spectroscopy at High External Pressures, Academic Press Inc., New York, 1984. [43] R. A. Crowell and E. L. Chronister, Chem. Phys. Lett., 195 (1992) 602. [44] S. A. Hambir, J. Franken, D. E. Hare, E. L. Chronister, B. J. Baer, and D. D. Dlott, J. Appl. Phys., 81 (1997) 2157. [45] N. Liver, A. Nitzan, and J. I. Gersten, Chem. Phys. Lett., 111 (1984) 449.
Handai Nanophotonics, Volume 2 S. Kawata and H. Masuhara (Editors) © 2006 Elsevier B.V. All rights reserved.
101 101
Chapter 6
Single molecule sensitivity in surface enhanced Raman scattering using surface plasmon M. Futamata", Y. Maruyamab a
Nanoarehitectomcs Research Center, National Institute of Advanced Industrial Science and Technology, 1-1-1 Higashi, Tsukuba 305-8562, Japan
b
Tsukuba Research Laboratory, Hamamatsu Photonics K.K., Tsukuba 300-2635, Japan 1. INTRODUCTION It is crucial to elucidate the nature of constituents at an atomic or molecular scale in order to realize a bottom-up nanotechnology, e.g. a single molecule device or a nano-chemieal factory. Surface enhanced Raman scattering (SERS) available under ambient conditions even at the solid/liquid interfaces attracts increasing interest according to the potentialities of single molecule sensitivity (SMS) and also nanoscale spatial resolution when combined with near-field microscopy. SERS is a well-known phenomenon for more than 30 years [1,2] that Raman signal from adsorbates on roughened metal surfaces is enhanced by a factor of 10 — 106 due to excitation of surface plasmon polariton (SPP) on roughened metals and/or due to the "first layer" enhancement including charge transfer resonance between adsorbates and metals. The enhancement factor obtained with a conventional Raman spectroscopy is a population-averaged value for enormous molecules sitting on numerous Ag particles with various shapes and sizes, and thus obviously insufficient to yield single molecule detection (SMD) in SERS. Recent progress in scanning probe microscopy (SPM) in addition to a highly-sensitive charge-coupled device (CCD) detector enables us to observe extremely weak spectral signal from an individual metal particle or even from single molecule. Accordingly, several groups reported vast enhancement in SERS, e.g. 1014, corresponding to SMS by probing only 'hot' particles with appropriate shapes and sizes [3-7], whereas mechanism of the enormous enhancement is still obscure. Obviously this is because there are no straightforward methods to elucidate electronic and vibrational spectra of individual molecules at local environment on metal nanostructures, e.g. spatial resolution of near-field spectroscopy is still at ca. 10 nm relevant for such
102 102
Y. Maruyama M. Futamata and Y.
purposes. Some groups claimed the importance of junction for SMS-SERS based on the fact that only touching metal nanoparticles give enormous enhancement and 'blinking' of the signal that intensity suddenly and repeatedly alters with time as predicted by theoretical simulation for the local electric field. In contrast, Nie et al. observed even isolated Ag or Au nanoparticles showed the blinking, which were immobilized on the substrate and followed by activation with halide ions [8,9]. As is well known, valuable information is extracted from Raman spectra such as a molecular structure, orientation or interaction with neighboring species. In contrast, fluorescence spectroscopy precedent as a SMD method yields less informative broad spectra albeit useful to identify individual molecules. Thus it is challenging to establish the SMD method with SERS in terms of application to various fields, such as elucidation of elementary reaction process at solid/liquid interfaces. The followings are commonly observed concerning the vast enhancement in SERS [3-7]: (1) only a small number of 'hot particles' emerge with prominent enhancement when surface coverage of adsorbates is between ca. several tens and one hundred per each Ag particle. (2) At lower surface coverage < ca. 1 molecule/particle, the blinking of Raman signal is observed for merely trace amounts of Ag particles with dye or other biomolecules like hemoglobin or DNA bases. The blinking is believed to be a single molecule phenomenon, since it is not plausible for many molecules to move or to interchange their orientation at the same time. Although there are no direct methods to prove it, various supporting evidences were reported. For instance, Nie et al. reported intermittent properties, i.e. complete ON/OFF features of the blinking rationalized only by motion of individual molecules [8, 9]. Kneipp et al. found Poisson distribution of the SERS intensity for extremely low surface coverage [4], which is quantized intensity corresponding to a discrete number of adsorbed molecules in the sampled volume, e.g. null, one or two molecules, in contrast to Gaussian distribution at much higher coverage. It should be noted that apparent SERS intensity fluctuation may occur during photochemical reaction of organic adsorbates using even rather weak laser irradiation under electronic resonances [10]. Therefore, experiments of the blinking in SERS signal should be performed under precisely controlled conditions to suppress such phenomena. Recently, Weiss and Haran [11] reported that the rate of spectral fluctuation is proportional to laser intensity and suggested a nonthermal process such as photo-induced desorption of adsorbates. Bosnick et al. [12] reported the SERS scattering intensity fluctuates due to motions of the molecule in and out of the hot spot and is highly localized around it. Also, thermal activation of the blinking was evidenced by the temperature dependence that the intensity fluctuation is suppressed at 77 K as described in Section 3.3 [13], which is compatible with the photo-induced molecular motion for particular molecules with appropriate electronic state. In addition, local electric field intensity evaluated by a numerical method, e.g. Finite Difference Time Domain (FDTD) method, yields vast enhancement of > 1010 at the junction
103 Single molecule sensitivity in surface enhanced Raman scattering using surface plasmon 103
of touching Ag particles with various sizes and shapes as in Section 3.2 [14] in contrast to modest enhancement of 104-10s at other ordinary sites or on isolated particles. Similar results were obtained with using different numerical simulations, e.g. finite element method [15], multipole expansion and boundary charge method [5], discrete dipole analysis [16]. Thus, the observation of frozen molecules at low temperature indicates that the blinking is due to thermal process, most probably due to thermal diffusion of adsorbates molecules from the junctions to other ordinary sites on Ag particles. Furthermore, in order to verify the above attributions, it is crucial to specify the adsorbed site of individual dye molecules on Ag nanoparticles. Recently significant spectral changes were observed in elastic scattering during inactivation process of SERS from hot or blinking Ag particles [17, 18]. Adsorbed molecules sitting at the junction of Ag particles could account for vast enhancement and distinct spectral variation in elastic scattering. In the former paper [17], two-dimensional FDTD method (FDTD-2D) was utilized to evaluate the effect of dye adsorption to an Ag nanowire in the elastic scattering spectra. Although the essential features were obtained with such calculations, three-dimensional analysis provides more accurate electromagnetic distribution, especially precise wavelength dependence in the near-field and far-field for the complicate nanostructures. Accordingly, FDTD-3D adopted here in Section 3.4 yields detailed information on electromagnetic coupling between dye and localized surface plasmon (LSP) of Ag nanoparticles that determines the excitation profile of SERS. At last, we tried to fabricate the metal nanostructures with SMS-SERS, i.e. trigonal silver nanoarray with sharp edges (Section 3.5). In our knowledge, silver or gold nanoparticles prepared by chemical reduction of AgNOs or HAuCLj have been exploited for most of single molecule detection or vast enhancement in SERS experiments [2-9, 11-13]. This is probably due to experimental feasibility to explore the optimum metal nanoparticles with particular morphology or sizes. However, it has not been established to fabricate such optimum nanoparticles with sufficiently high yield by the chemical reduction, since the process is only macroscopically controlled through the reaction temperature or mixing speed. Indeed, variety of silver particles with different shapes and sizes are formed by means of citrate salt as a reducing chemical, while NaBfti provide homogeneous, isolated spherical silver particles with modest enhancement. Obviously, only scarce metal nanoparticles show the blinking among huge number of particles prepared by the citrate method [2-9, 11-13]. Chemically etched metal surfaces could provide much higher occurrence for the blinking in SERS as suggested by Doering and Nie [9]. On the other hand, SERS spectrum is strongly dependent on the nanoscale morphology at the active sites that are not precisely controlled with the etching conditions. Therefore, more efficient fabrication method for the metal nanostructures with optimum morphology for SMS-SERS is requisite. Two dimensional array formation from polystylene nanospheres has been innovated by Nagayama et al.
104 104
M. Futamata and Y. Maruyama
[19-23]. Then, nanosphere lithography (NSL) with subsequent evaporation of metal was pioneered by Van Duyne et al. [24-28] to control the LSP resonance with respect to the efficient SERS-active substrates and optical sensors. Of late, Van Duyne group reported detailed excitation profiles of SERS under plasmon resonances for the trigonal metal nanostructures, e.g. pronounced enhancement > 109 observed for Fe(bpy)3-(PF6)2 including the electronic resonance effect [29], which proves the prevalence of NSL. In addition, the trigonal nanostructure, in principle, has sharp edges which possibly yield enormous electric field under the LSP resonance as anticipated by the numerical simulation [5, 14, 15]. After the optimization of the preparation condition to give sufficiently sharp nanostructures, it could be quite valuable to fabricate nanodevices with SMS-SERS. 2. EXPERIMENTS AND NUMERICAL ANALYSIS 2.1. Experimental set up for SERS measurement 2.1.1 Ag nanoparticles preparation AgNO3 was chemically reduced using excess amount of sodium citrate [30], and mixed with adsorbates such as rhodamine 6G (R6G) or DNA bases and NaCl in aqueous solution with particular concentration to control the surface coverage. Then, the Ag particles were immobilized onto 3-aminopropyl trimethoxysilane (APTMS) covered Si substrate using a spin-coater. Here NaCl substitutes residual citrate anions on the Ag particles with Cl" anions, which is essential for cationic dye to adsorb onto the Ag particles with electrostatic force. It was evidenced by an occasional appearance of SERS bands from citrate anions instead of dye at modest concentration of NaCl. Only a countable number of Ag particles are located in each sampled area on the Si substrate, which are separated with a grid of 50 nmx50 urn, by adjusting a concentration of Ag-dispersed solution. It enables us to detect Raman spectra from individual Ag particles, since a separation (ca. several micrometer) between neighboring Ag particles is much larger than the laser beam diameter at the sample position (§ ca. 1 \im). Additional optics for external configuration in Raman spectroscopy, a highly sensitive CCD camera with an image intensifier, a notch filter to observe Raman images and a precisely adjustable X-Y mechanical stage (with an accuracy of 1 |i,m in both directions) were installed to a conventional microscope (Renishaw © Ramascope, see Fig. 1). These are actually crucial to detect SERS images and SERS spectra from individual Ag particles between room temperature and 77 K, while compensating mechanical drift of sampled positions. Elastic scattering images and spectra of the individual Ag particles were observed with a CCD camera using white light source (Xe lamp) and a mask for dark field illumination as shown with the dotted line in Fig. 1, whereas Raman images were obtained under an external geometry using Ar+ laser at 488 nm (full line). Back-scattering geometry was used to measure Raman spectra with the same Ar+ ion laser through a microscope objective and a polychromator
105 Single molecule sensitivity in surface enhanced Raman scattering using surface plasmon plasmon 105
containing a set of notch filters and a CCD detector (see broken line). Sufficiently weak laser power, i.e. 1 ^,W/|xm2 for Raman imaging and 70 |iW/(im2 for Raman spectra was used, which did not yield any indication of photochemical reaction of constituents. With using a grid-like marker, we can correlate each Ag particle in the atomic force microscope (AFM) images, optical (LSP scattering and Raman) images and spectra, which enables us to explore the optimum Ag nanostructures. 2.1.2. Nanosphere lithography (NSL) Polystyrene (PS) nanosphere (diameter of 540 run, Duke) was employed to fabricate the metal nanostmcture using NSL developed by Van Duyne [24-28]. Glass substrate can be treated in weak alkaline solution to increase affinity with the PS nanosphere. Only a small amount of the PS nanosphere solution (0.5 ul) spread onto the glass substrate is sandwiched with another glass via a spacer at one end (a wedge angle of 5 - 6 °) [31]. Then, the sample is dried in atmosphere with relative humidity of ~75 % at 295 K. After drying for about 6 hours, the PS nanospheres formed a closest packed monolayer with a typical ordered region of several mmxl mm. The glass substrate with PS nanospheres is then mounted Raman/Optical Image age
PC CCD
Raman spectra
Monitor Notch filter
Xe lamp
Polychromator CCD "
/2 plate λ XI2 Mobile mirror
PC
Ar++ laser, 488 nm Ar
Interference filter
Fig, 1. Experimental setup for optical imaging, LSP extinction spectra, Raman imaging, and Raman spectral measurement based on a Raman microscope. Ag particles with adsorbates are immobilized onto the APTMS coated glass or Si substrates located under the objective (x 50).
into a conventional vacuum chamber to use as a deposition mould for the metal nanostructure. Silver is deposited onto the substrate with rate of ~0.4 nm/s under 10'7 Torr until the thickness of 50 nm. After silver deposition, the sample (silver/PS/glass substrate) is sonicated in an ethanol solution for 3 minutes to remove PS nanosphere, while retaining the silver nanostructure on the substrate. Continuous silver films (50 nm in thickness) are evaporated on the bare glass surface with the same condition to exploit as a reference. In addition, the continuous silver films are annealed at 473 K for 2 hours in the vacuum chamber
106 106
M. Futamata and Y. Maruyama
to evaluate actual efficiency of the trigonal silver nanostructures compared to flat, smooth silver films. 2.2. Numerical analysis of the local electric field and scattering cross section for metal nanostructures Maxwell equations are solved for spheres or ellipsoids with or without substrate [2, 32-34]. However, analytical solutions have not been obtained for other complicated structures such as triangular, tetrahedral particles with/without unsymmetrical protrusions or pit, because isolation of variables in the differential equations is substantially difficult. Therefore, the numerical simulation such as the FDTD method is valuable, which transfers the differential equations to difference equations. The stationary solutions can be obtained at given positions for the initial electromagnetic field with various wavelengths and propagating directions. In the FDTD method, the nanostructure is surrounded by virtual boundaries with an appropriate size, and inside of the area is separated into small rectangular meshes with a particular size (AxxAy meshes, see Fig. 2, here two-dimensional expression is given to simplify the explanation). Namely, the metal particles and surroundings are depicted as a collection of these small meshes with a particular size and dielectric properties [35]. In this system, curl in the Maxwell equations is given by the following equations for TM (transverse magnetic) field [36].
Hx{iJ+1/2)
J- :•-!-•
Ay
• • ; • • ; • -
Hy(i + 1/2,j)
1\ J)
i
Ez(i + 1J)
E2
i Fig. 2. Schematic image of a sample system and coordinate used in the FDTD calculation.
107 Single molecule sensitivity in surface enhanced Raman scattering using surface plasmon plasmon 107
-C^y {Uj){Hn-ul{Uj +1 /2) - H;-m{i,j -112)} here 1 a(iJ)At
U)
c
(in-
I
,j)=n;-U2a
Here o(i, j), e(i, j) and \k are conductivity, dielectric constant, and magnetic permeability, respectively. According to these equations, the local electromagnetic field En"' and Hn"1/2 at orthogonal coordinates x = x(i), y = y (j) and at time t = t n l , t'm are calculated, and then the H"+I/2 and En as a sequential and time-evolutional response to the incident electromagnetic field. The first order Mur scheme or PML (perfectly matched layer) was used as an absorbing boundary condition in conjunction with a recursive convolution method for metals with prominent dielectric dispersion in wavelength region studied here [36]. In the FTDTD-2D simulation in Section 3.2, metal nanostructures were supposed to be a nanowire with infinite height (length) perpendicular to the cross section with various shapes, e.g. circle, triangle or ellipsoid. Local electric field was evaluated using the maximum field intensity on metal surfaces with using a sufficiently small mesh size, e.g. 0.25 nmxO.25 nm or 0.1 nmxO.l nm, where variation of the field is rapid at junctions or surfaces of particles, contrary to less dense mesh size for outside the particles with 1 nmxl nm, typically. Such inhomogeneous meshes are quite useful and efficient to obtain the accurate values with rather short time especially in three-dimensional evaluations. Actual calculation was performed for metal nanoparticles with various sizes, shapes and ordered structures. We noted that the accurate wavelength dependence of the LSP resonance could not be reproduced in the 2D calculation, as the height of the nanostructures is not contained. However, it was also confirmed that accurate local field intensity is obtained for parallel polarization to the X-Y plane with detailed mesh sizes in much shorter processing time compared to the 3D simulation.
108 108
M. Futamata and Y. Maruyama
Scattering cross section, far-field response, in Section 3.4 from Ag nanoparticles with or without R6G adsorbates was evaluated at various wavelengths with the three dimensional FDTD (FDTD-3D) method using dielectric constants of constituents [37, 38]. Dielectric constants of the dye were expressed in a damped harmonic oscillator model, e = 1 + S / (coo2 - (B2 - ieaT), here an oscillation strength (S) of 5xl0 31 sec"2, damping (T) of 2.5xl0 14 sec"1, and resonance wavelength (COD) of 3.2xlO15 sec"1 (XQ = 589 nm) based on the experimental observations for rhodamine [39]. In order to discus the coupling efficiency between the LSP and dye absorption, the resonance wavelength of dye was tuned between 400 nm and 700 nm. We adopted the various model structures such as isolated spheres, spheroid or adjacent spherical Ag particles with/without dye molecules. Yee cells are built from 160x120x120 meshes (2.3xlO6 meshes) with 2 nm/mesh along x, y and z directions using PML (perfectly matched layer) absorbing conditions [38], as larger numbers of meshes or smaller sized meshes do not yield significant differences in scattering cross section. An incident light as a plane wave with linear polarization irradiates the cell at various wavelengths. Typically, one or two hours are necessary to yield a converged solution using a CPU (Pentium 4, 3.0 GHz) and Windows XP, which is roughly ten times longer than in 2D calculation. Finer meshes at the junction do not give significant differences in far-field scattering, whereas providing much accurate near-field distribution compared to poorly resolved meshes. Therefore, the mesh sizes of 2 nm are adopted through the present evaluation. 3. RESULTS AND DISCUSSION 3.1. Hot particles In SERS Optical images in the bright-field, Raman and topographic (AFM) images for the same Ag particles were observed as shown in Figs. 3a-3d using the facility in Fig. 1. At surface coverage of ca 300 R6G molecules/particle, only several Ag particles showed prominent SERS signal (bright spots in Fig. 3b) among many particles in the sampled area (ca. 20 umx20 um). All of the hot particles consist of Ag aggregates with a typical size of 1 um or touching several particles as clearly visualized in fee AFM images (Figs. 3c-3d). At the surface coverage of ca. 300 molecules/particle, the enhancement factor was estimated to be about 2x107 by comparing the Raman scattering intensity from the hot particles with that for bulk solution. The LSP extinction spectra are apparently broadened and extended to longer wavelength upon coalescence of Ag particles [40] due to the overlap and coupling of the LSP from different Ag particles. For example, one of the hot particles composed of two touching Ag spheroid with a similar size of ca. 210 nmxl60 nmxl20 nm (height) gives the main peak at 470 nm as well as shoulder peaks at ca. 490 nm and ca. 510 nm. Then, the Ar+ laser at 488 nm can resonantly excite the LSP to enhance Raman signal from adsorbates on Ag aggregates. This is in contrast to the observation that detectable SERS signal was not observed for isolated Ag particles that possess fairly sharp LSP bands at ca. 400 nm [40]. We also noted that
109 Single molecule sensitivity in surface enhanced Raman scattering using surface plasmon plasmon 109
slightly elongated Ag spheroid gives the LSP extinction at longer wavelength, which is in resonance with the same laser. Nevertheless, isolated particles are not hot. Accordingly, the junction of touching particles is crucial to give prominent enhancement in SERS. In addition, the Ag aggregates in Figs. 3c-3d are not always hot because of the polarization dependence of the local field under LSP resonance. (a) Optical image
18tim x 1B|iJTl
(d) Expanded image oftcl
Fig. 3. Correlation of optical image (bright-field) (a), Raman image (b) and AFM images ((c) and (d)) of R6G/Ag particles. The same area in the same sample was measured in (a), (b) and (c) (18 jimxl 8 |jm).
Fig. 4a shows hot particles consisting of two touching particles. Behind the bright spot, another particle is located, which is slightly shifted along the lateral direction, hi this case, the polarization parallel to the touching axis (lateral direction in the figure) shows significantly stronger intensity, whereas perpendicular polarization gives only humble (see Figs. 4b-4c). The polarization dependence of the SERS enhancement was confirmed for various touching particles with different configurations. These discussions are supported by theoretical simulation for the local electric field on Ag particles using the FDTD method as follows.
110 no
M. Futamata and Y. Maruyama (a) AFNI (b) Raman spectrum
Touching axis
•++• Pol.
(c-1) Raman (V-pol.)
(c-2) Raman (P-pol.) 1600
1400
1200
1000
Stokes shift (cm')
Fig. 4. Polarization dependence of SERS signal: (a) AEM image, (b) Raman spectra, and (c) Raman images. Behind a bright spot in (a), another particle is located, which is slightly shifted to the lateral direction. Vertical and parallel polarization to the touching axis gave distinct enhancement in SERS.
3.2. Local field evaluation on the Ag nanoparticles 3,2.1 Relevance of the FDTD method in SERS system At first, the relevance of the FDTD method was confirmed for the electromagnetic field around the metal particles in the near-field and far-field calculation. For this purpose, we adopt an Ag sphere with 10 nm in radius placed above the Ag flat substrate with a gap size of 0.5 nm (see Fig. 5), as the analytical solution was obtained in the Maxwell equations for this case using bispherical coordinates [32]. A plane wave is incident with the angle of 45 ° and p-polarization (see the inset of Fig. 5). Consequently, quite similar local field enhancement of4.2xl04 at 450 nm and 1.29xl04at400nmwas obtained by the FDTD method at the center between the Ag particle and substrate compared to the values of 4.2xlO4 at 450 nm and L3X104 at 410 nm obtained by analytical solutions [32]. Essentially the identical results were obtained for various gap sizes in these different calculations. While these were obtained by the two-dimensional calculation for p-polarization to the substrate, similar results were given by the three-dimensional calculation as well as for s-polarization. Thus, we can safely use the FDTD method to evaluate the local electric field in the vicinity of the metal particles. Note that the vast enhancement of 1010-10u for Raman scattering [41] was predicted for the gap size of 0 nm, which is comparable with those for touching Ag particles with various shape and sizes as described in the following section. Moreover, this result suggests tremendously large enhancement for Raman scattering from adsorbates at the gap between Ag particles and smooth Ag films, where the propagating SPP is excited using a prism.
111 Single molecule sensitivity in surface enhanced Raman scattering using surface plasmon plasmon 111
(a)
,
4x10*
^
I
Metiu
5x10 J_
FDTD
3 .E
hi r=10rim
" tj /
oP"*
S 2
I
jo.
^f _ |-
UJ
1
04
-M-p
J
400
\
|
I A
!.
f•«-4-
500 600 700 Wavelength (nm)
800
400 500 Wavelength (nm)
600
Fig. 5. Electric field intensity at the gap between the Ag sphere and substrate: (a) results by the analytical solution, (triangles) and FDTD (filled circles) for the Ag sphere with r = 10 run, and gap size of 0.5 nm, and (b) by FDTD for various gap size. Inserted figure is a schematic drawing of the sample configuration.
3.2.2 Local field and scattering cross section for the metal particles with respect to SERS activity (1) Isolated Ag circular and ellipsoidalparticles Here we do not hold any substrates beneath the metal particles to evaluate the electric field in the near-field and far-field for the isolated metal particles. In our real experiments, a glass substrate was used to immobilize the metal particles. However, dielectric substrate like a slide glass does not give significant changes in electric field intensity in contrast to the above result for metal substrate, while LSP extinction peak slightly shift to longer wavelength by a few tens nm, Accordingly, the dielectric substrate was extinguished to reduce size of the sample cell for saving computation time. Scattering cross section for an isolated Ag circular cylinder shows the peak at ca. 370 nm irrespective of size between 10 - 80 nm (Fig. 6a). The peak is clearly arisen from the LSP excitation, of which wavelength accords with the analytical solutions for spherical particles with the same size [42]. In contrast, two distinct peaks were obtained for triangular structure at 430 and 500 nm for the right-angle (80nmx40nm), or 400 nm (shoulder) and 450 nm for the equilateral right-angle (80nmx80nm) (see Fig. 6c). The peak for ellipsoid slightly shifts to longer wavelength with increasing the aspect ratio, e.g. 370 nm (2:1) shifts to 400 nm (3:1) or 430 nm (4:1, see Fig. 6b), which is much smaller than the observed values, i.e. 565 nm for the diameter (D) 95 nm and height (H) 48 nm (2:1) to 782 nm for D145 nm and H50 nm (3:1). For tetrahedral samples with a
112 112
M. Futamata and Y. Maruyama
size of 80 nm (width, W) x 80 nm (H) at the cross section, the LSP peak was obtained at 380 nm irrespective of polarization direction quite similar to those for circular particles as depicted in Fig. 6d In contrast to the results in two-dimensional simulations, (a)
600
0c 400 500 600 700 800 Wavelength (nm)
400 500 600 700 800 Wavelength (nm)
400
400
! 80 I / 4 5 P
80
400 500 600 700 800 Wavelength (nm)
D--
400 500 600 700 800 Wavelength (nm)
Fig.6. Scattering cross section fa isolated Ag particles with various shape and sizes: (a) circular, (b) ellipsoidal, (c) triangular and (d) tetragonal tubes with different shape, sizes and polarizatioa Hie symbols of square, triangle, and circle in (b) correspond to ellipsoidal particles with the aspect ratios of 4:1,3:1,2:1 respectively, for x- (filled) and y-polarized light (open).
much larger red shift of the LSP peak was obtained by the three-dimensional calculatioa As shown in Fig. 7a, the scattering cross section (SCS) peak shins from 380 nm (for sphere tx = r2 = 40 rim) to 450 nm (2:1 spheroid, ra = 40 nm, rb = rc = 20 nm) and 580 nm (4:1, r a =40 nm, rb = rc = 10 nm), which is consistent with the experimental
113 Single molecule sensitivity in surface enhanced Raman scattering using surface plasmon plasmon 113
date. In addition, these results fairly accord with analytical solutions for surface mode frequencies determined by geometrical factors [42]. For instance, the resonance condition for an ellipsoidal cylinder is given by eV Em ~ -1 at Xi = %i ~ 400nm, which splits into two branches for a prolate spheroid; eV Em «-1 at %\« 400 nm and eV £ „ , « -1 at X.2 » 400 nm (here, e', Em, X\ and Xa denote dielectric constant (real part) of particles, media, first and second resonance wavelength for the surface modes, respectively). It suggests that we should utilize three-dimensional simulations to characterize the LSP extinction spectra and local electric field for real metal particles and then to explore the optimum nanostructure. Nevertheless, two-dimensional calculation gives valuable insight into the vast enhancement at the junction as described later. The local electric field on Ag circular cylinder surfaces shows the maximum intensity (G) of 10 - 1 5 at ca. 380 nm for different sizes, while three peaks are obtained for right-angle triangular cylinder (80 nmx40 nm) at 380 nm (G = 180), 430 nm (G = 370), 500 nm (300) with different polarizations (see Figs. 8a, c, 9a, b, d). The equilateral right-angle triangular shows the maximum (G = ca. 500) at 430 nm with a shoulder at ca. 380 nm (Figs. 8c and 9c).
o.o 400
500 600 700 800 Wavelength (nm)
400
500 600 700 800 Wavelength (nm)
Fig. 7. Scattering cross section for an isolated ellipsoidal particle with different aspect ratio (a), and the maximum electric field for touching circular particles (b) obtained with the FDTD-3D simulation, to (a), results for a, b-polarized light are denoted with open and filled circles, although they are almost identical. Local electric field by FDTD-2D is also shown in (b, open circle).
114 114
M. Futamata and Y. Maruyama
400 500 600 700 800 Wavelength (nm) 500 -
400 500 600 700 800 Wavelength (nm)
400 500 600 700 800 Wavelength (nm)
Fig. 8. Maximum electric field for isolated Ag particles: (a) circular tabes with radii of 10,40,80 nm, (b) ellipsoidal particles with different aspect ratio between 2:1,3:1,4:1 for fixed long axis (40 nm), (c) triangles with different shape and (d) tetragonal particles (80 nm (W) x 80 nm (H)) with different polarization. Polarization parallel to the long axis was used for ellipsoidal particles.
With increasing the aspect ratio of ellipsoidal particles, where the longer axis is fixed to 40 nm and shorter axis is changed from 40 nm to 10 nm, significant spectral changes are not observed (as drawn in Fig. 8b). An isolated tetragonal particle (80 nm (W)x80 nm (H)) shows rather modest enhancement of ca. 110 at 380 nm. Thus, only triangular particles give vast enhancement at the sharp edge, while only modest enhancement was predicted for isolated circular, ellipsoidal and tetragonal particles. In addition, wavelength dependence of the scattering cross section is similar to that of the local field maximum for isolated particles. Slight differences of the peak wavelength
115 Single molecule sensitivity in surface enhanced Raman scattering using surface plasmon 115
and width, e.g. for triangular see Figs. 6c and 8c, can be explained based on the localization of the LSP field at the edge. The scattering cross section is given by the shape-averaged far-field intensity contributed from the entire Ag surfaces. Accordingly, a particular site like a sharp edge of triangular may not be dominant In contrast, the electric field maximum is determined by the local structure and the LSP resonance at the particular site, e.g. the enhancement is confined within a few nm from the edge, although these are not completely distinguished (see Figs. 9c-d). Thus the observed discrepancies between the resonant Rayleigh scattering and the SERS excitation [6] are rationalized. 110
-10
Fig. 9. Spatial distribution of the electric field on isolated Ag particles at the peak wavelength; (a) circular cylinder (r = 40 nm), (b) circular cylinder (r = 80 nm), (c) equilateral right-angle triangular (80 nm x 80 nm), (d) right-angle triangular particles (80 nm x 40 nm). Polarized excitation light was used at 380 nm for (a), (b), and at 430 nm for (c, with x-pol.) and (d, with y-pol.). See also Fig. 8. Electric field is shown as an amplitude enhancement relative to the incident field.
(2) Local field maximum and scattering cross section for two touching Ag particles Distinct spectral features were obtained for the local electric field maximum compared to the scattering cross section [14] as summarized in Figs. lOa-d. This is again because the local field maximum is determined by the LSP resonance and local nanostructure, whereas SCS is a shape-averaged far-field intensity contributed from the entire surface of Ag particles. Several peaks appear at 430 nm, 480 nm, 510 nm and 700
116 116
M. Futamata and Y. Maruyama
nm for touching circulars in addition to the original one at 370 nm for me isolated particle. Note mat much larger electric field G = 500, which is in the SMS level, is formed at 480 nm for the polarization parallel to the touching direction in contrast to modest values for isolated circular cylinders. Even for a vertical polarization, prominent factor of ca 270 is given at 440 nm. In addition, the maximum peak shifts to longer wavelength with the particle size, e.g. 410 nm at r = 10 nm, 520 nm at r = 80 nm (Fig. 11). 600
.... \P)im
I
I I I r=40x20nm !
UJ
400 500 600 700 800 Wavelength (nm)
400 500 600 700 800 Wavelength (nm)
400 500 600 700 800 Wavelength (nm)
400 500 600 700 800 Wavelength (nm)
800 -
Fig. 10. Local electric field maximum for two touching Ag particles: (a) circular, (b) ellipsoidal tubes, (c) right-triangular, and (d) equilateral right-triangular particles with different polarization.
117 Single molecule sensitivity in surface enhanced Raman scattering using surface plasmon plasmon 117
I
600
Size
k- j +
500 «"400 | f-300 •2. UJ
60'
f -(-
4-
0 r=40 nm, at 430 nm A r=80 nm, at 520 nm <> r=160 nm, at 580 nm
-S400 Q.
f
E
m
200 — u f
40 \ / \
100 —fflrfih -; MO
/i \Z/\
nf200
^
^
W I
600
(b)
400 500 600 700 800 Wavelength (nm) (c)
u
Isolated sphere
Tijiangular a^b=c=80 nm At 659 nm i
pn
:
Side-$lde (d=0.25i 0.5 n(t()
400 500 600 700 800 Wavelength (nm)
Fig. 11. Local electric field maximum for two touching Ag circular (a-b) and triangular (c-d) particles: (a) for various circle sizes, (b) spacing dependence for circular tubes with different radii, (c) spacing dependence for triangular and tetragonal, and (d) for tetragonal with different configurations. Spacing dependence was evaluated at the wavelength which gives the maximum electric field intensity at d (spacing) = 0 nm (see also Fig. 10).
Quite similar enhancement is obtained at 430 nm (G = 600) for touching ellipsoidal particles with the parallel polarization as well as the modest intensity (G = 60 at 400 nm and 520 nm) for vertical polarization (Fig. 10b). Rather complicate features at longer wavelength are obtained for triangular particles as the prominent enhancement at the edge is overlapped. The tremendous enhancement factors are obtained for triangular particles with parallel polarization, while slightly modest values similar to isolated particles are predicted for vertical polarization. Right-angle triangular particles
118 118
M. Futamata and Y. Maruyama
(80 nm x 40 nm) give vast enhancement G = 400 at 480 nm and G > 800 at >800 nm, while equilateral right-angle triangular give G > 600 at 560 nm for side-by-side and 700 nm for edge-to-edge with parallel polarization (see Fig. lOc-d). These features do not essentially depend on the particle sizes, e.g. see Fig. 1 la for spheres with a radii > 20 nm. As depicted in Figs, llb-c and 12, the enhancement rapidly decreases with increasing the gap size due to diminished LSP coupling and edge effect, i.e. G « (Go)/e at 1.5 nm gap size and G = 20 for 10 run, here Go the enhancement value for the contacting particles. The local field maximum decays with the similar way irrespective of particle sizes between r = 40 ~ 160 nm as shown in Fig. 1 lb. As anticipated, the local field maximum for the triangular edge does not depend on the gap size at the resonance wavelength, i.e. 500 at ca. 400 nm for the equilateral right angle particle. However, the enhancement arose from the LSP coupling at the junction with the resonance wavelength of 660 nm rapidly decreases even for triangles at the gap size < lnm similar to the circular particles (Fig. 1 lc). The sharp decay of the local field maximum is clearly explained by the peak shift of the LSP resonance to shorter wavelength and by the decreased coupling with increasing the spacing as shown in Fig. 10a In addition, these results indicate that the critical conditions to give the vast enhancement at SMS level are the LSP resonance and also the nanostructures such as edges or junctions, where the dense electric field is confined. This is rationalized by the following simulation that only modest electric field is obtained for two tetrahedral particles placed in a side-by-side configuration with sufficiently small spacing of d < 0.5 nm. In feet, the maximum enhancement (70 at 380 nm for a gap d = 0 nm) is quite similar to mat for isolated tetragonal particles (110 at 380 nm, see Figs. 1 Id and 8d). In addition, the scattering cross section for this configuration is similar to isolated one, as no additional bands are observed at longer wavelength. In contrast, the edge-to-edge configuration of the same tetragonal particles give die vast enhancement at ca 550 nm with a shoulder at 400 nm (Fig. 1 Id) as well as an additional peak in scattering cross section at ca. 540 nm (not shown). Conclusively, a sharp edge as well as the LSP coupling is indispensable to gave the enormously large enhancement Note that the junction between touching circular nanowires consists of moderate curvature at the metal and much sharper one at an air-side, which contributes electric field confinement enhanced by the LSP resonance. In addition, the local field maximum at a small protrusion with sufficiently sharp edge placed on a circular cylinder is quite similar to those for triangular edge or the junctions of particles, while the LSP spectra are similar to an isolated circular tube. This result suggests two plausible reasons for the discrepancies between the observed wavelength dependence of the optimum size and theoretical predictions [3]: (1) the second particle is located behind the first one, or (2) an isolated particle with small protrusions, which could not be imaged due to insufficient spatial resolution of the AFM. Thus, the spatial distribution of the LSP resonance and local electric field should be further studied with the SNOM method in addition to topographic image of the metal particles.
119 Single molecule sensitivity in surface enhanced Raman scattering using surface plasmon plasmon 119
20
(A) d=Onm
\
/ \
/
X-axis (nm)
20
Fig. 12, Electric field distribution for two adjacent Ag circular tubes (r = 40 nm) as a function of the spacing at 480 nm (wavelength): (a) gap d = 0 nm, (b) d = 1 nm, (c) d = 5 nm and (d) d = 2 0 nm.
Moreover, as shown in Fig. 7b, we confirmed by three-dimensional calculation feat fee similar enhancement of ca. 300 at 500 nm corresponding to SMS level in SERS is obtained at fee junction wife fee comparable wavelengfe dependence for two touching Ag spheres. Slightly smaller values compared to 2D calculation is due to rather large mesh sizes of 0.5 nmxO.5 nmxO.5 nm adopted to save computing time similar to fee SCS spectra. It indicates feat fee vast enhancement at fee junction is already obtained by fee two dimensional nanostructures, although distinct spectra for scattering cross section and the local field maximum are obtained for circular or ellipsoidal cylinder (2D) in comparison wife spheres or spheroidal particles (3D, e.g. see Fig. 7). 33. Origin of the Blinking 3.3.1 Blinking at room temperature As described above, Ag particles that yielded enormous enhancement are aggregates wife a typical size < ca. 1 |Xtn or touching particles, which give higher enhancement for polarization parallel to fee connection axis compared to fee vertical direction. Only a few of these particles showed the blinking. In addition, fee fluctuation
120 120
M. M. Futamata Futamata and and Y. Y. Maruyama Maruyama
of the peak frequency within ca. 10 cm"1 and narrower bandwidth (ca. 1/2) than feat for higher surface coverage were observed as shown in Figs. 13-14. It is noteworthy that these spectra were sequentially measured with an accumulation time of 1 sec to give sufficiently high signal to noise ratio wife the CCD, while actual blinking frequency is several Hz, slightly faster than the accumulation time. Nevertheless, intensity and peak frequency fluctuations were clearly observed, while prominent intermittent features were observed by Krug et al. [3] using avalanche photo diode. These observations suggest the existence of various adsorption sites on Ag particles with different interactions and enhancement Adsorbed molecules possibly diffuse between these so
*
counts
1700
1660 1620 Raman shift (cm1)
1700
1860 1620 Raman shift (cm'1)
sites. Fig. 13. Peak frequency fluctuation of the SERS spectra from R6G on Ag nanopartieles. Average surface coverage of R6G is 1 molecule per Ag particle. Each spectrum was obtained by sequential measurement of accumulation time of 1 sec at "k - 488 nm and 70 uW/pm2.
Before going into the temperature dependence, it is useful to discuss possible temperature increase of Ag nanopartieles by the excitation laser, which may cause significant difference in the LSP resonance and enhancement in SERS. At first, in our experiments with quite weak laser power of 70 (iW/jim2 at wavelength of 488 nm, the observed Raman bands are safely assigned to vibrational modes from original species. Accumulated spectra for a long duration of measurement, e.g. for 100 sec or longer, do not contain any pronounced Raman bands from plausible contaminants such as amorphous carbon [10]. Therefore, the blinking of SERS signal is attributed to R6G and adenine adsorbates. On the other hand, if the temperature of Ag particles increases even up to ca. 315 K, the morphology of the Ag particles could be irreversibly changed as reported by Semin et al. [43]. However, AFM measurement before and after the laser irradiation did not give any distinct change in nanoscale morphology of our Ag particles (not shown). The laser power (X = 488 nm, 100 (xW/[im2 for 10 min) adopted here is the maximal intensity in our experiments to measure SERS spectra, while only modest
121 Single molecule sensitivity in surface enhanced Raman scattering using surface plasmon 121
intensity of 1 jlW/nm2 was used for the SERS image measurements here. In contrast, Ag nanopartieles prepared with the same procedure significantly changed their morphology by annealing at 350 K or at 400 K for 1 hr in accordance with the observation for Ag island films [43]. Thus, invariant morphology of Ag particles under the laser irradiation indicates only negligible temperature increase of the samples during our SERS experiments. Namely the LSP resonance of Ag nanopartieles and thereby the enhancement are not modulated by the excitation laser, hi accordance with these results, the temperature increase of Ag particles is estimated to be negligible (< 4.5 K) for our laser power used for SERS spectral measurements (< 100 |lW/jim2) using the equation 3 in Ref. 11 based on stationary heat diffusion from Ag nanopartieles to hydrating water layer, hi addition, similar blinking features were observed for SERS spectra of adenine on the Ag nanopartieles at RT [7], which has no electronic transitions in visible wavelength region in a bulk solid or solution state. From these observations, it seems that photochemical bleaching, relaxation via triplet electronic stote or morphology changes by laser irradiation are not concerned with the blinking. The origin of the blinking is possibly due to thermal process such as thermal diffusion of individual molecule on the Ag particle presuming physisorbed molecules, of which binding energy is comparable with mermal energy. This is rationalized, if (1) the Ag surface has different sites with distinct enhancement, and (2) sufficiently large thermal energy to overcome an activation barrier for diffusion. (a) 2000
300
1
FWHM = 19 c m '
O
—X
o o
01
Intensity (cou
fhsoo -
n
1646cm-1
-
J
• • 1640 Raman shift (cm 1 ) I
1680
1600
1630
1640
Raman shift (cm"1)
1600
Fig. 14. SERS peak profile (a) from Ag particles with higher surface coverage, and (b) from blinking particles. Average sur&ce coverage of R6G in (a) and (b) are 300 molecules and 1 molecule per Ag particle, respectively. SERS spectra were obtained with the same condition as in Fig. 13.
Concerning the first point, it was demonstrated by numerical simulations using the FDTD method (see Section 3.2, [14, 40]) that vast enhancement of > 10 n in Raman scattering is obtained at a junction between two touching Ag particles with various shapes and sizes in addition to an edge of isolated trigonal prism under the LSP resonance (see Figs. 9 and 12). The vast enhancement sharply decays with increasing
122 122
Y. Maruyama M. Futamata and Y.
the gap size (Fig. 11, 12) [14]. Other sites apart from the junctions or edges of the touching particles and of isolated triangular prisms give only modest enhancement of < 30. These results by the numerical simulations agree with the experimental observations, since only Ag aggregates show the vast enhancement with parallel polarization to touching axis. Thus, we may attribute the blinking to thermal diffusion of adsorbed molecules on the Ag surface between the junctions with vast enhancement and other ordinary sites with modest enhancement If the blinking arises from thermal diffusion of adsorbates on the Ag particles with respect to the second point, the fluctuation frequency should be decreased or blinking is completely suppressed with decreasing the temperature according to simple consideration of hopping. Therefore, we measured the temperature dependence of the blinking in SERS signal from R6G at a surface coverage of ca 3 molecules/Ag particle. (a)298K
^
(b)77K
Fig. 15. Temperature dependence (I) of blinking: (a) at RT, (b) at 77 K. Darkened spot was observed at 77 K through the experiments (for ea. 10 min.), indicating blinking is suppressed at inactive sites.
3.3.2 Blinking at low temperature At first, the blinking particles were found at room temperature, and then cooled to 77 K, As clearly depicted in Figs. 15a-15b, a bright spot became completely darkened and never turned to a bright image at 77 K in contrast to repeated intensity changes at
123 Single molecule sensitivity in surface enhanced Raman scattering using surface plasmon 123
room temperature, e.g. bright spot with intensity fluctuation between 1.6 - 3.4 sec., and 3.9 - 4.4 sec. We also observed alternative cases that the blinking spot at room temperature, e.g. bright spot between 1.3 and 2.1 sec. did not change its intensity at 77 K through the measurement, >10 sec. in reality, as shown in Figs. 16a-16b. Both of these observations enable us to conclude that the blinking in SERS signal at RT is suppressed at 77 K, indicating the blinking is a thermally activated phenomenon: when individual dye molecule is immobilized at the sites with vast enhancement, bright invariant spot was observed at 77 K, whereas at modest enhancement sites, dark images were given. It should also be noted that roughly 1/3-1/4 of blinking Ag particles were frozen, suggesting most of the blinking particles have much smaller activation energy for the process compared to thermal energy at liq. N2 temperature. This is not surprising, since each adsorbed molecule can possess different bound energy on polycrystalline Ag particles according to locally different surface electronic state. Moreover, Raman spectra from R6G on Ag at blinking (room temperature) and at frozen (77 K) conditions, were safely assigned to intramolecular vibrations of R6G, e.g. 1653 (§ C-C sir.), 1582 ($ C-C str.), 1539, 1510 ($ C-C str.), 1358 cm"1 (0 C-C str.) in good agreement with the former report [45]. Rather poor signal to noise ratio of the SERS spectra compared to the previous one [40] is due to lower optical throughput and/or collection efficiency for the sample in a liq. N2 cryostat Occasional intensity difference of these SERS spectra is due to rather long accumulation time of 1 sec with respect to the blinking frequency of a few Hz. (a)298K
30 frames /sec
(b)77K
124 124
M. Futamata and Y. Maruyama
Fig. 16. Temperature dependence (H) of blinking: (a) at RT, (b) at 77 K. Bright spot was observed at 77 K through the experiments (for ca. 10 min.), indicating blinking was suppressed at active sites.
Interestingly, the frozen particles at 77 K recovered the blinking, when they were warmed to room temperature as shown in Fig. 17. Thus, the suppression of the blinking is intrinsic and reversible with the temperature variation between RT and 77 K. It clearly suggests that the temperature dependence observed here is not an experimental artifact, such as irreversible photochemical reaction of adsorbates by excitation light. Consequently, the blinking is thermally activated, most probably due to thermal diffusion of adsorbed molecules between the particular sites with vast enhancement and with modest enhancement on Ag surfaces. These sites are attributed to the junction (ca. 2-3 nm [14, 40]) of touching particles and other ordinary sites far from the junctions based on the theoretical simulation as shown in Figs. 11 and 12. Relative intensity changes of SERS bands during the blinking [6,12], can also be explained by orientation changes of molecules during diffusion with respect to the anisotropic electric field at the junctions [14,46].
(a)
(b)
(c)
Fig. 17. Temperature dependence (El) of blinking: (a) at RT, (b) at 77 K and (c) after warmed up to RT. Blinking was suppressed at 77 K and recovered at RT after wanned up, indicating it is thermally activated.
Weiss and Haran [11] reported for R6G on Ag particles using 532 nm excitation that (1) background intensity as well as SERS signal shows the intensity fluctuation, (2)
125 Single molecule sensitivity in surface enhanced Raman scattering using surface plasmon plasmon 125
fluctuation rate is proportional to laser power, while thermal effect is negligible in their experiments (excited at 532 nm with 1 |i,W/|m2) as the sample temperature does not increase at all. Then, they concluded the origin of the increased fluctuation is not a thermal but a photochemical process, possibly due to molecular diffusion that are mediated by desorption triggered by election tunneling between the metal surface and molecules. To avoid a possible confusion, it should be noted that the laser power dependence was studied by Weiss and Harran at a constant temperature, while we explicitly changed the sample temperature at a fixed laser power. As described in Section 3.1, in our experiments temperature of the Ag particles was not significantly raised by the excitation laser (at 488 nm with <100 |xW7(Jm2). Clearly, photochemical activation and thermal activation suggested here are not exclusive, but compatible for particular molecules such as R6G or other dyes which have an appropriate electronic resonance with the excitation light. Thermal diffusion can also work for other molecules which have no electronic transitions in visible wavelength at adsorbed states. With respect to vast electric field gradient at the junction, an optical confinement of molecules is prospected in analogous to the optically biased diffusion of R6G molecule in solution [47]. However, as predicted by theoretical simulation for surface plasmon enhanced optical forces [48], even if the enormously large electtic field is formed at the junction of touching Ag particles, only modest trapping force is induced for R6G adsorbates under our experimental condition (< 100 p,W/|im2at 488 nm, see also Fig. 1 in Ref. 48). In fact, they obtained the trapping energy of ca 0.03 feT/flOO fiWum'22) at the junction of two touching Ag particles with a size of 40 nm in diameter at X = 540 nm, roughly similar to the above estimation based on the trapping potential in solution [47]. Conclusively, the optical trapping force works to immobilize adsorbates onto the junction, but the value is much smaller than thermal energy. Thus, thermal diffusion is not frozen by the optical confinement effect At this stage, it is quite important to get definite evidences that adsorbed molecules are sitting at the junction when vast enhancement in SERS is observed. On this issue, it was convinced here based on the correlation between elastic scattering spectra and SERS activity together with the FDTD-3D simulation. 3.4. Critical importance of the junction for SMS-SERS 3.4,1 Elastic scattering experiments As reported in our recent paper [17], the elastic scattering spectra from the touching particles provide marked differences according to the on/off feature of SERS activity. For example, the Ag particles consisting of a few ellipsoids as in the AFM image (Fig. 18c) give the elastic scattering peaks at 530 nm and 730 nm according to the LSP excitation. In addition, another peak appears at 625 nm when tiny amount of dye molecules, e.g. 30 molecules/Ag particle, adsorb to give pronounced SERS signal as in Figs. 18a and 18b. Actually, the enhancement factor observed for the Raman bands of R6G at 1654,1574,1510 or 1365 cm"1 is estimated to be 108-109 compared to the signal intensity at bulk solution state. Very interestingly, the scattering peak at 625 nm disappears when the SERS activity is lost after duration of measurement for typically 30 min. This is probably due to diffusion to marginal sites or desorption of the
126 126
M. Futamata and Y. Maruyama
dye molecules as depicted in Figs. 18a, 18b and 18c. In addition, intermediate scattering spectrum was observed for partially active state. These distinct elastic scattering spectra are inherently related to the SERS activity, since identical features were observed for every hot particle. Moreover, these features were confirmed for the blinking particles as indicated in Fig. 19. Although two distinct LSP peaks were observed at ca. 475 nm and at ca. 630 nm according to rather complicate morphology of individual particles, appreciable spectral changes were observed at ca. 500-600 nm. Namely, definite scattering peak emerged at ca. 590 nm by the addition of R6G is extinguished when the SERS activity is lost. Experimental errors such as fluctuation of the signal intensity due to instable alignment of the optics are negligibly small, as confirmed for Ag particles at entirely inactive state (not shown). The additional peak in elastic scattering spectra observed at SERS-active state probably arises from the electronic absorption of the dye molecules, located at slightly shorter wavelength, i.e. ca. 560 nm as discussed in the next section. The LSP does not give such additional peaks unless quite formidable coalescence or morphology variations are induced by the dye adsoiptioa This is not the case, since only a few molecules are attached and desorbed from each Ag particle (B)
20a
525
627 730 "
1600 1400 1200 1000 800 Stokes shift (cm' 1 )
1 jim x 1
S00
400
/
500
500
• i BOO 700 Wavelength (nm)
600 700 Wavelength (nm)
800
800
immobilized on the glass substrate, of which morphology and aggregation state are invariant through the measurement as confirmed by AFM images. Fig. 18. Topography and scattering spectra fiom the same hot Ag particles (30 molecules/Ag particle): (A) SERS spectra, (B) elastic scattering spectra, ( Q Differential scattering spectra in B between SERS-active and —inactive particles, and (D) AFM images of the hot particles. Dotted line in (D) denotes the absorption of dye at bulk state.
127 Single molecule sensitivity in surface enhanced Raman scattering using surface plasmon plasmon 127
Concerning the observed spectral changes, it was reported independently by Craighead & Glass [39] and Garoffet al.[49] in the early stage of SERS study mat the extinction spectra of Ag evaporated films varies with deposition of thin dye film (thickness < lnm). They could qualitatively reproduce the observed spectra based on Garnet or Mie theory: (1) obvious splitting of the LSP peak, in other words new peak appeared at slightly longer wavelength than bulk absorption peak of dye molecules, and (2) feint changes in the LSP peak. Thus, together with analytical solution, the effective medium theory that describes averaged dielectric properties based on the volume fraction of the constituents is conveniently utilized to yield macroscopic optical response from composite or stratified samples. These efforts encourage us to evaluate electromagnetic field in near-field and in far-field for various nanostructures and locations of tiny amount of molecules with using numerical simulations.
800 1600 1400 1200 1000 800
1 JJJIl X 1
600
400
500
600 700 Wavelength (nm)
800
500
600 700 Wavelength (nm)
S00
Fig. 19. Topography and scattering spectra from the same blinking Ag particles with tiny amount of dye (3 molecules/Ag particle): (A) SERS spectra, (B) elastic scattering spectra, (C) Differential scattering spectra in B between SERS-active and -inactive particles, and (D) AFM images of the hot particles. Hie dotted line in D denotes the absorption of dye at bulk state.
3.4.2 Numerical simulations We adopted the three-dimensional FDTD method to evaluate the scattering cross section for various Ag nanoparticles at various wavelengths between 300 - 800 run. Small amount of the dye molecules, e.g. 2x2x2 nm3 or 4x4x4 nm3, are placed at
128 128
M. Futamata and Y. Maruyama
different sites on the isolated or touching Ag particles to investigate the effect of adsorbed positions. At first, an isolated sphere with a radius (r) of 40 nm was evaluated as presented in Fig. 20a. An extremely weak but detectable peak appears at ca. 550-580 nm for thick dye films (> 4 nm) along with the invariant LSP peak of Ag particles at ca 400 nm. Stronger coupling was obtained for a larger Ag sphere such as r = 80 nm, which gives a new peak at 650 nm together with a broad LSP resonance at 500 nm, as the integrated intensity is about 10 times larger than that for the smaller sphere (r = 40 nm) as shown in Fig. 20b. It suggests importance of the peak separation of the LSP resonance and absorption of dye for the electromagnetic coupling. (a) Sphere (r40)
15x10 -
6x10' dye 8 nm dye 4 nm dye 2 nm dye 1 nm dye 0.5 nm dye 2x2x2 n m Bare
500 600 700 Wavelength (nm)
800
- ! 7— dye 4 nm -ft— dye 2nm -D- dye 1 nm Bare
400
500 600 700 Wavelength (nm)
800
Fig. 20. Elastic scattering spectra calculated for the Ag isolated spheres with various thickness of dye: (a) with a radius (r) of 40 nm and (b) r = 80 nm sphere. Distinct forward (open) and backward (filled) scattering were obtained only for the larger sphere (b).
Similar to isolated spheres, pronounced coupling is obtained for the isolated Ag spheroids with various aspect ratio of 160:80:80 ((a), in nm for the diameters dx: dj,: d j , 80: 80:40 (b) and 80: 40: 40 (c) as presented in Fig. 21. The LSP extinction peaks are observed at 380 nm and 550 nm for a bare particle in (a) with tilted polarization from the x-axis (by 45 °), where the former and latter peaks originate from the LSP resonance along the y- and x-axis, respectively (Fig. 21a). An additional peak gradually grows at ca, 470 nm, while the LSP peak at 550 nm shows red-shift with increasing dye thickness. Obviously, much larger effect is observed for the prolate with the size of 160 nmx80 nmx80 nm compared to smaller prolate with the same aspect ratio (80 nmx 80 nmx40 nm) or the oblate (80 nmx40 nmx40 nm) (see Fig. 21b, c). Again it is due to proximity of dye absorption (590 nm) and LSP resonance for the larger prolate (160x80x80, at 550 nm), because the oblate or smaller prolate with the same aspect ratio gives the LSP peak at 410 nm or 450 nm far away from the dye absorption (we will be back to this point). In these cases, rather thick dye adsorption >1 nm, which covers Ag surfaces fully and homogeneously corresponding to > 105 molecules/particle,
129 Single molecule sensitivity in surface enhanced Raman scattering using surface plasmon plasmon 129
is necessary to detect the additional peak at 470 nm. Distinct spectral fluctuations are not obtained for much smaller amount of adsorbates such as 2 nmx2 nmx2 nm (see Fig. 21a) in contrast to adjacent Ag particles (vide infra). In addition, these pronounced spectral changes for isolated particles with different shapes and sizes were never observed as well as variations in the LSP resonance wavelengths in 2D simulations. Thus, only modest spectral variations were obtained in elastic scattering with dye ,5 (b)
1.5x10'
800
400
500 600 700 Wavelength (nm)
800
adsorption for isolated Ag particles irrespective of their sizes and shapes. Fig. 21. Elastic scattering spectra calculated far the Ag isolated spheroids with various thickness of dye films: (a) with a size of 160 nm (x) x 80 nm (y) x 80 mn (z), (b) 80 nm x 80 nm x 40 nm, (c) 80 nm x 40 nm X 40 nm. Incident beam was illuminated along z-direction with the incline (45 °) polarization.
Next, two adjacent Ag nanospheres with a diameter of 80 nm were adopted to evaluate the effect of dye adsorption at various positions including junction or marginal sites. Actual SERS active Ag particles consist of 3-4 particles with different shape and sizes as shown in Figs. 18-19. However, the model structure is relevant to discuss the effect of dye adsorption onto the junction, since we are studying single molecule phenomena plausibly arose from an individual junction. While the FDTD-2D method provides essential features on the effect of dye adsorption onto elastic scattering spectra, accurate wavelength dependence of the scattering spectra is obtained only with the 3D calculation. Two adjacent bare Ag spheres (r - 40 nm) with various gap sizes (g, nm) provide two distinct peaks attributed to isolated LSP resonance at 370 nm and coupled LSP at 430-550 nm as presented in Fig. 22a. With increasing the gap sizes, the coupled LSP peak shifts to shorter wavelength according to diminished coupling, i.e. 550 nm,
130 130
M. Futamata and Y. Maruyama
490 ran, 450 nm and 430 nm for g = 2 nm, 4 nm, 10 ran and 20 nm, respectively. In other words, the LSP peak results in pronounced red-shift with increasing the coupling between neighboring particles in contrast to minute variations in 2D evaluations, e.g. peak shift of 420 nm for g = 2 nm or 380mn for g = 8 nm in addition to long tail extended to infrared region. (a) 15x10 4
_
)
W
„
i
A 55CJ
i430yR9 i\
\ ; JW |1 \ti \V \L\
"£10 c CO
o co
1 1 —•— 0 nm \ -V— 2nm | -O-4nm j - D - 1 0 nm I -&— 20 nmj~
,4 (b) 15x10 [ 800 X-pol - • - dye (gap) —o— dye (gap, Az = +4nm) •-V- dye (gap, Az = +8nm).
- n - Bare Y-pol - dye (gap) -, - © - Bare
ip •/ aa. H \
5
01
400
500 600 700 Wavelength (nm)
500 600 700 Wavelength (nm)
400
800 d)
15x10
|
-
c CO O 5 co
or 500 600 700 Wavelength (nm)
800
! I I I ig=10nm: - D - R6G, - • - Bare ;g=20nm -A— Bare
j-
Tjrfl'H
10x10*
400
n M
800
l-
1 400
n 500 600 700 Wavelength (nm)
ii 800
Fig. 22. Elastic scattering spectra calculated for the Ag neighbouring spheres (diameter of 80 nm) with different gap sizes ( g = 2 nm, 4 nm, 10 nm and 20 nm) using the FDTD-3D method: (a) bare particles, (b) with dye filling the gap of 4 x 4 x 4 nm3, (c) dye molecules with a volume of g x g x g nm3 are attached to fill the gap of the same Ag particles, (d) dye molecules with a volume of 4 x 4 x 4 nm3 placed at the center of gap (g = 10 nm or 20 nm).
Significant spectral changes are observed by filling the gap with dye as shown in Figs. 22b, c. For instance, an additional peak is observed at 750 nm for the gap (4 nm) when illuminated with the polarized light parallel to the touching axis, here the dye molecules with the volume of 4x4x4 nm is used (see Fig. 22b, although smaller volume of dye such as 2x2x2 nm3 provided significant spectral changes, larger volume of dye was employed to demonstrate the difference clearly). The LSP peak at 370 nm and 500 nm shows only little spectral changes. It should also be noted that 1) the additional peak is not observed for other location of dye molecules, e.g. the left end or top position of the left side particle (see Fig. 22c) as in the FDTD-2D calculation. 2) Incident light with y-polarization (vertical to the touching axis) does not give such
131 Single molecule sensitivity in surface enhanced Raman scattering using surface plasmon plasmon 131
spectral changes (Fig. 22b, overlapped completely by the bare spectrum) [5,14,15]. 3) The additional peak position and intensity are quite sensitive to the detailed location of dye at the gap, e.g. slight migration (by 8 nm) along the y-axis within the gap causes a weak shoulder at ca. 600 nm, while minor diffusion (4 nm) does not modify the 750 nm peak (Fig. 22b). These results are rationalized, since the marginal positions even in the gap or y-polarized light do not increase the LSP coupling of neighboring Ag particles. At larger gap sizes, which are filled by the dye molecules, the LSP peak manifests a blue shift with diminished intensity, e.g. 530 nm (g = 2 nm), 470 nm (g = 4 nm), 430 nm (g = 10 nm), or 370 nm (g = 20 nm) as in Fig. 22c. These peak deviations are significantly larger compared to the bare particles due to the increased LSP coupling through conductive dye molecules. Monotonous peak shift of the additional peak is not observed according to distinct coupling efficiency, as the LSP peak positions for the bare Ag particles are already dissimilar at different gap sizes. In addition, only faint spectral changes without the additional peak are obtained for small amount of dye (4x4x4 nm3) located at the center of larger gap (g = 10 nm or 20 nm) as shown in Fig. 22d, indicating negligible coupling of the LSP at each Ag particle. The dye filling Ihe gap with prominently small sizes increases the LSP coupling to give pronounced red-shift of the LSP peak together with the additional peak (see Fig. 22b). If quite large amount of molecules such as 20x20x20 nm3 or more adsorb and continuously detach from the junction of Ag particles presumably due to photochemical decomposition or evaporation under laser irradiation, one can expect drastic red-shift of the LSP peak and simultaneous blue-shift of the additional peak with diminished intensity. Much smaller amount of dye like 4x4x4 nm3 (= 64 nm3 < 1/100X(20X20X20 nm3)), in contrast, would provide only slight spectral changes that are elimination of the additional peak at 750 nm and concomitant LSP peak shift from 500 nm to 470 nm. These are in good agreement with our empirical observations for extremely low surface coverage. It is noteworthy that the absorption of dye is crucial to yield drastic spectral variations, since transparent materials with fairly high refractive index do not give similar changes. For instance, water molecules with the dielectric, constant of e - 1.82, or SiO2 of e = 2.10 with a size of 4x4x4 nm3 filling the gap (g = 4 nm) do not give additional peaks at longer wavelength (see Fig. 23c) [50]. Furthermore, the additional peak position shifts monotonously from 580 nm to 750 nm by tuning the absorption wavelength (ko) from 400 nm to 700 nm, in contrast to the invariant LSP peaks at 370 nm and 470 nm as depicted in Figs. 23a and 23b. This is clearly explained by variation of the implicit coupling between the LSP and dye. Much efficient coupling is obtained for the dye absorption (Xo) between 400 and 500 nm with the LSP resonance at 470 nm, whereas gradual decrease with increasing their separation, at Xo > 500 nm, as evidenced by weakened peak intensity. Accordingly, only dye located at the junction with the absorption band close to the LSP peak gives the additional peak with parallel polarization in elastic scattering spectra. Increased coupling is afforded by favorable electric conductivity (0), e.g. ca. 3.3xlO4 at X,= 450 nm or 1.8xlO6 at X = 589 nm (at absorption peak) as estimated from the equation, 0 = to EQ E", where co, £0 and e" are frequency of the excitation light, free space dielectric constant and imaginary part of dielectric constant of dye [17]. Identically, we confirmed enhanced absorption of dye at
132 132
Y. Maruyama M. Futamata and Y.
the same wavelength (750 nm) as in Fig. 23d. Apparently one may address that rather large amount of dye molecules, 4x4x4 ran3, or 2x2x2 nm3 are necessary to provide significant spectral changes in comparison with 2D calculation (lxl nm) [17]. However, this is unambiguously not correct since the Ag nanowire (with infinite length) adopted in 2D is enormous compared to the nanoparticles (with definite height). In addition, the amount of adsorbates to give distinct spectral changes in the 3D simulation, being dependent of absorption coefficient of dye, is still sufficiently small, e.g. 2x2x2 nm3 of R6G molecules roughly corresponds to a volume for 10 molecules (presumably 1 nm3 for each R6G molecule assuming parallel orientation of its molecular plane to the Ag surface) is comparable with the averaged surface coverage of 3 molecule/Ag particle (= 6 molecules/junction) in our experiments. These are explicitly qualitative analysis sufficient for the evaluation of the effect of dye adsorbed on adjacent Ag particles. It should be noted that quantitative analysis for the elastic scattering of metal nanoparticles is still under progress, e.g. LSP resonance peak position could be reproduced involving the effect of substrate or image dipole [51]. As the next step, the quantitative analysis of the scattering cross section will be studied based on the actual size and morphology of the hot particles. 15x10
(a) - O - 590 nm - D - 400 n -©— 450 nm I -&- 500 nm —®— 650 nm —v— 700 nm
400
500 600 700 Wavelength (nm)
800
400
500 600 700 Wavelength (nm)
800
15x10
400
500 600 700 Wavelength (nm)
800
Fig. 23. Elastic scattering spectra for adjacent Ag nanospheres (r = 40 nm) with a gap size of 4 nm calculated with FDTD-3D: (a) dye molecules filling the gap wifli different absorption peak wavelength (Xo) between 400 nm and 700 nm, (b) plot of the peak position as a function of absorption peak wavelengths, (c) effect of various species filling the gap, (d) absorption spectra obtained for R6G filling the gap. Parallel polarization was used to the touching axis in these calculations.
133 Single molecule sensitivity in surface enhanced Raman scattering using surface plasmon plasmon 133
To summarize the section, it should be noted again that the SERS activity correlates with the spectral changes in elastic scattering for the same Ag particles: 1) vast enhancement in SERS was obtained only for Ag touching particles with the polarization parallel to the touching axis. The blinking arises from thermal diffusion of adsorbed molecules between adsorption sites with distinct enhancement. 2) The additional peak at 590-620 nm was observed in elastic scattering spectra only when the same sample gives the vast SERS activity. In contrast, exactly the same spectra as bare Ag particles were obtained for inactive SERS particles. These spectral changes are attributed to the adsorbed dye at the junction as confirmed by the FDTD-3D simulatioa Consequently, the adsorption of dye molecules at the junction of touching Ag particles gives vast enhancement in SERS and additional scattering peak. This is primarily based on the so-called electromagnetic mechanism, while implicit electronic interaction between the Ag nanoparticles and adsorbates is exploited for the efficient LSP coupling. Our target is to explore the optimum nanostructures with SMS in SERS in the frame work of SPP resonance. The electromagnetic enhancement is much useful and convenient for various applications compared to the chemical enhancement, e.g. charge transfer resonance at specific sites on atomiealry-roughened surfaces is relevant only for particular adsorbates and metal surfaces. Further efforts are prerequisite on the correlation between SERS activity and elastic scattering spectra to clarify; (1) if ordinary molecules with no electronic transitions in visible wavelength such as DNA bases give the additional scattering peak upon adsorption or not. Alternatively, when adsorbed molecules possess specific electronic interaction with Ag surfaces to form charge transfer state, the additional peak will be observed. If not observed for blinking particles with ordinary adsorbates, variety of molecules could be characterized at a single molecule level merely with the electromagnetic mechanism. It is also necessary to study (2) if additional enhancement is obtained with using optical current (dynamic charge transfer) as pointed out by Otto [52,53]. He suggested that the blinking could be induced by thermal fluctuation in orientation of adsorbates such as ethylene at the junction which determines the excitation eflBciency of vibrational modes with tunneling current. In analogy, fluorescence of ZnTBP (Zn-tetra-buthyl-phenyl-porphyrin) molecules on Cu(100) surface was observed with using the tunneling current in STM [541. The authors in Ref. 54 presumably attributed the multiple peaks separated by 800 cm" to the excited vibrational modes of TBP with the tunneling current. On this issue, quantitative evaluation of the local electric field or enhancement factor of SERS signal is necessary based on the actual metal nanostructures with using a reliable numerical method involving Stokes shift for the LSP resonance and phase shift in Raman scattering together with precise experimental study on the SERS intensity for various molecules at SMS level [55]. It is also useful to discuss the chemical enhancement with respect to SMS-SERS. To our knowledge, there are two distinct approaches to explore the optimum metal nanostructures; (1) to find an appropriate particle with a fixed excitation wavelength [3-7] or (2) to find an optimum wavelength for particular metal particles [8]. The former group reported that only touching particles exhibit the vast enhancement based on the AFM measurement and optical spectroscopy, while the latter accounted even for
134 134
M. Futamata and Y. Maruyama
isolated particles with particular sizes display the blinking or vast enhancement at different excitation wavelengths. For example, Doering and Nie ensured isolate! gold ellipsoidal particles provide the bunking with particular sizes that are resonant with the excitation wavelength [9]. In this case, isolate! Ag particles composed of spheres, ellipsoid or cylinder should have only modest enhancement like 10 by means of the LSP resonance as reported earlier [5, 14, 15]. Prominent enhancement with LSP resonance in addition to chemical enhancement is expected at sharp edge structures as in the triangular prism [14, 15]. Accordingly, they could increase the number of blinking particles by the addition of halide ions to the Ag particles immobilized onto the substrate. It indicates plenty of active sites are induced for the chemical enhancement and/or extremely localized surfece plasmon at the sharp edges by slight dissolution of Ag. Thus, for isolated metal particles, the chemical enhancement and/or nanoscale sharp edges are possibly utilized to yield SMS in SERS, although further conclusions await detail nanoscale characterization on the surface morphology and electronic state. This issue is still unintelligible on account of insufficient spatial resolution in near-field spectroscopy [56].
3.5. Fabrication of metal nanostructures With gradual evaporation of water from polystyrene suspended solution that is intruded between the glass plates, monolayer of the nanospheres was formed. Topography of the closest packed monolayer of PS nanospheres evidences efficient growth of quite large domains with a size of about 100 |imxl cm, although the fabrication method is simple and convenient The monolayer domain was feasibly identified using diffraction properties of such ordered particles with a nanometer size, as also confirmed by scattering peak centered at 450 - 500 nm in wavelength. After silver deposition followed by sonication to remove PS spheres, the trigonal silver nanostructure appeared as in Fig. 24. Each trigonal structure is periodically located at a hexagonal position on a glass substrate with trigonal sharp edges as shown in Fig. 24b. Unfortunately the image encompasses the blank area where the tiigonal nanospots were washed out by the sonication. This can be improved by optimizing the sonication conditions, while adjusting the adhesion of silver on the glass substrate by predepostion of Ti. Also, continuous silver peninsular structures were observed along the domain boundary of the ordered PS layers. Thus, rather large occupied area of 12 % (in average) was obtained for the trigonal structure compared to 9.4 % for completely ordered structure (theoretical area occupied by the trigonal silver nanostructure relative to the entire substrate surfece). Because the peninsular structure contains sharp edges (of course with smaller number), it can contribute to afford the vast enhancement in SERS. Accordingly, the prepared silver nanostructure is not perfect but still promising for the SERS blinking study. The trigonal nanostructure has the following geometrical sizes in average: height (thickness) of 80 nm, length of the trigonal base plane of ~ 190 nm, nanostructure spacing of 355 nm (310 nm for the closest packing model). The top of the trigonal nanostructure is quite flat with roughness < 5 nm without containing granular junctions. SERS spectrum of R6G adsorbed on the Ag trigonal prism was measured for an accumulation time of 10 s to afford enough high signal to noise ratio,
135 Single molecule sensitivity in surface enhanced Raman scattering using surface plasmon plasmon 135
whereas it is rather long compared to the blinkingfrequenciesof several hertz. However, the accumulation time of 10 s is sufficiently short to distinguish the blinking region from dark (inactive) region. The blinking spot gave the SERS spectra from R6G (not shown), e.g. 1651 cm"1,1575 cm'1,1514 cm"1,1361 cm"1,1195 cm"1 and 770 cm"1 well correspond to those from bulk R6G spectra at 1647 cm"1,1568 cm"1,1534 cm"1,1504 cm"1,1365 cm"1,1195 cm"1 and 771 cm"1, respectively.
Fig. 24. Topography of fee silver nanostrueture prepared by NSL: (a) 2-D array of trigonal nanoprism and (b) expanded image of fee trigonal nanoprism at hexagonal positions. Size and spacing of fee nanoprism can be controlled by fee diameter of PS and inclination of fee substrate to fee evaporation source.
Since the R6G solution of 0.5 pi with 10"10 M was spread onto the silver nanostrueture with a typical diameter of about 1 cm, the number of molecule on the sampled area of ~ 10x10 pm2 is estimated to be ca. 40. Under these conditions, the SERS spectra were observed only when and where the SERS blinking occurred. At first, the blinking probabilities on the trigonal silver nanostructures and on continuous silver films, which were prepared with the same evaporation condition to 50 nm in thickness, were compared. In contrast to the trigonal structures, Ag continuous films consist of granules with junctions as shown in AFM images (not shown), which can contribute to give vast enhancement and the blinking similar to nanoparticles prepared by chemical
136 136
M. Futamata and Y. Maruyama
reduction. However, it is reasonable to compare the efficiency of these structures to verify the validity of nanostrueture fabricated for the SMD with SERS. Then, it was compared with the results from annealed silver continuous films with much flat and smooth surfaces. Probability of the SERS blinking was evaluated in each sampled area of 9.66x9.66 urn2 for 10 s as shown in Fig. 25. The area was divided into 7x7 divisions (each division has a size of ~ 1.3 ^imxl.3 (jm at the sample position, which is slightly larger than spatial resolution of the microscope). As clearly seen in Fig. 25, the trigonal nanostrueture gives quite high yield that > 90 % of unit area showed the blinking. Although the bunking was observed even for the continuous silver film without annealing, almost 35 % of the unit area is free from the blinking. Actually, the probability on the uigonal nanostrueture is much higher than that on the continuous silver films. Especially the probability of more than 5 spots (1 spot =1.3 pmxl.3 jim) per area (9.66 |imx9.66 |im, composed of 7x7 spots) on the trigonal surface is about 2.5 times higher compared to the continuous silver surface. Namely, the percentage of the spot with the probability more than 5 spots/area occupied 60.5 % for the trigonal nanostrueture, while 25.9 % for the continuous silver film. Furthermore the total number of the SERS blinking spot is 653 in case of the trigonal nanostrueture compare to 325 on the continuous silver film without annealing. In contrast to the continuous silver film, the trigonal nanostrueture occupies only a small area of the substrate, 9.3 % as estimated from the geometrical data for the closest packing of PS. Normalizing the occupied area for the present trigonal silver nanostrueture, the probability of the SERS bunking observation is about 25 times higher than that for the continuous silver film. Even if we adopt the experimentally observed occupancies of about 12 % for the trigonal nanostrueture, the probability of the blinking is approximately 17 times higher man the continuous silver films. These results suggest that the bunking at the trigonal structure does not result from the same origin in the continuous films but from more efficient mechanism as predicted by theoretical evaluation for the local electric field on sharp edges. Moreover, the blinking in SERS was not observed on the continuous silver films after annealed at 473 K for 2 h. As evidenced by AFM measurement (not shown), the surface morphology of the silver continuous films is prominently modified by the annealing: (1) as-evaporated film consists of rather fine particles with a diameter of 100 nm or less, but (2) after annealing, the silver granular structure grows to form much larger and smooth particles with a typical diameter of 250 nm or more. Namely, the sharp junctions on as-grown samples were extinguished, whereas coalescence and partial melting yield larger granules with blunt contact. As described above, the junctions of touching metal nanoparticles as well as the sharp edges of trigonal silver nanostrueture provide vast enhancement, viz. > 1010, of the local filed under the LSP resonance, whereas curved surfaces give only modest enhancement, viz. 104-105, similar to sphere or ellipsoidal particles. According to disappearance of such junctions, the vast enhancement in SERS and the blinking were suppressed for the annealed silver films. We noticed that detail morphology of the trigonal structures, e.g. packing density or sharpness at the edge is determined by adhesion and wettability of PS and Ag on the glass substrate as well as evaporation conditions. These properties can be controlled by
137 Single molecule sensitivity in surface enhanced Raman scattering using surface plasmon plasmon 137
pretreatment of the glass substrate using various chemicals with different hydrophilicity. For instance, heating of the glass substrate in peroxosulfete solutions removes organic contaminants to yield hydrophilic surfaces, while coating by silane coupling chemicals with methyl-end groups gives relatively hydrophobic surfaces accommodating the PS layers. In our case, relatively hydrophobic glass substrates (beneath the PS) rinsed simply by pure water were used to immobilize the PS monolayer with the hydrophilic covering glass pretreated in alkaline solution. Faint but significant differences in hydrophilicity of the overlaid and underlaid glass substrates benefit efficient and smooth drying for water layer with the lateral capillary forces. In addition, the vacuum chamber provides parallel Ag beam during evaporation to give homogeneous deposition at centered and marginal region of the restricted trigonal spaces between the densely packed PS particles. Accordingly, terrace of the trigonal nanostaictures prepared here are sufficiently smooth with roughness < ~ 5 nm even without annealing, which does not contain explicit granular structures. In fact, roughness of the trigonal structures is negligibly small compared to the annealed continuous Ag films. Thus it indicates that the blinking of the trigonal structure does not arise from die junctions of granular structure as in continuous films but from the sharp edges. The trigonal Ag nanostrueture is quite promising to detect a single molecule by means of SERS, while detailed study on correlation between the edge shape and the enhancement factor in SERS is now being progressed using various pretreatment and evaporation conditions. Advanced technologies such as electron beam lithography or focus ion beam could yield metal nanostructures for SMS-SERS that are applied to various fields such as rapid DNA sequencing, single molecule analysis in a living cell, or characterisation of individual species in molecular devices.
(a)
S
10 15 Spot numbers
5
10 15 Spot numbers
20
Fig. 25. Probability (frequency) of the SERS blinking in the sampled area (9.66 x 9.66 um2): (a) for the silver surface with trigonal nanostructures, and (b) for the continuous silver film without annealing.
138 138
M. Futamata and Y. Maruyama
4 CONCLUSION In good accordance between experiments and theoretical evaluation, only touching metal nanoparticles provide enormous electric field to yield SMS, whereas isolated particles, except nanostructures with sharp edges, gave only modest enhancement This is attributed to the increased coupling of the LSP of individual particles at the junction. Blinking is raised from thermal activation, most probably thermal diffusion of adsorbed molecules between the junction and marginal region. An elastic scattering peak at ca. 600 nm was extinguished during inactivation process of enormous enhancement in SERS of Ag touching particles with adsorbates. Numerical simulations using the FDTD-3D method proved mat this peak originates from implicit electromagnetic coupling between the LSP of the Ag particles and absorption of dye located at their junction. Consequently, critical importance of the junctions of Ag particles was evidenced with respect to single molecule sensitivity in SERS. Trigonal silver nanoprism was fabricated with the nanosphere lithography that can provide enormous electric field at their sharp edges under the LSP resonance. In fact, it yielded much higher probability of the blinking compared to the continuous film. Accordingly, the trigonal silver nanoprism or other metal nanostructures could be soon prepared with using advanced technologies to achieve SMS in SERS.
ACKNOWLEDGMENT The authors appreciate Dr. Mitsuru Ishikawa (AIST) and Dr. Yoshinori Yamaguchi for useful collaboration. This research was financially supported in part by Grant-in-Aid for Scientific Research (B) 14340189 by Japan Society for the Promotion of Science (JSPS), by New Energy and Advanced Industrial Technology Development Organization (NEDO), and also by Core Research for Evolutional Science and Technology (CREST) project of Japan Science and Technology Corporation (JST).
REFERENCES [I] A. Otto, I. Mrozek, H. Grabhom, and W. Akemann, J. Phys.: Condens. Matter, 4 (1992) 1143. [2] ' M. Keiker, Surfece Enhanced Raman Scattering, Proc. SPIE, MS10 (1990). [3] J. T. Krug, G. D. Wang, S. R. Emory, and S. Me, J. Am. Chem. Soc., 121 (1999) 9208. [4] K. Kneipp, H. Kneipp, L Itzkan, R, R. Dasari, and M. S. Feld, Chem. Rev., 99 (1999) 2957. [5] H.Xu,J.Aizputua,M.Jail,andP.A P ell,Phys.Rev.E,62(2000)4318. [6] M. Michaels, M. MimaL and L. E. Bras, J. Phys. Chem. B, 104 (2000) 11965. [7] Y. Maruyama, M. Ishikawa, and M. Futamata, Chem. Lett, (2001) 834. [8] S. R. Emory and S. M. Me, J. Phys. Chem. B, 102 (1998) 493. [9] W. EDoering and S. Me, Anal. Chem., 75 (2003) 6171. [10] A. KudelskiandB. Pettinger, Chem. Phys. Lett., 321 (2000) 356. II1] A. Weiss and G. Haran, J. Phys. Chem. B105 (2001) 12348. [12] K. A. Bosnick, J. Jiang, and L. E. Bras, J. Phys. Chem., B106 (2002) 8096.
139 Single molecule sensitivity in surface enhanced Raman scattering using surface plasmon plasmon 139 [13] Y. Maruyama, M. Ishikawa, and M. Futamata, I Phys. Chem., B108 (2004) 673. [14] M. Futamata, Y. Maruyama, and M. Ishikawa, I Phys. Chem., B107 (2003) 7607. [15] I P. Kottmann, O. J. F Martin, D. R. Smith, and S. Schultz, Chem. Phys. Lett, 341 (2001) 1. [16] K Hao, and G. C. Schatz, J. Chem. Phys., 120 (2004) 357. [17] M. Futemata, Y. Maruyama, and M. Ishikawa, J. Phys. Chem., B108 (2004) 13119. [18] Similar spectral changes for much larger amount of molecules on several tens of Ag nanoparticles were reported in aqueous media by T. Itoh, K. Hashimoto, A. Ikehata, and Y. Qzaki, Appl. Phys. Lett, 83 (2003) 5557. [19] K. Nagayama, Colloids and Surfaces A, 109 (1996) 363. [20] N. D. Denkov, O. D. Velev, P. A. Kralchevsky, L Blvanov, H. Yoshimura, and K. Nagayama, Langmuir,8(1992)3183. [21] S. Dimitov, C. D. Dushtrin, H. Yoshimura, and K. Nagayama, Langmuir, 10 (1994) 432. [22] S. Dimitrov and K. Nagayama, Langmuir, 12 (1996) 1303. [23] P. A. Kralchevsky and K. Nagayama, Adv. Colloid M. ScL, 85 (2000) 145. [24] L. A. Dick, A. D. McFarland, L. C. Haynes and R. P. Van Duyne, J. Phys. Chem. B, 106 (2002) 853. [25] L A. Dick, A. J. Hayes, and R. P. Van Duyne, J. Phys. Chem. B, 104 (2000) 11752. [26] T. R. Jensen, M. L. Duval, K. L. Kelly, A. A. Lazarides, G. C. Schatz,, and R. P. Van Duyne, J. Phys. Chem. B, 103 (1999) 9846. [27] L. Haynes and R, P. Van Duyne, Nano Lett, 3 (2003) 939. [28] L. Haynes and R. P. Van Duyne, J. Phys. Chem. B, 105 (2001) 5599. [29] L. Haynes and R. P. Van Duyne, J. Phys. Chem. B, 107 (2003) 7426. [30] P. C. Lee andD. P. Meisel, J. Phys. Chem., 86 (1982) 3391. [31] T. Yamasaki andT. Tsutsui, Jpn. J. Appl. Phys., 38,5916 (1999). [32] P. K. Aravtod and H. Metiu, Surf. ScL, 124 (1983) 506. [33] N. Liver, A. Nitzan, and J.I. Gersten, Chem. Phys. Lett, 111 (1984) 449. [34] H. Chew and M. Kerker, J. Opt. Soc. Am. B, 2 (1985) 1025. [35] K. S. Yee, IEEE Trans. Antennas Propag., 14 (1966) 302. [36] A, Taflove (Ed), Computational Eledrodynatnkxthe finite-difference time-domain method (2nd Ed), Artech House, Norwood 2000. [37] D. Palik, Handbook of Optical Constants of Solids; Academic press, London, 1998, P.351. [38] K. S. Kunz and R, J. Luebbers, The Finite Difference Time Domain Method for Electromagnetics, CRC Press, Boca Raton, 1992. [39] H. G. Creighead and A. M. Glass, Opt. Lett, 6 (1981) 248. [40] M. Futamata, Y. Maruyama, and M. Ishikawa, Vihrational Spectrosc. 30, (2002) 11965. [41] As the first order approximation, the vast amplitude enhancement of > 330 for the incident field obtained at the junction yields the Raman scattering enhancement of >1010. As the scattering intensity from induced Raman dipole is resonantly enhanced by the LSP excitation of metal particles as well as the incident channel, the Raman scattering intensity is approximately proportional to fourth power ofincident electric field iRa™1* lE^xEi | z = |Ej | 4 x | a | 2 ( h e r e , E s , a, Ej are scattering field Raman tensor and incident field intensity, respectively). The value 1010 corresponds to the SMS as it yields detectable signal of 5-10 counts/sec with our fecility [40]. [42] F. Bohren and D. R. Hoffinan, Adsorption and Scattering of Light by Small Particles, John Wiley & Sons, 1983 New York. p. 344. [43] D. J. Semin,; A. Lo, S. E. Roark, R. T. Skodje, and K. L. Rowlen, J. Chem. Phys, 105 (1996) 5542. [44] J. P. Kottmann, O. J. F. Martin, D. R.Smith, and S. Schultz, Phys. Rev. B, 64 (2001) 235402. [45] P. Hildebrandt and M. Stockburger, J. Phys. Chem., 88 (198) 5935. [46] F. J. Garcia-Vidal and J. B. Pendry, Phys. Rev. Lett., 77 (1996) 1163.
140 140
Y. Maruyama M. Futamata and Y.
[47] M. A. Osbome, S. Balasubramanian, W. S. Furey and D. Klenerman, J. Phys. Chem. 102 (1998) 318). Loral electric field induced at the junction is estimated to be about 30 p,Wx (3 nm/1 um)z x 10s = 27 jiW (= laser power at the sample in $1 um x relative are of the jurctionxenhancement factor). It gives the trapping potential of ca. 0.0008 - 0.003 eV by assuming the same parameters as in solution at a first approximation, which is comparable to thermal energy of 0.026 eV at room temperature. [48] H. Xu and M. Kail, Phys. Rev. Lett., 89 (2002) 246802. [49] S. Garoff; D. A. Weitz, T. J. Grarrrila, and C. D. Hanson, Opt Lett, 6 (1981) 245. [50] A weak peak appeared at ca 420 run for SiCb and H2O at the opposite side of the above additional peak for dye. Possibly, this is due to quite high refractive index of the materials at rather small gap sizes, as phase velocity of light is increased by a factor of ca. 1.5 for S1Q2 (reftactive index of n = 1.45). Thus the gap size of 4nm used apparently decreased to about 2 nm for SiCh adsorbates. This is supported by the observation that similar peak was obtained at 420 nm for smaller gap sizes (g = 2 nm) without any adsorbates as presented in Fig. 22a. Accordingly, these are not concerned for the additional peak at longer wavelength that are experimentally observed for hot or blinking SERS particles with tiny amount of adsotbates. [51] H. Tamaru, H. Knwata, H. T. MiyazaM, and K. Miyano, Appl. Phys. Lett, 80 (2002) 1826. [52] A. Otto, Phys. Stat. Sol., (a) 188 (2001) 1455. [53] A. Otto, A. Bruckbauer, and Y. X. Chen, J. Mol. Struct, 661-662 (2003) 501. [54] X-L. Guo, Z-C. Dong, A. S. Trifonov, K. Mild, S. Mashiko, and T. Okamoto, Nanotechnology 15 (2004) S402 and references Iherein. [55] So far, concerning SMS-SERS it seems rather primitive approximation was adopted to estimate the enhancement fector in Raman scattering with LSP resonance at various metal nanostructures. For instance, distinct LSP resonance for the excitation and Stokes shifted light has never been taken into account as Otto suggested. Actually, the enhancement value was estimated simply by the fourth power of the amplitude enhancement obtained with various numerical methods. Also the real enhancement value is affected by the orientation of molecules irrespective of electronic resonance or not At least both of these factors should be involved in more accurate estimation. [56] W. E. Moemer, in Single Molecule Spectroscopy, R. Rigler, M. Qrrit, T. Bassche (Eds.), Springer, Berlin, 2002, Chap. 2.
Handai Nanophotonics, Volume 2 (Editors) S. Kawata and H. Masuhara (Editors) © 2006 Elsevier B.V. All All rights reserved. reserved.
141 141
Chapter 7
Enhanced Raman scattering mediated by metallic surface-particle gap modes S. Hayashi Department of Electrical and Electronics Engineering, Faculty of Engineering, Kobe University, Rokko, Nada, Kobe, 657-8501, Japan 1. INTRODUCTION 1.1 SERS and hot sites Research on surface-enhanced Raman scattering (SERS) has started around 1974, when Fleishmann et al. [1] reported on the Raman scattering of pyridine molecules adsorbed on Ag surfaces roughened by an electrochemical treatment. Since then, a variety of Raman data for various molecules adsorbed on various metallic surfaces have been reported [2,3]. Various metallic systems including Ag and Au island films, colloidal particles and roughened surfaces have been proved to be SERS active, leading to the Raman enhancement of ~104 to ~106. After a long debate, the enhancement of local electric fields near the metallic surface upon excitation of surface plasmon modes has widely been accepted as a major mechanism, although the chemical mechanism such as resonant Raman scattering mediated by a charge transfer state cannot be neglected in some cases. It should be noted that the Raman measurements performed at that period were macroscopic and the enhancement factors obtained were averages over macroscopic areas. In the second stage of SERS research, which started around 1997 and continues to develop up to now, local detection of Raman scattering was introduced by using the confocal microscope and the near-field microscope [4-7]. The enhancement factor as large as 10 w was reported in conjunction with the ability of single molecule detection. From the local Raman measurements, researchers have pointed out the existence of so-called hot sites, i.e., the sites which give rise to an extremely high enhancement. Typical examples of hot sites proposed so far are a gap between two metallic spheres [8,9], tips of metallic nanorods [10], corners of triangular prisms [11], etc. Although attempts to directly observe the hot sites have been reported previously [12], clear demonstration of the hot sites proposed with a sufficiently high spatial resolution has not yet been achieved. Clear experimental identification of hot sites may
142 142
S. Hayashi
further develop the SERS research and allow the application of hot sites to various optical phenomena including the fluorescence, second harmonic generation and other nonlinear optical processes. The present article is devoted to discuss the local field enhancement in a system of a metallic particle placed very near to a metallic surface. In such a metallic particle-surface system the gap between the particle and the surface becomes a hot site under an appropriate excitation condition. In what follows, the gap mode excited in the system is first discussed and evidence for SERS mediated by the gap mode [13] is presented.
medium
Fig. 1. System of a metallic sphere placed near a metallic surface.
1.2. What are the gap modes Let us consider a system of a metallic sphere of radius R placed on a metallic substrate at a distance D as shown in Fig. 1. In the particle-surface system surface plasmon modes are greatly modified from those of an isolated particle and the surface alone. As described in detail in our previous review paper [14], electromagnetic theories predict the existence of so-called gap modes, whose electric fields are strongly localized at the gap between the particle and the surface and strongly enhanced compared to those induced at an isolated particle or at the surface alone. When D and R are much smaller than the wavelength of light, retardation effects in electromagnetic fields can be neglected and distributions of electric fields in the system can be calculated within the electrostatic approximation. Note that results obtained within the electrostatic approximation depend only on the ratio D/R and not on the absolute values of D and R. Figures 2(a), 2(b) and 2(c) show electric field distributions calculated within the electrostatic approximation for a Ag sphere with i?=10nm placed on a Ag surface. The surrounding medium was assumed to be air and the literature values of the dielectric function for Ag [15] were used. In the calculation, p-polarized light incident on the surface at 45° was assumed and the distance D was varied
Enhanced Raman scattering mediated by metallic surface-particle gap modes
143
from 5nm to 0.5nm. The ordinate of the figure represents the absolute value of electric field induced at a point (x» z) divided by that of incident light. D=2nm (b) D=2nm
D=5nm (a) D=5nm
•«•
D=0.5nm (c) D=0.5nm
•t?
Fig. 2. Electric field distributions calculated for a Ag sphere (R = lOnm) placed on a Ag surface with D=5nm (a), 2nm (b) and 0.5nm (c).
Figure 2 demonstrates that as D decreases the electric fields are more and more localized at the gap and a very large enhancement of ~102 is achieved. It should be noted that the results shown in Figs. 2 (a), 2(b) and 2(c) are obtained for different wavelengths 365, 400 and 477nm, respectively, since the resonance wavelength depends strongly on D as shown in Fig. 3. In fact, the wavelength dependence of the maximum field intensity (relative to that of incident light) is plotted for three different D values in Fig. 3. We see that the main resonance peak shifts to longer wavelengths as D decreases. For small D values, subsidiary resonance peaks appear at the shorter wavelengths. As pointed out in ref.[14], the main resonance arises from the first-order gap mode, while the subsidiary resonances from higher-order gap modes. The results presented in Fig. 2 were calculated at the resonance wavelengths of the first-order gap mode. Since the results of calculation in electrostatic approximation depend only on the ratio D/R, results presented in Figs. 2 and 3 can be converted to those for various particle
144 144
S. Hayashi
diameters with a fixed particle-surface distance. When the distance is fixed at D=2xan, then the results presented in Figs. 2 and 3 can be converted to the cases of i?=4,10 and 40nm, respectively. 300
400
500
D=5nm
600
700
800
Ag particle on Ag substrate (R= 10nm)
D=2nm
D=0.5nm
400
500
600
700
800
Wavelength (nm)
Fig. 3. Dependence of the maximum field intensity on the wavelength.
From various calculations similar to those presented in Figs. 2 and 3 we can extract characteristic features of the gap modes, which can be summarized as follows: (1) A metallic particle-surface system supports a series of gap modes when the particle-surface distance is sufficiently small (£W?<~1). Higher-order gap modes become important for smaller values of DIR, (2) The resonance wavelength of the gap modes shifts to longer wavelengths as DIR decreases. (3) The maximum enhancement factor obtained upon excitation of the gap mode is larger than that achieved by an isolated particle or a surface alone. (4) The enhancement factor is larger for an excitation field perpendicular to the surface than for an excitation field parallel to the surface. Judging from these points, the gap in the particle-surface system is considered to be a very promising hot site to enhance various optical processes such as Raman scattering, fluorescence and second harmonic generation. However, there has been no clear direct observation of the enhancement caused by the gap modes. In the following sections, we discuss the enhanced Raman scattering mediated by the gap modes based on our experimental results [13].
Enhanced Raman scattering mediated by metallic surface-particle gap modes
145 145
2. EXPERIMENTAL PROCEDURES 2.1. Sample preparation The sample prepared is schematically shown in Fig. 4. First, a 40nm thick Ag film was vacuum evaporated onto a glass substrate. Secondly, a 5nm thick film of copper phthalocyanine (CuPc) was deposited also by vacuum evaporation. The thicknesses of these films were monitored by a quartz microbalance. On top of the CuPc film, a drop of Au colloidal solution was put and dried to finally obtain a CuPc thin film sandwiched by Au particles and an Ag film. We used commercially available Au colloidal particles (British Bio Cell International) with various diameters ranging from 20 to 250nm. Scanning electron microscopic (SEM) observation confirmed that these colloidal particles are spherical.
Au particle
O
O
P
CCuPc UPC metal glass
Fig. 4. Samples prepared.
2.2 SNOM-Raman measurement system Figure 5 schematically shows a SNOM-Raman measurement system constructed in our laboratory. The sample described in the above is attached to the basal plane of prism (BK-7) with the aid of an index matching oil and illuminated by laser light incident from the prism. This is the Kretchmann configuration of attenuated total reflection (ATR). The incident light was the 488,0nm line of an Ar+ ion laser and 632.8nm line of a He-Ne laser. The incident angle was fixed at the excitation angle of surface plasmon polaritons, which corresponds to the dip position of angle-scan ATR spectra. A sharpened tip of an optical fiber prepared by chemical etching was mounted on a x-y-z piezoelectric stage and scanned over the sample surface at a constant height mode. A SEM image of the tip is also shown in Fig. 5. In order to obtain SNOM images, light collected by the optical fiber was sent to a photomultiplier and the out put signal was fed into a lock-in amplifier. A piezoelectric stage driver and the lock-in amplifier are controlled by a computer and SNOM images are constructed in the computer. In order to obtain Raman spectra, a part of collected light is reflected by a beam splitter and sent to a Raman spectrophotometer equipped with a notch filter, a single monochromator and a CCD detector. The present SNOM-Raman measurement system allows us to measure Raman spectra of local sites on a sample which are well characterized by the SNOM imaging. The spatial resolution was estimated to be around 500nm.
146 146
S. Hayashi
photomultiplier Imaging
lock-in amp. piezo control notch filter
beam splitter |
monochromator /CCD CCD control
fiber probe sample
x
Raman spectroscopy
incident light p-pol.
Fig, 5, SNOM-Raman measurement system and tip of optical fiber prepared by chemical etching.
3. RESULTS AND DISCUSSION 3.1. Absorption measurements Before presenting Raman data, we first summarize the results of absorption measurements. Figure 6(a) shows absorption spectra of Au particles spread over a glass substrate obtained by a double-beam spectrophotometer (Shimadzu UV-3101PC). We see that as the diameter increases, the surface plasmon band slightly shifts to the red and broadens. This behavior of the surface plasmon band is in general agreement with the results of Mie calculation [16]. When the particles are spread over a silver surface the surface plasmon band changes dramatically as shown in Fig. 6(b). To obtain the spectra shown in the figure, reflection spectra were measured with unpolarized light by setting the incident angle at 5°. In the presence of Au particles the intensity of reflected light is decreased due to the absorption caused by the particle-surface system. Measured reflectivity R was converted to absorbance by a relation, - logl0 R. In Fig. 6(b) we see that as the particle diameter increases, the absorption band shows a larger shift than that observed for particles placed on the glass substrate. In fact, the band is located around 530nm for the particles 20nm in diameter and the main band shifts up to 1 lOOnm for the particles 250nm in diameter. Furthermore, for particles with the diameter larger than lOOnm subsidiary bands appear at the shorter wavelength side of the main band. We note here that these behaviors of the bands agree fairly well with the characteristics of the gap modes predicted by electromagnetic theories summarized in Section 1 of this article and also with our previous experimental results obtained for Ag particle-Al surface system reported in ref.[14]. We attribute the main band to the first-order gap mode and the subsidiary band to the higher-order gap mode. The results of reflection measurements
Enhanced Raman scattering mediated by metallic surface-particle gap modes
147
presented here will greatly help the interpretation of Raman results presented later.
Diameter Diameter 20nm 20nm
/ ~\ r \
/
(b) Au particles on surface Ag surface
(a) Au particles on glass substrate
20nm
Absorbance Absorbar ce (a.u.) (a.u.)
—•-"••—'
50nm 50nm
Diameter Diameter
—fcrt
/ 50nm
150nm 100nm
/ 100nm
150nm 150nm
j
150nm
/ 250nm 250nm 400
500
250nm 600
700
Wavelength (nm)
800
500
1000
1500
2000
Wavelength (nm)
Fig. 6. Absorption spectra of Au particles spread over a glass substrate (a) and reflection-absorption spectra of Au particles spread over a Ag surface.
3.2. SNOM image and Raman measurements Figure 7 (a) shows a SNOM image obtained for a Au particle lOOnm in diameter placed on a 5nm thick CuPc film deposited onto a 40nm thick Ag film. The 632.8nm line of the He-Ne laser with a power of 50mW was used as the incident light in the Kretchmann ATR configuration. The tip of the optical fiber was scanned at a constant height mode setting the height from the CuPc surface around 3Q0nm. The image obtained appears to be somewhat noisy showing fragments of fringes around the round image of the particle. It is very likely that these fringes are caused by the interference between the incident surface plasmon polaritons excited by the ATR method, which propagate from the right to the left on the image, and those scattered by the presence of the Au particle. In spite of the appearance of the noisy signals the present SNOM system allows us to resolve the single Au particle. Figure 8 compares Raman spectra obtained by placing the tip of optical fiber just on top of the Au particle and far from the particle on the surface. These spectra were obtained with the CCD exposure time of 120sec. When the tip is placed on top of the particle, the Raman spectrum obtained clearly exhibits peaks attributable to the molecular vibrations of CuPc. On the other hand, when the tip is placed far from the particle, Raman peaks are very weak and almost at the detection limit of our measurement system. We can clearly see that the presence of the Au particle enhances the Raman intensity by a factor of 102.
148 148
S. Hayashi
Light intensity (a.u.)
1D0
e •
tc
0
-1.0
-0-5
-00
Raman intensity (counts)
Fig. 7. SNOM image of a Au particle placed on CuPc/Ag surface (a) and profiles of SNOM image and Raman intensity along the lines A-A' (b) and B-B' (c). •
400 --
B
I
•
I
•
I
•
I
•
i
• •
i
Au particle 100nm on Ag film 632.8nm Excitation
300 -
on Au particle
A
n
t 200 100
iJ V V
V
\L
~
\wr
on Ag surface 1100
1200
1300
1400
1500
1600
Raman Shift (cm " 1 )
Fig. 8. Comparison of Raman spectra obtained by placing the tip of optical fiber just on top of the Au particle and far from the particle on the surface.
Enhanced Raman scattering mediated by metallic surface-particle gap modes
149 149
In Figs. 7(b) and 7(c), profiles of the SNOM image (intensity of light collected) along the lines A-A' and B-B' indicated in Fig. 7 (a) are plotted together with the intensities of the 1530cm*1 Raman line measured point by point along the same lines. We can see that the variation of the Raman intensity coincides very well with the SNOM profiles and takes a maximum at the top of the Au particle. Since the CuPc molecules are located in the gap between the Au particle and the Ag film, the enhancement of Raman scattering presently observed is thought to be caused by the excitation of the gap mode. Similarly to a widely accepted mechanism of SERS, which involves the in-going resonance (at the frequency of excitation light) and out-going resonance (at the frequency of scattered light) to surface plasmons, the Raman enhancement presently observed is thought to arise from the steps summarized below. First, the light incident in the ATR geometry excites the gap mode of the Au particle-CuPc-Ag surface system and induces strong electric field in the CuPc layer at the gap (excitation of the gap mode at *»,). Second, the strong electric fields oscillating at eo, excite Raman dipoles of CuPc molecules oscillating at the Raman frequency atR. Third, the Raman dipoles act as the excitation source of the gap mode at the frequency mK. Since the gap mode is a radiative mode, the gap mode excited at mR now decays radiatively. What we observe through the optical fiber tip is the radiative decay of the gap mode oscillating at a)R, but is not the Raman light coming directly from the CuPc molecules. In order to confirm the Raman enhancement mediated by the gap mode, we examined the dependence of the enhancement factor on the particle size by using two different excitation laser wavelengths, i.e., 488.0 and 632.8nm. Figures 9(a) and 9(b) show the size dependence of the enhancement factor for the 1530cm'1 Raman line obtained for 488.0 and 632.8nm excitations, respectively. For the 632.8nm excitation, since CuPc molecules are under a resonance Raman condition, Raman signals of CuPc films as thin as 5nm deposited onto glass substrate could be detected by our SNOM system. Therefore, the enhancement factor was obtained by simply dividing the Raman intensity for the gap-surface system by that of 5nm CuPc film on a glass substrate. For the 488.0nm excitation, Raman signals from 5nm CuPc films on glass substrate were too weak to be detected by our SNOM system. Therefore, we prepared a 5nm CuPc film deposited onto a 5nm Ag island film, which gives sufficiently strong Raman signals to be detected by our SNOM system. The Raman intensity obtained for the gap-surface system was compared to that of the CuPc film on the Ag island film, and then converted to the enhancement factor by referring to the enhancement factor of Ag island film determined in our previous paper [17]. It should be noted that the enhancement factor estimated for 632.8nm excitation is considerably lower than that for 488.0nm excitation. This result is in good agreement with previous results of non-resonant and resonant SERS experiments [18], in which the enhancement factor in resonant SERS was shown to be several orders of magnitude lower than that in non-resonant SERS.
150 150
S. Hayashi 1
1
1
1
1
1
1400 Q1200
Excitation (a) 488.0nm Excitation
•21000
-
^ 160 o 45 120
(b)632.8nm Excitation Excitation
| 800 -
8 80
§ 600 -g 400
m 40
HI
200
0 t
i
T
T
i
50
100
150
200
Diameter (nm)
i 250
Oti0
_|_ 50
100
150
200
250
Diameter (nm)
Fig. 9. Dependence of enhancement factor on the particle diameter for 488.0nm excitation (a) and 632.8nm excitation.
As can be seen in Figs. 9(a) and 9(b), the enhancement factor depends strongly on the particle size. For the 488.0nm excitation, the smallest particle in our experiment with the diameter of 20nm gives rise to a very large enhancement factor, while the other particles do not result in large enhancements. For the 632.8nm excitation, on the other hand, the Au particle lOOnm in diameter brings about the largest enhancement. The behaviors of the enhancement factor seen in Figs. 9(a) and 9(b) can be explained very well by the mechanism of Raman enhancement summarized previously. An important point is that the Raman enhancement arises from two resonances, i.e., resonance of the incident light to the gap mode at m, and that of the scattered light to the gap mode at coK. As presented in Fig. 6(b), the position of the reflection-absorption peak attributed to the first-order gap mode of Au particles placed on the Ag surface depends strongly on the particle diameter. Namely, the peak shifts to longer wavelengths as the diameter increases. When the Raman scattering is excited by 488.0 and 632.8nm laser lines, the wavelengths corresponding to the 1530cm'1 Raman line are 527.4 and 700.6nm, respectively. These wavelengths are indicated as vertical broken lines in Fig.6 (b). We now compare the location of the gap mode peak with these wavelengths. We see that the scattered light wavelength 527,4nm is in good resonance with the gap mode for the particle 20nm in diameter and becomes off resonant when the particle diameter increases. This explains very well the size dependence of the enhancement factor shown in Fig. 9(a). Furthermore, we see that the wavelengths of 632.8 and 700.6nm coincide rather well with the relatively broad peak of the gap mode for the lOOnm particle. As the particle diameter increases, the first-order gap mode peak shifts to longer wavelength and the wavelengths are off resonant with the first-order gap mode. Although a subsidiary peak corresponding to the higher-order modes appears as the diameter increases and the wavelengths coincide more or less with the subsidiary peak, the enhancement caused by the higher-order modes is expected to be smaller than that
Enhanced Raman scattering mediated by metallic surface-particle gap modes
151 151
caused by the first-order mode as suggested from the results of calculation shown in Fig. 3. All these facts can explain the size dependence of enhancement factor presented in Fig. 9(b). The size dependence of enhancement factor and its good correlation with the resonance behaviors of the incident and scattered light to the gap mode confirm the Raman enhancement mediated by the gap mode, 4. CONCLUDING REMARKS In this article, we focused our attention to the gap modes of a metallic particle-surface system and gave experimental evidence of enhanced Raman scattering mediated by the gap modes for CuPc molecules lying between a Au particle and a Ag surface. From the dependence of Raman enhancement factor on the Au particle size and the excitation wavelength, it was suggested that the resonance of incident and scattered light to the gap mode is important to obtain a high enhancement factor. In so-called apertureless SNOM, a sharpened metallic tip is brought very close to an object and scattering of near fields around the metallic tip is detected to obtain SNOM images and to perform spectroscopy. According to what is described above, if the object is placed on a metallic surface a high enhancement of near fields can be expected, since the sharpened metallic tip can be approximated by a sphere. However, it should be noted that the incident light must be tuned to the gap mode of the tip-surface system, which depends on the size of the tip and the tip-surface distance. Although a variety of experiments about apertureless SNOM has been published, to the author's knowledge, attention was not paid on the resonance wavelength of the gap mode in the tip-surface system. Further optimization of the performance of apertureless SNOM may be achieved by appropriately tuning the incident wavelength to the resonance wavelength of the gap mode. As demonstrated in this article, the metallic particle-surface system is relatively easy to prepare. Applications of the gap mode enhancement to various optical phenomena, such as the fluorescence, second harmonic generation and other nonlinear optical processes, may be straightforward. REFERENCES [1] M. Fleischmann, P. J. Hendra, and A. J. McQuillan, Chem. Phys., Lett., 26 (1974) 123. [2] A. Otto, Light Scattering in Solid IV, Topics in Applied Physics, 54, eds. M.Cardona and G.Guntherodt, Springer-Verlag, Berlin, 1984, Chap.6, p.289. [3] M. Moskovits, Rev. Mod., Phys., 57 (1985) 783. [4] S. Nie and S. R. Emory, Science, 275 (1997) 1102. [5] S. R. Emory and S. Nie, Anal. Chem., 69 (1997) 2631. [6] K. Kneipp, Y. Wang, H. Kneipp, L. T. Perelman, I. Itzkan, R. R. Dasari, and M. S. Feld, Phys. Rev. Lett., 78 (1997) 1667. [7] K. Kneipp, H. Kneipp, V. B. Kartha, R. Manoharan, G. Deinum, I.Itzkan, R. R. Dasari, and M. S. Feld, Phys. Rev., E 57 (1998) R6281.
152 152
S. Hayashi
[8] H. Xu, E. J. Bjerneld, M. Kail, and L. Borjesson, Phys. Rev. Lett., 83 (1999) 4357. [9] H. Xu, J. Aizpurua, M. Kail, and P. Apell, Phys. Rev., E 62 (2000) 4318. [10] A. Bouhelier, M. R. Beversluis, and L. Novotny, Appl. Phys. Lett., 83 (2003) 5041. [11] E. Hao and G. C. Schatz, J. Chem. Phys., 120 (2004) 357. [12] V. A. Markel, V. M. Shalaev, P. Zhang, W. Huynth, L. Tay, T. L. Haslett, and M. Moskovits, Phys. Rev., B 59 (1999) 10903. [13] S. Hayashi and T. Konishi, Jpn. J. Appl. Phys., 44,5313 (2005). [14] S. Hayashi, Near-Field Optics and Surface Plasmon Polaritons, Topics in Applied Physics, 81, ed. S.Kawata, Springer-Verlag, Berlin, 2001, p.71. [15] D. W. Lynch and W. R. Hunter, Handbook of Optical Constants of Solids, E.D.Palik (ed.), Academic Press, Orlando, 1985, p.350. [ 16]T. Okamoto, Near-Field Optics and Surface Plasmon Polaritons, Topics in Applied Physics, 81, ed. S.Kawata, Springer-Verlag, Berlin, 2001, p.97. [17] S. Hayashi and M. Samejima, Surf. Sci., 137 (1984) 442.
Handai Nanophotonics, Volume 2 S. Kawata Kawata and H. Masuhara Masuhara (Editors) (Editors) © 2006 Elsevier B.V. All rights reserved.
153 153
Chapter 8
Surface plasmon enhanced excitation of photofunctional molecules in nanospace towards molecular plasmonics A. Fujii and A. Ishida Laboratory of Material Science, Department of Human and Environmental Science, Kyoto Prefecture University, Shimogamo Hangi-Cho 16, Sakyo-Ku, Kyoto 606-8522, JAPAN. 1. INTRODUCTION The next-generation optoelectronics devices will consist of semiconductors, organic molecules, metal complexes, and biological molecules that are regularly positioned in nanospace [1]. One of the most important key technologies is probably near-field optics, which enables light to be handled between the nanospace and far-field [2]. Among optical near-fields, surface plasmon phenomena in particular are expected to be able to be adopted as interfaces of molecules, light, and metals in nanospace, as they can localize the energy of light near the surfaces of metals [3]. Therefore, plasmonics must be one of the most promising key technologies in the development of the next-generation optoelectronics devices. The study of interactions between plasmons and molecules is extremely important, as it provides much information that is essential to the practical application of plasmons. It also provides spectroscopic information through probing of the properties of plasmon electromagnetic fields using molecules. However, study in this field has only recently begun and there are few reports. An example of an application of the surface plasmon (SP) that is commercially the most successful is the surface plasmon resonance (SPR) sensor [4]. It has become an essential tool for biochemical and physiological studies, as it is capable of detecting molecular binding interactions, such as immune reactions, with high sensitivity and in real time. However, the SPR sensor only observes changes induced in the refractive index on the surface of the gold film by molecular interactions; hence the energy of the incident light is consumed as heat in the end. The electromagnetic fields of the SPs are very attractive as a source of excitation of the thin film of photoredox-responsive
154 154
A. Fujii and A. Ishida
molecules that are immobilized on metal film surfaces, because it propagates over a long distance on the film surfaces of such metals as gold and silver and, compared with the incident light, is enhanced several tens of times, even in the visible region [3]. Also, an interface based on electronic conduction can be constructed between the molecule and the external circuit, because gold and silver can be used as electrodes. On the other hand, quenching of photoexcited states cannot be avoided if the immobilized molecule is directly excited on the metal film [6], and this will cause the acquired excitation energy to be spoiled. Therefore, the pioneering of an effective method of preventing quenching is required to apply plasmons as excitation sources of photofunctional molecules. Furthermore, recent attention has been drawn to the localization of plasmon electromagnetic fields and the control of the spatial distribution in which metal nanoparticles [8], nanorods [9], and the nanostructure of the metal surface are used [10-13]. These new plasmonic media will provide new kinds of biochemical and chemical analyses and photochemical reactions, which will open up new practical applications. The science of the plasmon has now entered its second generation, which is directed toward plasmonics. In this chapter, the abilities and limitations of the SP fields as an excitation source and the studies based on the new tactics we are developing for breaking the limit are described. 2. HISTORICAL BACKGROUND 2.1. Excitation of fluorescent molecules by SP fields Studies of the interactions between SP fields and organic photofunctional molecules date quite far back; the oldest known was by researchers at IBM in 1978 [14]. At that time, they had already conducted angle-resolved measurements of fluorescence to reveal the interaction between SP fields and fluorescent molecules in the excited state. However, for around 20 years afterward, application of the SP fields to fluorescence spectroscopy was not attempted. This is in contrast to the fact that the application of Raman spectroscopy was actively studied in the same era. Only after SPR spectroscopy for the purpose of biochemical analyses came onto the market was the application of enhanced plasmon fields to fluorescent molecules attempted. The use of a silver film that was illuminated by attenuated total reflection as a substrate enabled the excitation of fluorescent molecules in enhanced plasmon fields. Using this technique, the observation of a biological molecular motor protein and multiphoton excitation of single crystal of a fluorescent dye [15], both of which are difficult by conventional methods, were attempted and reported on. However, there were fewer developments after these advances than expected. The reason seems to be the fact that it is difficult to avoid several serious problems that these researchers confronted in plasmon-
Surface plasmon plasmon enhanced excitation of photofunctional photofunctional molecules molecularplasmonics in nanospace towards molecular plasmonics
155 155
excited fluorescence spectroscopy, including quenching of fluorescence by the metal and denaturing by strong binding interactions between the metal and protein molecules.
Conventional SPR
SP enhanced excitation
Detection of change in the dielectric constant light energy . = >
thermal loss
Effective excitation of photofunctional molecules Suppression of reflection loss Active use of remarkable field enhancement Field localisation by nano-constructed surface
Scheme 1. Concept of SP enhanced excitation of photofunctional molecules
There is a strong interaction between the photoexcited state of molecules and metals that exist close to each other all the time. In the study of photoelectrochemical reactions, researchers have been aware of some bizarre occurrences in experiments - such as the fact that the surface roughness of the gold and silver electrodes on which the molecules are immobilized is strongly correlated with the generation efficiency of the photocurrent; that even a bare electrode that does not possess immobilized molecules has a weak photovoltage upon photoirradiation; or that the long-wavelength part of the action spectrum is enhanced on an electrode with a rough surface - and have imagined that plasmons must be involved in these occurrences. Despite its expected superior properties, however, no attempt to actively apply SP fields as an excitation source was reported until the 1990s. 2.2. Excitation of semiconductor photocells by SP fields The first example of success in applied photoelectrochemical research using a plasmon as the excitation source occurred in the research conducted by Hayashi et al. on the development of an organic semiconductor photocell that could be excited by SPs [16]. They studied the SP excitation of a copper phthalocyanine photocell and demonstrated the intensity of the SP field; accurate measurement of the angular dependence of incident light on this field was theoretically expected. Almost at the same time, in the branch of
156 156
A. Fujii and A. Ishida
photoelectrochemistry that mainly uses organic photoredox-responsive molecules, researchers started to actively study the self-assembly monolayer (SAM) that is fabricated by the reactions between gold or silver and organic compounds that possess a thiol group [17]. The SAM is very superior as a method of producing photoresponsive electrodes, as it is easy to fabricate and durable, and various systems can be constructed by combining several kinds of molecules. However, it has a major disadvantage in that, being a monolayer, its light absorbance is extremely small, so most of the incident light is wasted by reflection or transmission when conventional direct photoillumination is used. We proposed the application of SP electromagnetic fields - for the first time as far as we know - as an effective answer to this problem and have developed our research accordingly [18-23]. We have demonstrated that the SP, which is enhanced several tens of times compared with the incident light and propagated over a long distance on the surface, enables highly efficient excitation of the SAM. The results inspired new applied research on SP-excited fluorescence analyses [24]. In the next section, we will describe mainly the results of our research and the properties of the SP as an excitation source for thin films, and its limitations. 3. EXCITATION OF THE PORPHYRIN SAM BY SP FIELDS 3.1. SAM as a photofunctional thin film Gold and silver possess moderate chemical reactivity and are especially useful in reactions with thiol and disulphide groups. For example, simply by making a bare metal film in contact with a solution of alkanethiol with a long alkyl chain or its disulphide, oxidation of the thiol group or degradation of the disulphide bond is induced, and the molecule is immobilized on the surface of the metal by covalent bonding [25]. On the surface, many molecules are immobilized one after the other after the initial reaction of a small number of molecules; therefore, the Van der Waals force that works between neighbouring molecules regulates the spaces between the molecules. As a result, on the film that is obtained the molecules are arranged regularly, standing out from the surface and oriented at a certain tilting angle, like a brush. This is the reason why this kind of film is called a SAM. Accurate observation by several research groups using scanning probe microscopy or angle-resolved X-ray photoelectron spectroscopy has revealed the dynamic process of this SAM formation reaction [26]. A SAM of alkanethiol that is formed in this way is extremely dense and sturdy. For example, in electrochemical analysis by cyclic voltammetry, modification of the electrode with an alkanethiol SAM blocks the transmission of redox-responsive species in the solution completely and eliminates the redox peak [27]. Alkanethiol can be used only for the purpose of protecting the surface
Surface plasmon plasmon enhanced excitation of photofunctional photofunctional molecules molecularplasmonics in nanospace towards molecular plasmonics
157 157
of the metal as a simple insulation film. However, the use of molecules of which the headgroups can be excited by photoirradiation is connected to alkanethiol provides a SAM that can function as a photoredox-responsive electrode. O NH N N HN Au d = 50 nm
BK7
< 10–55 Pa Pa <10"
N H
(CH2)n–S
N HN
N HN
N HN
NH N
NH N
NH N
HN
HN
HN
N HN NH N
2
1a, n=3
O
O
O
HN
O
1b, n=10 S
S
S
S
Au
Au
Au
Au
0.5mMDCE 0.5 mM DCE
Scheme 2. Formation of photofunctional self-assembly monolayer (SAM) on gold; in this case, the photofunctional moiety is a porphyrin.
It was because of this that several research groups have conducted much research on organic photocells or photosensors based on SAMs, and these groups have demonstrated the effectiveness of the photoelectrochemical application of SAMs [28]. Although their practicality as photocells might not be high because of concerns about durability, the use of SAMs with molecular recognition properties provides an attractive possibility for the application of molecular sensors [29]. However, there is a big disadvantage in using SAMs that are photoresponsive: the majority of the incident light is wasted by reflection or transmission when direct illumination is used for excitation, because the molecular layer is extremely thin and the absorbance is usually very small. This property of SAMs becomes a major issue in practical application, as it decreases the sensitivity of photoresponsive molecular sensors that are based on SAMs. In light of this history, the development of a method that would enable effective excitation of monolayer and multilayer thin films with small absorbances was eagerly wanted. We considered that excitation using SP electromagnetic fields were most suitable for this purpose and therefore began our research. 3.2, Preparation of the porphyrin SAM The properties required of molecules to verify the effectiveness of excitation by SP electromagnetic fields are appropriate photoredox properties and an absorption spectrum that possesses individual absorption bands over a wide wavelength region of visible to near-IR. The freebase porphyrin is appropriate for this purpose, as it possesses appropriate photoredox properties and has discrete absorption bands, called the Soret band and the Q bands, at wavelengths of around 400 nm and above 500 run, respectively [30]. Therefore,
158 158
A. A. Fujii and A. Ishida
we composed a disulphide that possessed a porphyrin freebase as terminal groups [31] and dipped a gold film into the solution to form a SAM [18-23]. The SPR (Fig. 1) and transmission absorption measurement clearly showed the irreversible formation of a porphyrin SAM merely by several minutes' dipping of a gold film in the solution of the porphyrin disulphide. The porphyrin SAM obtained as a result was very sturdy and was stable for more than several months, even in the air. 1.000 in air, n= 1.000 1.0
0.8
1.362 in water, n= 1.362 0
before modification
700 8
600
6 400
0.4
4
0.2
2
300
photocurrent / nA
reflectance
500 0.6
200 100
after 1 min modification 0.0 0 40 42 44 46 48 50 5066 66
0
68 68
70 70
72 72
74 74
76
θ/Þ
Fig. 1. SPR reflectivity curves for the gold film on a BK-7 right angle prism (open circle) and that modified by -S(CH2)ioCONH-Por (closed circle) in air (left) and water (right) and dependence of the photocurrent on the incident angle (right, open triangle).
Determination of the thickness of the film by SPR measurement suggested that the molecules were immobilized, regularly arranged, and standing up from the surface at a similar tilting angle as that seen in the alkanethiol SAM. The absorbance in the transmission absorption spectrum was very small, which means that the majority of the light would be wasted by transmission if conventional direct photoirradiation were applied. The area occupied by porphyrin in the film was 20 nm2. The shoulder found in the Soret band of the absorption spectrum, with significant broadening, together with the weak shoulder found at 475 nm, suggests (as mentioned later) the formation of H- and J-aggregates by stacking of the porphyrins [32]. The fact that SPR measurement by 633 nm excitation showed effective light absorption clearly indicated that the porphyrin in the SAM effectively damped the SP electromagnetic fields those were excited by the wavelength. This result led us to expect electronic excitation of the porphyrin molecules in the SAM, as the porphyrin chromophore possesses strong photoabsorption at 633 nm.
Surface plasmon enhanced excitation of photofunctional molecules in nanospace towards molecular plasmonics
159 159
3.3. SP enhanced fluorescence of the porphyrin SAM When SP excitation was conducted at an incident angle of 55° with ppolarised 420 nm light, a fluorescence spectrum characteristic of a porphyrin freebase was obtained, similar to the one obtained by direct illumination (Fig. 2). Despite the fact that the same intensity of excitation light source was used, the fluorescence intensity was far stronger compared with direct photoexcitation [18]. This result, as we expected, demonstrated that SP excitation effectively suppress reflection loss and vastly improved excitation efficiency. 0.4 fluorescence intensity Tluor
:
C10 SP A C 1 0SP
r
0.3 §0.3
\
c
d eoueose
C10 direct I
0.2
0.1
C3 SP direct
\v A\
: spJ/ n 0 600
650
700 wavelength nm wavelength // nm
750
Fig. 2. The fluorescence spectra of the CIQ and Cj porphyrin SAMs by SP excitation and conventional direct illumination, clear differences in the intensity indicate the efficiency of SP excitation and remarkable quenching in the Cj SAM having much shorter methylene chains than the Cio SAM.
SPs can be excited only by p-polarised incident light, and the electromagnetic fields obtained as a result are p-polarised. If the porphyrin chromophores in the SAM are excited by direct electronic interaction with the SP electromagnetic fields and the porphyrin chromophore is fixed rigidly in the SAM, then the fluorescence obtained must have a strongly preserved ppolarisation. The result of fluorescence depolarisation measurement of the porphyrin SAM was as we had expected [20]. On the other hand, fluorescence in a film when Cw porphyrin disulphide was partly doped to a SAM of butanethiol gave only small p-polarisation. This result suggests that a porphyrin chromophore that possesses Cio legs that stick out on a SAM of thin butanethiol is rotating or vibrating [20]. It is possible to observe the motion of the fluorescent molecules in a thin film by using SP excitation like this. This method is probably effective for real-time observation of the molecular dynamics in living cells or in a molecular assembly that possesses vesicle-like flowability.
160 160
A. A. Fujii and A. Ishida
3.4. Photocurrent generation of the porphyrin SAM by SP excitation Porphyrin is a photoredox-responsive molecule and its SAM can be used to constitute a wet solar cell by being dipped in electrolytic solution that contains an electron acceptor. All wet solar cells in the past were excited by conventional direct photoillumination, so that incident light could be transmitted to the molecular layer only once or twice. As a result, excitation efficiency was very low, as most excitation light was wasted, and this was a serious technical issue that needed to be addressed. Conventional direct photoillumination of the porphyrin SAM actually gives only a weak photocurrent. In comparison to this, SP excitation gave a peak current that was more than ten times larger (Fig. 3) [19]. Of course, this was because of the excitation efficiency was remarkably improved by effective suppression of reflection loss. Therefore, SP excitation is effective not only for fluorescence but also for photocurrent generation. This result is bound to be extremely useful for the development of molecular sensors and artificial photosynthesis where photoelectrochemical reactions are applied.
1.6 1.6 1.4 1.4 1.2 1.2
I 0.8 0.8 -
40
,
• "
20 10 10 SP excitation
1
I 0.6
3
[MV2+] / mmol dm dm"-3 [MV
is. ^
^ 1.0 1.0 -
photocurrent /
2+
•
0
µ A
, 1
0.4 "0.4
,
0.0
illunination direct illunination
1
0.2
. 1 . . . .
0
4 light on
5
1 . . . . I . . . . I . . . ._
10
15
20
25
time/s time /s
Fig. 3. Time profiles of the photocurrent of-S(CH 2 ) 10 -CONH-Por SAM in oxygen saturated 0.1 mol dm"3 aqueous solution of Na2SC>4 upon the SP excitation at 6 = 73° in the absence and presence of 10, 20, and 40 mmol dm"3 of MV 2+ at 26°C. The lowest line is a time profile by direct illumination in the absence of MV 2+ .
3.5. Field enhancement effect of SP excitation The biggest advantage of SP excitation is the field enhancement effect, by which the electromagnetic fields are far stronger than incident light appears on the surface. The longer the wavelength, the larger the effect, and the effect is much larger for silver than for gold [3,5]. This field enhancement effect is reflected in the fluorescence excitation spectrum and photocurrent action
Surface plasmon plasmon enhanced excitation of photofunctional photofunctional molecules in nanospace towards molecular plasmonics
161 161
spectrum, and we were able to prove the effectiveness of SP enhanced excitation by measuring them. A film sample that has very small absorbance, like the SAM, usually gives a fluorescence excitation spectrum and photocurrent action spectrum very similar to the transmission absorption spectrum. However, the fluorescence excitation spectrum and photocurrent action spectrum by SP excitation were both very different in shape from the transmission absorption spectrum, and their long wavelength regions were significantly enhanced [18, 21]. We assumed that the product of absorbance of porphyrin SAM and the field intensity [5] gives a spectrum that shows the efficiency of SP excitation, and that the spectrum is rather similar to both the fluorescence excitation spectrum and the photocurrent action spectrum. However, both of these latter spectra have even more enhanced long-wavelength regions than the spectrum assumed from field intensity. This even larger enhancement effect than expected may be ascribed to the grain structure on the surface of the gold that induces the particle plasmon. In fact, a porphyrin SAM formed on a rough gold surface evaporated under a low vacuum condition gave an even more enhanced long-wavelength region of the spectrum compared with that formed on flat-surfaced gold evaporated under an ultrahigh vacuum condition. 8 0.05 0.05
absorption action fluorescence excitation excitation fluorescence
6
absorption absorption
0.04 0.04
0.03 0.03 3)
absorbance
4
0.02 0S>2 x4 2 0.01
0 400 400
500 500
600 700 600 700 nm wavelength / nm
800 800
0 900 900
Fig. 4. Spectroscopic properties of the Cio porphyrin SAM; difference absorption spectrum between the modified and bare gold films, dotted line; fluorescence excitation spectrum by SP excitation monitored at 725 nm, grey solid line; corrected action spectrum of the photocurrent generation by SP excitation in the presence of 40 mmol dm"3 methylviologen, black solid line.
The Soret band of the action spectrum split into two bands, the new band at 395 nm and an original Soret band at 425 nm, and a small band at a wavelength
162 162
A. Fujii and A. Ishida
of around 470 nm was also found (Fig. 4). These were the bands that originally did not exist in the absorption spectrum of the porphyrm chromophore measured in the solution. We assumed that these new bands were attributable to the molecular assembly formed by the stacking of the porphyrin chromophore and referred to as the H and J aggregates, respectively [21]. Furthermore, several discrete bands were found in the near-infrared region. These long-wavelength bands might be attributable to the highly ordered J aggregates [32]. These longwavelength bands cannot be observed at all in a normal transmission absorption spectrum, as it is too weak, but in the SP excitation action spectrum they could be observed clearly because of the remarkable field enhancement effect. 3.6. Quenching of photoexcited states: a serious problem of SP excitation As described above, we were able to demonstrate the effectiveness of SP excitation by our research of the fluorescence of porphyrin SAMs and photocurrent. However, it became obvious that we had a task that could not be avoided to prevent quenching of the photoexcited states of a molecule by the gold. It is well known that metals such as gold and silver quench the photoexcited states of molecules extremely rapidly by energy transfer [33]. By theoretical research, the energy transfer quenching of photoexcited molecules on the surface of the gold is in inverse proportion to the cube of the distance. Recently, several groups, in experiments using film spacers, obtained results in which the fluorescence intensity was in inverse proportion to the cube of the distance from the surface of the gold [34]. Therefore, to effectively prevent quenching, sufficient distance must be secured between the photoresponsive part of the immobilized molecule (the chromophore) and the surface. As the SP propagates several tens of nanometers in a vertical direction [3], even molecules on quite thick molecular film can be sufficiently excited. However, it is not easy to secure sufficient distance, because of the limitations of synthesis, if SAMs that employ thiol or disulphide are used. The length of the methylene chain that connects the chromophore and the thiol group that is anchored to the surface of the gold is around 14 at the maximum, and to extend it further requires synthetic reactions that are extremely difficult, which lacks actuality. The fluorescence quantum yield of the porphyrin SAM that we obtained using a chain length of 12 (the Cio derivative) was indeed merely 0.0003 and significantly lower than that of porphyrin in solution [20]. Therefore, to use SPs effectively as excitation sources it is essential to find an efficient tactic for transmitting sufficient SP electromagnetic field to the photoresponsive molecules, as well as to prevent quenching by gold.
Surface plasmon enhanced excitation of photofunctional molecules in nanospace towards molecular plasmonics plasmonics
163 163
4. STRATEGY TO PREVENT QUENCHING UPON SP EXCITATION To expose the photoresponsive molecules to sufficiently intense SP electromagnetic field at the same time as preventing quenching by gold, it is necessary to immobilize the molecule that is to be excited at the most appropriate distance from the gold film. This distance is the one that has sufficient field intensity but without quenching. Therefore, there are two tactics. One is the traditional method of securing vertical distance from the gold using a spacer. Several research groups have avoided quenching by using this method and have promoted research on SP excited fluorescence analysis [24]. The other is a new tactic that we have proposed, whereby horizontal distance from the gold is secured by using a nano-constructed gold film. If a spacer is used to secure vertical distance, then the properties of the molecular environment that is obtained will depend strongly on the properties of the spacer. If the spacer is rigid and the surface is smooth, then the chromophore that is immobilized there can efficiently couple with p-polarised SP field, enabling an application that utilises fluorescence polarisation. However, it is usually fairly difficult to construct a spacer as smart as that. If a thick film is created by using a polymer, many molecules can be easily immobilized on the surface, but the structure of the surface of the polymer is usually irregular, so that the environment that surrounds the molecules immobilized on it will probably be very irregular. Therefore, a spacer layer that consists of a polymer should probably be applied for processes such as fluorescence analysis or photochemical synthetic reaction, which place high priority on sensitivity (proportional to the number of immobilized molecules), without using fine properties such as polarisation measurement. Several groups have investigated SP excited fluorescence analysis and have demonstrated that the dynamic interactions between molecules can be accurately measured by immunoassay and DNA assay [24]. As mentioned later, our interest in this tactic is not the analysis but the photochemical synthetic reaction, in that the photochemical reaction is induced by SP excitation of a thylakoid membrane. On the other hand, some interesting phenomena have recently been reported in relation to localisation of plasmon electromagnetic fields by metallic apertures (nanoholes), such as the extraordinary efficient light transmission that is observed when a nanohole of a rim diameter less than the wavelength of light is made in a film of a metal such as silver and gold [10-13]. Localisation of the plasmon electromagnetic fields in the nanohole is interesting, in terms of not only spectroscopy but also its practical application as an excitation source. This is because a nanohole that is formed on a transparent substrate such as glass and polymer can be used as a vessel for photochemical reaction or fluorescence analysis and can also be used to secure distance from the metal by the
164 164
A. A. Fujii and A. Ishida
immobilisation of a photoresponsive molecule on the bottom substrate or by packing it inside the nanohole, thus effectively preventing quenching. ii-TAS fibre sensor light emitting device
spacer
etc. integration
"nanowell"
metal film glass substrate
Scheme 3. Strategy to prevent quenching upon SP excitation left, application of a spacer layer or a joint right, application of a nanowell
This tactic is expected to be applicable to micro total analysis systems (jxTAS) of detecting reactions and molecular interactions. However, there is a possibility that the complexity of distribution of the field localised in the aperture can cause difficulty in smart applications that utilise polarisation. Therefore, for systems that use nanoholes as well as ones that use spacers, it is probably necessary to start this kind of research with applications that place high priority on sensitivity. Talcing this background information into account, we studied fluorescence properties by immobilisation of fluorescent molecules on the bottom surface of the nanohole. Furthermore, to try the development of even more practical applications in immunoassay and DNA assay, we also investigated the light scattering and transmission properties of a gold nanohole the diameter of the wavelength of light, prepared on a glass substrate [35]. An overview of our achievements by these two tactics is given below. 5. SP ENHANCED EXCITATION OF THYLAKOID MENBRANE 5.1. Characteristics of photosynthetic system and difficulties faced in artificial photosynthesis In the photosynthetic system of plants we find a photocatalytic reaction system that is completely optimised. Plants have realised extremely efficient photoinduced charge separation and, subsequently, extremely efficient electron transport, by using a protein membrane structure that is highly organised [36]. The successful application of so called "artificial photosynthesis", which is an attempt to construct a complex molecular system that induces photoinduced charge separation efficiently, is yet to be reported, despite the enthusiastic
Surface plasmon enhanced excitation of photofunctional molecules in nanospace towards molecular plasmonics
165 165
research carried out around the world over the past 30 years. The most serious problem is the low efficiency of the initial charge separation as a result of back electron transfer. To overcome this, complex molecules that try to use multistep electron transfer have been constructed by extremely difficult synthetic reactions [37]. Although those compounds have been providing very interesting and useful results for spectroscopy, sadly there have been hardly any practical achievements. In other words, the results of research over the past 30 years clearly show that the performance of artificial molecular systems is currently far behind that of natural photosynthetic systems. Separation Photoinduced Charge Separation Ps-l hv hν Electron A
A00 . . Transport Transport
\
+
hν \
Ps-ll Electron l' .Transport Transport
4H+++O 4H O22 \ A
...:.
ft
"
Excess Electron
P700
2H22 O -JMn
'P680 P680
Damage
Oxygens Reactive Oxygens
Photoinhibition
immobilization on gold surface microfluidics system substrates H2O
Cycle Calvin Cycle FNR 2NADPH
RBISCOetc RBISCO etc CO2
2NADP
sugar
suaar
Supplemented Artificially Supplemented Enzyme Redox Enzyme substrates
useful products
SP excitation products useful products
O2
Scheme 4. Mechanism of photoinhibition under high-light condition and concept of semi-artificial photosynthetic microchip using SP enhanced excitation, in which the excess electrons are effectively transferred to the artificially supplemented redox enzyme giving useful products.
Is photosynthesis really an optimised photoinduced molecular transformation system? The answer is NO. It is a fact that the efficiency of photoinduced charge separation system is extremely high. However, the reaction rate of the following enzymatic reaction system, called the Calvin-Benson cycle, is significantly slow and spoils the performance of the whole system. Maybe this is due to the fact that plants used to live under a low-fluence environment in the early stage of their evolution. Under high light levels (for example, usual
166 166
A. Fujii and A. Ishida
daylight), the photoinduced charge separation system of the plant supplies more electrons than the following enzymatic reaction system requires. The excess electrons generated as a result are captured by oxygen, generating reactive oxygen species such as superoxide, which cause serious damage to the plant body (for example, by browning of the leaves) [38]. Plants have several protection systems to prevent such photoinhibition, one of which is an electron consumption cycle called photorespiration. Photorespiration is a reaction system that prevents damage to the plant; to do this it prevents the occurrence of reactive oxygen species by consuming excess electrons with the energy of light. This is a non-productive system that wastes an amount of energy that cannot be ignored as a proportion of the energy acquired from light. However, as long as a stable and perpetual source of energy from the sun is used, there is little need to improve the energy efficiency. When plants had to go out into high light levels, they probably added the photorespiration system as protection against damage, instead of fundamentally re-creating the enzymatic reaction cycle of the photosynthetic system. Therefore, we can assume that we should use the tactic of keeping the photoinduced charge separation system and replacing the enzymatic reaction system with one with higher performance if we want to use SP excitation to construct a system that improves on the system of photosynthesis. Therefore, as a next step, we planned a method of realising this tactic [39]. 5.2. Concept of a SP excited semi-artificial photosynthetic reaction system Enzyme electrodes can be created on metal electrodes by immobilising redox enzymes, as occurs with enzymes in the Calvin-Benson cycle. The electrode is applied as a sensor for quantitative analysis of molecules such as glucose, as it gives the current signal accompanying the enzymatic reaction [40]. Contrary to this, the enzyme can cause a synthetic or degradation reaction if a voltage sufficient to induce the reaction is supplied. Therefore, useful compounds should be able to be produced by driving the redox enzyme by the electrons generated by photoexcitation if the photoinduced charge separation system of the photosynthetic system, an appropriate redox enzyme, and a mediator that transports electrons can be immobilized together on the surface of the gold. The thylakoid membrane is several micrometers long and several tens of nanometers thick. As described above, the SP electromagnetic field is suitable for exciting molecules positioned at a distance of several tens of nanometers from the surface, and we can expect effective excitation of the thylakoid by the enhanced electromagnetic field of SP. 5.2.1. Immobilisation ofthylakoids on gold and the SP excited Hill "s reaction Thylakoid that contained a small amount of intact chloroplast was isolated from spinach, then purified and immobilized on the surface of a gold film by using several kinds of adhesive layer, such as polylysine. The transmission
Surface plasmon enhanced excitation of photofunctional molecules in nanospace towards molecular plasmonics plasmonics
167 167
absorption spectrum of the modified gold film showed characteristic absorption bands in the chlorophyll, clearly showing that the thylakoid was immobilized on the surface. PS II, among photosynthetic systems, oxidises water and generates electrons. PS I continues to photoexcite electrons successively and acquires a strong reduction potential electrochemieally. In other words, electron generation occurs from both PS II and PS I by photoexeitation of the thylakoid. Therefore, if an electron-accepting reagent is added, it captures electrons generated by PS II and PS I, thus reducing the amount, and the activity of the photosynthetic system can be assessed. This is the so-called "Hill's reaction" [41]. The electron-accepting reagent used this purpose was a blue compound called DCIP (dichlorophenol indophenol). It captures two electrons and gives a transparent dihydro derivative after reduction, so the electrons generated can be quantified by measuring the weakening of the colour of the solution. A modified gold film was attached to a glass prism with index matching oil, and consumption of DCIP was observed by the changes in absorbance at a wavelength of 600 nm during SP excitation by ATR illumination. As soon as SP excitation began the DCIP started to decrease; the thylakoid was excited by the SP electromagnetic field, clearly showing that electrons were generated by the water oxidation [39]. The most effective adhesive layer was polylysine, which was chemically immobilized on the surface of the gold and presumably immobilized the thylakoid by electrostatic adsorption. On the contrary, although a gold film on which cystamine was immobilized adsorbed the thylakoid strongly and gave a high immobilisation density of thylakoid, its efficiency in generating electrons under these conditions was clearly less than that of polylysine. When the immobilisation density becomes excessive, there is a fear of inviting damage to the thylakoid by superoxide generated from the released electrons, because the electron-capturing efficiency declines owing to blockage of the diffusion of DCIP. When excitation was continued even after the DCIP has completely consumed, the photosynthetic system was damaged and deactivated irreversibly. However, if excess amounts of DCIP were present, then electrons continued to be generated for over 24 hours. This result clearly shows that a chemical reaction system that oxidises water by SP excitation of immobilized thylakoid possesses a sufficiently practical value as an electron source in enzymatic reaction systems.
168 168
A. A. Fujii and A. Ishida Cl
hν
H2O
- 2 e-
PS-II
PS-I
+ 2 e-
Cl
H2O
1/2 O2
N
O
ONa O Na
DCIP λmax=600 nm
thylakoids
DCIP2H DCIP2H colorless
λex= 632.8nm 12.5 mW
X= 600 nm λ= LED
θ=73
mt, DCIP [/DCIP~|
Multichannel Spectrometer
7
immobilised thylakoids on gold film
Scheme 5. Mechanism of Hill's reaction using DCIP as an electron acceptor and measurement setup of electron generation from immobilized thylakoids on gold surface upon SP excitation
[DCIP] arb. unit
60
40
20 0
0
30
60 90 time// min time
120
Fig. 5. Reduction of 0.1 mmol dm"3 DCIP by electron generation from immobilized thylakoids on gold surface upon SP excitation.
5.2.2. Driving a redox enzyme by SP excitation of thylakoid The fact that immobilized thylakoid generated electrons efficiently by SP excitation suggests that the electrons obtained as a result can drive a redox enzyme if it is transported to the immobilized redox enzyme together with the thylakoid on the surface. The enzyme used in this research was FNR, which reduces NADP to NADPH by assistant of Fd, which is an electron mediator. Chloroplasts contain this enzyme, but isolated thylakoid loses the majority of the enzyme during the isolation and purification processes, and the enzyme's reactivity is very low in a system where only thylakoid is immobilized. Therefore, FNR and Fd must be immobilized together with thylakoid to induce effective reaction of FNR by SP. A film that was obtained as a result of immobilizing a mixture of thylakoid, FNR, and Fd on polylysine that itself had
Surface plasmon plasmon enhanced excitation of photofunctional photofunctional molecules in nanospace towards molecular plasmonics
169
been immobilized on the surface of gold was dipped in a buffer solution that contained NADP, and consumption of the NADP was observed by photoabsorption measurement at 340 nm during SP excitation. NADP was consumed exponentially over several hours, beginning straight after excitation [39]. The result clearly shows that SP excitation induced an enzymatic reaction that reduced NADP. H 2O
hν
e-
- 2 e- PS-II
NADP
?imax =340 nm nm λ max=340
Fn FNR
PS-I
NADPH
1/2 O2
λex= 632.8nm 12.5 mW
θ =73
λ= 340 nm Xe-lamp + monochlometer
NADP
Multichannel Spectrometer
Thylakoids + Fn + FNR on poly-lysine coated Au film
Scheme 6. Immobilization of thylakoids with Fa and FNR and measurement setup of NADP reduction.
[NADP] / arb. unit
60 40 20 0 0
5
10 time / min time/
15
Fig. 6. Reduction of 0.4 mmoldm"3 NADP by SP excitation of immobilized thylakoids, ferredoxin, and ferredoxin NADP reductase on gold surface.
The reason why chloroplasts that retain their activity, or chlorophyll in the thylakoid, show only very weak fluorescence is that the energy is consumed for chemical reactions. Usually faint fluorescence in chloroplasts is studied using a device called a PAM [42]. However, there is no PAM commercially available that can be applied to such a thin film, and we are currently developing one. Because of this, unfortunately, fluorescence by SP excitation of chloroplasts, which is essential for solving the dynamics of the reaction, has not yet been observed. Although the yield of NADP reduction reaction that was observed per
170 170
A. Fujii and A. Ishida
one chlorophyll molecule depended strongly on the properties of the adhesive layer, the properties of the adhesive layer were not necessarily directly related to the properties of the excited state of the thylakoid, as the yield is determined by various factors. The fact that an adhesive layer that possessed an amino group gave a high reaction efficiency suggests that the density of the film (as determined by coulombie interaction of the thylakoid membrane and the adhesive layer, and strongly related to the transmission and diffusion of the solute) on which the thylakoid was immobilized had a bigger impact on the reaction efficiency than did the properties of the excited state. As described above, we demonstrated that fluorescence quenching, which is a disadvantage of SP excitation, can be avoided effectively by using a thick film, and its application can be extended to photochemical reactions as well as to the fluorescence analyses reported so far. It is probably possible to apply it to not only systems that use chloroplasts but also photocatalyst and photochemical data storage, as long as the molecules are capable of photoabsorption in visible light. 6. LOCALISATION OF THE SP FIELD AND ITS APPLICATION The excitation of molecules by using SP electromagnetic fields can be widely applied to such procedures as analyses, chemical reactions, and photoenergy conversion; however it has, at the same time, a disadvantage of quenching in the excited states. One effective method of avoiding this is to secure a distance from the gold in a vertical direction by using a thick buffer layer, as described in sections 4 and 5. In considering the fact that quenching can be prevented by securing distance from the gold, we can think of another tactic: securing horizontal distance by creating a nanosized microstructure in the gold film, in other words a space where no gold exists, such as a hole or a slit. The topic that has recently drawn the most attention in plasmon research is the phenomenon that nanoholes formed in silver or gold film - in other words, apertures of a diameter less than the wavelength of light - endow the property of extra-ordinal light transmission. On the basis of theoretical [13] and experimental research [10-12], several research groups have suggested that localisation of the plasmon field in such apertures is involved in this extra-ordinal light transmission. Interestingly, it is reportedly possible to select the wavelength as well the intensity of light that is transmitted by controlling the aperture size and alignment of the nanohole. These results suggest the possibility of localising light of a required wavelength inside the aperture by designing the aperture precisely, thus enabling its use in various applications. A nanosized aperture can be used as a microvessel for analyses and reactions that use light. One thing we must be aware of here is the importance of microapertures in gold, created on a glass substrate. The structure obtained here is a gold well with an aperture the size of the wavelength of light and possessing a glass-bottomed
Surface plasmon plasmon enhanced excitation of photofunctional photofunctional molecules in nanospace towards molecular plasmonics
171 171
surface. We named this structure a "gold nanowelF. The bottom surface of this gold nanowell is glass, so various kinds of molecules can be immobilized on the surface by chemical modification. If a gold nanowell with a bottom surface on which photoresponsive molecules are immobilized is SP-excited, the plasmon electromagnetic field that is propagated will be localised in the nanowell and the enhanced electromagnetic field obtained as a result is sure to excite the molecules. The important thing here is that the distance of propagation of the plasmon into free space is much longer than the distance where quenching by energy transfer to gold is sure to occur. In other words, many immobilized molecules are too far away from the gold surface of the nanowell wall to be quenched, so that quenching can be prevented at the same time as efficient excitation by the plasmon electromagnetic field is achieved. The development of such new methods of high-efficiency excitation of molecules in microspots on glass substrates is important, not only for the development of microarrays of DNA and protein [43] or display material, but also in the microfluid system [44] that has recently seen significant developments. h ν hv
and localisation scattering and of propagating SP SP field
excitation by calisation of of plasmon be photoillumination direct photoillumination interaction between the the neighboring nanowells
metal
metal glass substrate
expected functions
future applications
light antenna fluorescence enhancement addressing nanovessel nanoarray μ -TAS H-TAS fibre sensor light emitting device etc.
Scheme 7. Concept of nanowell
6.1. Fabrication of the gold nanowells and their optical properties A nanowell that can localise the SP electromagnetic field efficiently on the basis of precise design can be fabricated if a highly developed nanofabrication system is used. However, the design of the field distribution inside the nanowell is still difficult, and the devices required for the nanofabrication process are extremely expensive. We fabricated nanowells using an easy method, called the
172 172
A. A. Fujii and A. Ishida
"projection method", which uses latex beads [45]. To explain this method simply, aqueous suspension of latex beads that have the same diameter as the aperture required are scattered on a glass substrate that has a substituent for immobilizing molecules by chemical modification on its surface, and then gold of the required thickness is evaporated on top of it. After the evaporation, the latex beads covered by the gold are removed by ultrasonication in water, revealing the gold nanowell. In this method, of course, the configuration becomes random, so that interactions between the nanowells cannot be used and therefore the field intensity of the plasmon may be fairly weak, but it is satisfactory for assessing the function of the well as a cuvette. We studied the properties of SP scattering and light transmission of a gold nanowell created by the projection method, by affixing a coverslip that had a nanowell on its surface to a glass prism with index matching oil [35]. We illuminated monochromatic light while scanning the wavelength in Kretschmann's configuration of incident light at 55 degrees and then observing the intensity of the light of the same wavelength emitted from the gold surface that contained the nanowell. We expected that this measurement would explain the relationship between the scattering efficiency and the aperture size of the nanowell, because if the nanowell scattered an SP propagated at the surface of the gold and generated by ATR illumination, then the light scattering by the surface would be observed.
scattering intensity
BK-7 prism _ _ _ \ 5! 55 p-polarised monochromatic light
nanowells gold nanowells scattered light
monochromator monochromator
300 200 100 0 400
6 00 500 600 wavelength // nm wavelength nm
700
800
Fig. 7. SP scattering spectra of flat gold film and films with 500 and 600 nm nanowells by ATR illumination; inset shows optical configuration of the measurement.
In the scattering spectrum (Fig. 7), several peaks were observed over the whole visible region in common with all gold films, and they gave a characteristic fine structure to the spectrum. The spaces between peaks were 50 nm, which was almost consistent with the thickness of the gold film, suggesting
Surface plasmon enhanced excitation of photofunctional molecules in nanospace towards molecular plasmonics
173 173
that this fine structure resulted from multireflection between the two surfaces that sandwiched the gold film. Smooth gold film, in contrast, showed only a very weak scattering spectrum that possessed a fine structure. In contrast, a gold film that possessed a nanowell gave a far stronger scattering spectrum than did smooth gold film over the whole visual region. A 500 nm well showed an especially intense peak at around 550 nm and individual peaks at around 650 nm and 700 nm. This suggests that these intense peaks resulted from scattering of SPs by the nanowell, as the peaks did not exist with the smooth gold film [35]. To study the interaction between light and nanowells in the near-field, we scanned the surface of a gold film that had a nanowell while grazing laser light on an AFM chip made of silicon nitride; we then observed the transmission of light scattered from the very end of the chip. The very end of the chip was pyramidal-shaped, 2 |j,m on all sides and 1.4 |xm high, and the light source became a mixture of far-field light and near-field light under grazing illumination. Although far-field light overlaps near-field light as a very intense background, the effect of far-field light can be removed to a certain extent by detecting the focus of the optical microscope with an avalanche photodiode after matching it with the bottom surface of the well. A two-dimensional (2D) plot of the intensity of transmitted light can thus be expected to contain information related to interactions between the near-field light and the nanowell. Topographic Image 1000 nm
contact mode AFM 85nm 85 nm
532 nm 1μ VJ 3+ Nd3+ -YAG Laser -YAG
Nikon Eclipse 300
nanowells
t
500 nm
0.0 nm
x60 × 60 NA1.2
0 nm 0 nm
position controllable pinhole turret
Perkin Elmer SPCM module AQR-14 APD module
500 nm
1000 nm nrr 1000
Near-Field Image
BP520 1000 nm
•10mV -10 mV
t
500 nm
-14 mV
0 nm 0 nm
500 nm
1000 nm
Fig. 8. Topographic and near field image upon grazing illumination of a 600 nm gold nanowell on a glass substrate and the measurement setup; dark portion in the near field image represents intense light scattering.
174 174
A. Fujii and A. Ishida
The 2D plot obtained was an image that had strong contrast at the rim of the nanowell (Fig. 8). This can be understood as follows. Scattered light from the very end of the chip does not transmit during scanning of the flat surface of the gold film, as it is shielded by the gold film. Also, the light intensity is low when the very end of the chip comes to the bottom surface of the gold film, as the light does not reach to the very end of the chip. In contrast, we consider that strong transmitted light was observed on the rim of the nanowell because the light scattered at the very end of the chip was scattered by the edge of the rim of the nanowell and reached the back side efficiently. There is a possibility that the plasmon is involved in the scattering of near-field light, as pointed out by some groups. Plasmons and near-field light can be transformed mutually, so the fact that the light was scattered effectively at the edge suggests that the SP electromagnetic field that was propagated was scattered at the rim and transformed to near-field light. To explain the interaction between the SP electromagnetic field and the nanowell we would need to conduct an experiment whereby we would suck up near-field light with the c-mode probe of a near-field scanning optical microscope. However, we have not yet been able to do this because of experimental difficulties. From the results described above, we determined that the gold nanowell efficiently scatters an SP that is formed when light of a slightly longer wavelength than the aperture size is launched. Because the wavelength of the SP coincides with that of the evanescent field, this result suggests that a gold nanowell efficiently scatters incident light that roughly corresponds with its aperture and induces a large amount of light transmission. In other words, it seems that near-field light of a slightly longer wavelength than the aperture size is localised strongly inside the gold nanowell. As a result, we can expect that there will be applications of the process whereby the photoresponsive molecules immobilized or packed inside the well are effectively excited by this concentrated and enhanced near-field light. 6.2. Fluorescence analysis inside the nanowell Fluorescence analyses offer superior characteristics, such as high sensitivity (enabling the detection of single molecules), a wide dynamic range, linearity that gives good quantification, and a simple protocol. Especially in microarrays that enable high-throughput analysis, the use of fluorescence detection is essential, and it should be useful for developing ji-TAS, too [44]. The microplates currently in use have wells as big as 1 to 5 mm and cannot provide the throughput that is required in post-genomic research. Also, the dot size of these microarrays is more than 100 Jim [43]. It may be possible to improve the throughput by making smaller wells and dots smaller and closer to the wavelength of light, but the fluorescent signal will be buried in the background noise if there is no technology that improves the efficiency of
Surface plasmon plasmon enhanced excitation of photofunctional photofunctional molecules in nanospace towards molecular plasmonics plasmonics
175 175
photoexcitation and fluorescence detection in nanospace. Therefore, fluorescence analysis that uses nanowells is expected to be an effective new solution. 6.2.1. Immobilisation of fluorescent molecules in nanowells We aminated the surface of the glass with a silane-coupling reagent before scattering latex to immobilize the molecules in the glass part of the bottom surface of the nanowell [35]. An amino group can easily be chemically modified at ambient temperature by mixing it with a solution of a compound that possesses a succinimidyl group, and also with a compound that possesses an amino group, by using a bridging reagent such as glutaraldehyde. This amino group is effective as an adhesive layer against gold film, too. The first thing we did was to observe the fluorescence from fluorescent dye molecules that were immobilized on the bottom surface of the well by modifying the well with a fluorescent dye that bound specifically with the amino group. This observation was performed to assess the efficiency of immobilisation of the fluorescent molecules against the amino group that remained after the latex particles had been removed by ultrasonication in water. As a fluorescent dye, we used TR-SE (Texas Red succinimidyl ester mixture of the isomers), a derivative of Texas Red that is planned to be used in immunoassay and DNA assay. This dye has maxima of photoabsorption and fluorescence in the red region, where plasmon enhancement can be effectively used, and also it is not prone to photobleaching, which often becomes a serious problem in fluorescent spectroscopy [46]. TR-SE possesses a succinimidyl group as a side chain that is specific to binding with an amino group. We dipped a coverslip that possessed a gold nanowell with an aperture size of 600 nm in TR-SE solution, making the TR-SE react with the amino group on the bottom surface. We used index matching oil to bond two right-angle prisms on a coverslip that possessed a modified nanowell to form an optical waveguide, and then applied SP excitation with p-polarised laser light of 532 nm. We then obtained a fluorescence microscopic image of the nanowells, along with a clear fluorescence spectrum (Fig. 9). The result clearly showed that a sufficient quantity of amino groups was present to immobilize the molecules on the bottom surface of the nanowell. To assess the number of TR molecules immobilized, an aminated coverslip was dipped in TR-SE solution and the variations in the absorption spectrum and fluorescence spectrum were observed with time. The absorption intensity of the TR at its maximum absorption wavelength increased as the dipping time increased and then reached a plateau after 10 minutes. Contrary to this, the fluorescence intensity peaked after 5 minutes, decreased dramatically, and then stabilised at a value of about one-third of the maximum value. Thus TR-SE is immobilized after binding with the amino group and the number increases
176 176
A. Fujii and A. Ishida
gradually and reaches saturation. We consider that the fluorescence intensity decreased because energy migration started to happen between adjoining TRs when the number of immobilized TR molecules became excessive. If we assume that immobilisation of molecules on the surface of the coverslip and the bottom surface of the nanowell progressed at similar speeds, then on the basis of the results of measurement of the absorption spectrum of the coverslip we can say that 110 000 TR molecules would have been immobilized on the bottom surface of a 600 nm well during 5 min modification.
Fig. 9. Fluorescence image of TR immobilized 600 nm gold nanowells upon ATR illumination of 532 nm p-polar light
6.2.2. Immobilisation of antibody andfluoroimmunoassay in nanowells As the immobilisation of fluorescent molecules was successful, we next tried to immobilize antibody on the bottom surface for fluoroimmunoassay [47]. We treated gold nanowells of 600-nm aperture with glutalaldehyde for immobilisation of the aldehyde group on the ammo terminated bottom surface first, and we then dipped the nanowell in a buffer solution of anti-IgG FITC conjugate. Fluorescence microscope observation revealed fluorescence of the FITC from the nanowell, clearly showing that FITC-anti-IgG was successfully immobilized on the bottom surface of the nanowell. After that, we dipped the nanowell containing the immobilized antibody in casein solution, thus performing blocking to prevent adsorption of antibody by the non-specific interactions that can cause background signals when conducting fluoroimmunoassay. The fluoroimmunoassay sample used was IgG Texas Red conjugate (TRIgG) and the control was BSA Texas Red conjugate (TR-BSA). Two pairs of nanowells were SP-excited, but their phosphoric acid buffer solutions were individually dipped. In the case of TR-IgG, the fluorescence intensity increased and then plateaued in several tens of seconds (Fig. 10). In the case of TR-BSA no increase in fluorescence intensity was observed. This obvious difference clearly shows successful immunoassay inside the nanowell. In other words, TRIgG in solution was trapped in the FITC-anti-IgG immobilized on the bottom
Surface plasmon plasmon enhanced excitation of photofunctional photofunctional molecules in nanospace towards molecular plasmonics
177
surface of the nanowell and the TR molecules entered the localised SP field, which caused excitation, and fluorescence emitted, whereas TR-BSA, which had no specific interaction against anti-IgG, was not trapped into the SP field.
NH22-terminated NH 600 nm nanowells
glutaraldehyde
anti-Human IgG anti-Human IgG
in water 5% in 4°C, overnight 4˚C,
ng / ml 100 μg in phosphate buffer buffer 4˚C, 4 C, 16 16 hh
Red conjugated conjugated Human IgG Texas Red lgG ^.g / ml ml in in phosphate buffer 300 μg
Red conjugated conjugated BSA Texas Red BSA ng / ml ml in in phosphate buffer 300 μg
fl u or e sc e nc e i n t en s i ty
Scheme 8 Protocol of fluorescence immunoassay in nanowells
0.15
TR-IgG TR-IgG
0.10 0.05 TR-BSA 0.00 -0.05
40
60 60 time / s time/s
80 80
100 100
Fig. 10. Fluorescence immunoassay in 600 nm gold nanowells
One thing we have to remember here is that although the sample solutions of both TR-IgG and TR-BSA showed very strong fluorescence by far field photoirradiation, the background fluorescence of the solution, which could have seriously affected the measurements, could hardly be observed by SP-excitation. If background fluorescence shows up strongly, then one would expect the fluorescence intensity to shoot up straight after dipping. The slow increase after dipping of the TR-IgG suggests that it was a result of immune reaction, not background fluorescence. The measurement system used this time was called
178 178
A. Fujii and A. Ishida
quasi-confocal optics, and the background fluorescence of the solution was effectively removed by making the light that led to the photodetector come into focus on the bottom surface of the nanowell. In addition to this, the location of excitation of the localised SP was limited to the inside of the nanowell. These two effects enabled an analysis that gave a high signal-to-noise ratio. Also, the fluorescence signal represented the fluorescence from only 20 nanowells and was separated from the field of vision of the microscope by a pinhole. Unfortunately, a clear signal could not be obtained from only one nanowell, but improvement of the optical system may make this possible. Generally, analysis in microspace gives a narrow dynamic range, and this often causes serious problems in practical applications. By conducting observations using these nanowells at a series of antibody concentrations and verifying the linearity, we obtained linearity between the fluorescence intensity and concentration in the concentration range of 200 ng/mL to 60 jig/mL. This range suggests that fluoroimmunoassay in nanowells has sufficient practicality. However, expansion to even more dilute concentrations is particularly desirable [48], so improvements in the microfluid and optical systems are probably essential. Also, the reaction presently used to immobilize the antibody was a method that could not control orientation, which may have caused only a part of the immobilized antibody to bind with the antigen. In fact, a quantitative assessment by analysis of the absorption spectrum suggested that binding efficiency of the antigen by the immobilized antibody was only 20 % when the surface of the coverslip was dipped in excess antigen. Also, we suspect that close distances between antibodies when antibody is excessively immobilized may inhibit the approach to antibody by diffusion of antigen or decrease the fluorescence intensity by energy migration because of the presence of many antigens in a small region. The use of orientation control and diluting material in immobilisation of the antibody is essential for efficient detection of the specific binding interaction between antigen and antibody in nanospace. 6.2.3. Immobilisation ofoligonucleotide andDNA analysis in nanowells Detection of hybridisation of DNA is one of the most important analyses in genetic engineering [49]. The practical application of the microarray was a revolution in the improvement of research efficiency. However, the packaging density of the microarray is not satisfactory, as described above, and we await further improvements. Therefore, we tried fluorescence DNA analysis in nanowells [47]. Three kinds of oligonucleotides of 15 base pairs were used in this experiment. The probe oligonucleotide was an FITC conjugate, and its fluorescence made it possible to check immobilisation on the bottom surface of the nanowell. The target oligonucleotide was a TR conjugate that possessed a base sequence that was complementary to that of the probe oligonucleotide. The
Surface plasmon plasmon enhanced excitation of photofunctional photofunctional molecules in nanospace towards molecular plasmonics
179
control oligonucleotide was a TR conjugate that was a target oligonucleotide whose base sequence had been scrambled. We immobilized the aldehyde group by treatment of the amino group on the bottom surface of a gold 600 tun nanowells with glutalaldehyde first, and we then dipped it in a buffer solution of probe oligonucleotide. Fluorescence microscope observation revealed fluorescence of FITC from the nanowell, clearly showing that the probe oligonucleotide was immobilized on the bottom surface of the nanowell. glutaraldehyde
5% in water
NH22 -terminated -terminated ^ I / o i n w a ! e L 4˚C, overnight 600 nm nanowells 4 C ' o v e m l g h t
=
15mer5'-NH 15 mer 5’-NH22 oligonucleotide probe oligonucleotide
|xg / ml 10 µg ml in phosphate buffer buffer 4˚C, 4°C, 16 16 hh
5'-Texas Red 15 mer 5’-Texas Red complementary target oligo in TE buffer, rt 10 nM in
probe oligo nanowells immobilised nanowells
5'-Texas Red 15 mer 5’-Texas Red non-complementary control oligo
in TE buffer, buffer, rt 10 nM in Scheme 9. Protocol of fluorescence DNA assay in nanowells
fl u or e s c e n c e i n te n s it y
0.15
ta rg e t ooligonucleotide lig o n u c le o tid e target 0.10
0.05
0.00
ccontrol o n tro l ooligonucleotide lig o n u c le o tid e
-0.05 0
40
80 time / s
Fig. 11. Fluorescence DNA assay in 600 nm gold nanowells
120
180 180
A. Fujii and A. Ishida
For fluorescence DNA analysis the same optical system as used for immunoassay was used. Two pairs of nanowells were SP-excited, but their hybridisation buffer solutions were dipped individually. In the case of the target oligonucleotide, the fluorescence intensity then increased and plateaued in several tens of seconds (Fig. 11). On the other hand, no increase of fluorescence intensity was observed in the case of the control oligonucleotide. Although there were relatively many noises, this result indicates that Hie probe oligonucleotides immobilized on the bottom surface of the nanowell and the target oligonucleotides were hybridised in the nanowell. The TR molecules were excited by the localised SP field and showed fluorescence, whereas the contrast oligonucleotide could not hybridise and its TR molecules did not enter the SP field. Like immunoassay, DNA assay was successful. An oligonucleotide with minimal number of base sequences was used this experiment in an effort to recognise the patterns characteristic of the genes related to several diseases. It is probably necessary to verify the effectiveness of nanowell analysis for oligonucleotides that have longer chains. As the plasmon electromagnetic field propagates over several hundred nanometers, it seems impossible for a fluorescence dye that is conjugated to DNA to go outside the effective area of the plasmon electromagnetic field because the DNA would be too long, even if there were more than 300 base pairs. Indeed, the immobilisation reaction of DNA may be a serious problem in analysing long DNAs. It is possible that DNA that is too long can adhere strongly to the surface of the substrate due to electrostatic interaction, as DNA has a negative electric charge on its surface. To prevent this, countermeasures such as decreasing the number of active sites and positive charged sites on the bottom surface of the nanowell by using an appropriate diluting material will be necessary. The use of diluting material is necessary to secure spaces between the probe DNAs and to effectively induce hybridisation with the target DNA. 6.3. Future development of the nanowell We demonstrated the effectiveness of fluorescence analysis in nanowells, as described above. The analysis of individual samples in individual nanowells in arrays that are optically precisely designed and fabricated by a nanofabrication process and then regularly arranged is a key technological goal. Although there are many technical problems that need to be solved in the fields of optics, chemistry, and mierofluidics, the significant progress in the development of inkjet printers leads us to expect that it will be possible to perform separate injection of sample solutions on a nanoscale, a development that will be essential to realisation of the nanoarray. In considering the limits of the detection system, it is probably more realistic to use a bundle of several nanowells than to use a single nanowell. In that case, the advantages of being
Surface plasmon enhanced excitation of photofunctional molecules in nanospace towards molecular plasmonics plasmonics
181 181
able to efficiently conduct enhancement and localisation of the SP electromagnetic field are achieved at the same time by the regular arrangement of several nanowells. There is no need to use round wells, either. The precise design of well shape on the basis of simulations of field distribution is sure to improve performance. In particular, nanoslits are able to self-aspirate sample solutions by capillary force and will therefore be useful when integrated into ]iTAS. It is certain that the application of nanowell technology will improve the density of current microarrays by several thousands to several tens of thousands of times and realise development of bioanalysis microchips based on nanoarray. 7. CONCLUSION We have described the excitation of photoresponsive molecules by SP, from a basic demonstration to the application in practical fluorescence analysis. Our research is pioneering work in the application of SP electromagnetic fields. However, there are still many parts of which the details are unclear. One of them is the relationship between the fluorescent excited state and the SP. The plasmon must be strongly involved in emission processes such as the excitation process that we have researched in detail. In the case of excitation by the ATR method, we expect that emission of fluorescence will occur to the side of the prism at a particular angle. Unfortunately, these observations are difficult and have not yet been done precisely, because the emissions to be measured are mixed with the excitation light even if a notch filter is used. Also, the fact that the maxima of the fluorescence excitation spectrum and the photocurrent action spectrum of the porphyrin SAM show a long wavelength shift from those of the absorption spectrum suggests that molecules receive the SP-enhanced field directly; however, to check whether similar effects occur inside the nanowell, the excitation spectrum of the molecules immobilized inside the nanowell must be analysed precisely. Emissions from nanowells are not easy to measure, as they are far weaker than those from flat surfaces. Also, the liquid that is encapsulated in a microspace such as a nanowell probably has very different properties to the bulk liquid, and the possibility of this affecting the accuracy of the analysis and the efficiency of the reaction is strong. We intend to continue our research on these remaining problems. Although there is a need to conduct such important applied research as fluorescence analysis in nanowells promptly, it is also essential that we accumulate basic data on things such as the interactions between the molecules and electromagnetic fields, interactions among molecules, and interactions between the molecules and substrate, to enable the SP to play a role in the achievement of a breakthrough in the application of nanospace.
182 182
A. Fujii and A. Ishida
ACKNOWLEDGEMENT We thank OUT co-workers (Haruka Takedatsu, Ayaka Shinjo, and Tomonori Shibata) for their important contributions to the works reported herein. This work is supported by PRESTO project of Japan Science and Technology Co., Nakatani Foundation, Kansai Research Foundation, and the Grant-in Aid for Scientific Research {KAKENHI) from Ministry of Education, Culture, Science and Technology of Japan (Priority Area "Molecular Nano Dynamics" (No. 17034054) and Basic Research (C) (No. 16510111 and 14650814)). REFERENCES [1] M. A. Reed and T. Lee, Molecular Nanolectronics, American Scientific Publishers, 2003; idem. Proc. IEEE, 37 (1999) 652. [2] M. Ohtsu, K. Kobayashi, Optical Near Fields, Springer-Verlag, Berlin, 2004. [3] H. Rather, Surface Plasmons on Smooth and Rough Surfaces and on Gratings, Springer Tracts in Modern Physics, Springer-Verlag, Berlin, 111 (1988) 4. [4] L. G. Fagerstam, A. Frostell, R. Karlsson, M. Kullman, A. Larsson, M. Malmqvist, and H. Butt, J. Mol. Recognition, 3 (1990) 208 [5] Y. Kuwamura, Y. Yokota, M. Fukui, and O, Tada, J. Phys. Soc. Jpn., 51 (1982) 2962. [6] R. R. Chance, A. Prock, and R. Silbey, Adv. Chem. Phys., 37 (1978) 1. [7] Y. Dimitriev and E. Kashchieva, J. Mater. Sci., 10 (1975) 1419. [8] U. Kreibig and M. Vollmer, Optical Properties of Metal Clusters, Springer, Berlin, 1995. [9] F. Kim, J-H. Song, and P. Yang, J. Am. Chem. Soc, 124 (2002) 14316; Y. Niidome, K. Nishiok, H. Kawasaki, and S. Yamada, Chem. Comm. (2003) 2376; M. Tsuji, M. Hashimoto, Y. Nishizawa, and T. Tsuji, Mater. Lett., 58 (2004) 2326; M. Tsuji, Y. Mshizawa, M. Hashimoto, and T. Tsuji, Chem. Lett., 33 (2004) 370; J. Perez-Juste, L. M. Liz-Marzan, S. Carnie, D. Y. C. Chan, and P. Mulvaney, Adv. Funct. Mater. 14 (2004) 571; J. Perez-Juste, I. Pastoriza-Santos, L. M. Liz-Marz&i, and P. Mulvaney, Coord. Chem. Rev. (2005) in press. [10]R. Gordon, A. G. Brolo, A. McKinnon, A. Rajora, B. Leathern, and K. L. Kavanagh, Phys. Rev. Lett., 92 (2004) 1037401; A. G. Brolo, R. Gordon, B. Leathern, and K. L. Kavanagh, Langmuir, 20 (2004) 4813; J. Lindberg, K. Lindfors, T. Setatala, M. Kaivola, and T. Friberg, Opt. Exp., 12 (2004) 623; A. Moreau, G. Granet, F. I. Baida, and D. Van Labeke, Opt. Exp., 11 (2004) 1131; W. L. Barnes, A. Dereux, and T. W. Ebbesen, Nature, 424 (2003) 824; B. Dragnea, J. M. Szarko, S. Kowarik, T. Weimann, J. Feldmann, and S. R. Leone, Nanolett. 3 (2003) 3; E. Altewischer, M. P. van Exter, and J. P. Woerdman, Nature, 418 (2002) 304; T. W. Ebbesen, H.J. Lezec, H. F. Ghaemi, T. Thio, and P. A. Wolff., Nature, 391 (1998) 667. [11]C. Sonnichsen, A. C. Duch, G. Steininger, M. Koch, G. von Plessen, and J. Feldmann., App. Phys. Lett., 76 (2000) 140. [12]M. Xiao andN. Rakov, J. Phys., Condens. Matter, 15 (2003) L133. [13]K. Tanaka, M. Yan, and M. Tanaka, Opt. Rev. 8 (2001) 43; idem, ibid, 9 (2002) 213; idem, Appl. Phys. Lett., 82 (2003) 1158. [14]I.Pockland, J. D. Swalen, R. Santo, A. Brillante, and M. R. Philpott, J. Chem. Phys., 69 (1978)4001.
Surface excitation of photofunctional molecules Surfaceplasmon plasmonenhanced enhanced excitation of photofunctional molecules in nanospace towards molecular plasmonics in nanospace towards molecularplasmonics
183 183
[15]H. Kano and S. Kawata, Opt. Lett, 21 (1996) 1848. [16]S. Hayashi, K. Kozara, and K. Yamamoto, Solid State Commun. 79 (1991) 763. [17]J. Zak, H. Yuan, M. Ho, L. K. Woo, and M. D. Porter, Langmuir, 9 (1993) 2772; E. U. Thoden van Velzen, J. F. J. Engbersen, P. J. de Lang, J. W. G. Many, and D. N. Reinhoudt, J. Am. Chem. Soc, 117 (1995) 6853; K. Shimazu, M. Takechi, H. Fujii, M. Suzuki, H. Saiki, T. Yoshimura, and K. Uosaki, TMn Solid Films, 273 (1996) 250. [18]A. Ishida, Y. Sakata, and T. Majima, Chem. Comm. (1998) 57. [19]A. Ishida, Y. Sakata, and T. Majima, Chem. Lett., (1998) 267. [20] A. Ishida and T. Majima, Chem. Comm., (1998) 1299. [21]A. Ishida and T. Majima, Nanotechnology, 10 (1999) 308. [22]A. Ishida and T. Majima, Analyst, (2000) 535. [23]A. Ishida and T. Majima, Chem. Phys. Lett., 322 (2000) 242. [24]K. Tawa, D. Yao, and W. Knoll, Biosens. Bioelectr. (2005) in press; K. Tawa and W. Knoll, Nucleic Acids Res. 32 (2004) 2372; F. Yu, D. Yao, and W. Knoll, Anal. Chem. 75 (2003) 2610; D. Kambhampati, P. E. Nielsen, and W. Knoll, Biosens. & Bioelectron. 16 (2001) 1109; T. Liebermann, W. Knoll, P. Sluka, and R. Herrmann, Colloids Surf. A 169 (2000) 337; T. Liebermann and W. Knoll, Colloids Surf. A, 171 (2000) 115. [25]A. Ulman, an Introduction to Ultrathin Organic Films From Langmuir-Blodgett to SelfAssembly; Academic Press, Boston, 1991; D. Witt, R. Klajn, P. Barski, and B. A. Grzybowski, Curr. Org. Chem., 8 (2004) 1763. [26]E. Delamarche, B. Michel, and Ch. Gerber, Langmuir, 10 (1994) 4103; G, E. Poirier and J. J. Tarlov, J. Phys. Chem., 99 (1995) 10966; K. S. Birdi, Self-Assembly Monolayer (SAM) Structures, Plenum Press, New York, 1999. [27]M. D. Porter, T. B. Bright, D. L. Allara, and C. E. D. Chidsey, J. Am, Chem. Soc, 109 (1987) 3559. [28]D. Hirayama, K. Takimiya, Y. Aso, T. Otsubo, T. Hasobe, H. Yamada, H. Imahori, S. Fukuzumi, and Y. Sakata, J. Am. Chem. Soc., 124 (2002) 532; S. Yamada, T. Tasaki, T. Akiyama, N. Terasaki, and S. Nitahara, Thin Sold Films, 438/439 (2003) 70; N. Terasaki, T. Akiyama, and S. Yamada, Langmuir, 18 (2002) 8666. [29]A. J. Thiel, A. G. Frutos, C. E. Jordan, R. M. Corn, and L. M. Smith, Anal. Chem., 69 (1997) 4948; C. E. Jordan, A. G. Frutos, A. J. Thiel, and R. M. Corn, Anal. Chem., 69 (1997)4939. [30]A. Stone and E. B. Fleischer, J. Am. Chem. Soc, 90 (1968) 2745. [31]T. Akiyama, H. Imahori, and Y. Sakata, Chem. Lett., (1994) 1447. [32]M. Kasha, Radiat. Res., 20 (1963) 55; R. F. Khairatdinov, and N. Serpone, J. Phys. Chem. B, 103 (1999) 761; N. C. Maiti, S. Mazumdar, and N. Periasamy, J. Phys. Chem. B, 102 (1998) 1528. [33]R. Rossetti, L. E. Nrus, J. Chem. Phys., 76 (1982) 1146. [34]Z-J. Zhang, A. L. Verma, N. Tamai, K. Nakashima, M. Yoneyama, K. Iriyama, and Y. Ozaki, Thin Solid Films, 333 (1998) 1. [35]A. Ishida and A. Fujii, Chem. Comm., (2005) 608 [36]L. Bogorad and I. K. Vasil, The Molecular Biology of Plastids, Academic Press, San Diego, 1991; C. H. Shaw (ed), Plant Molecular Biology - a practical approach, IRL Press, 1988. [37]D. Kim and A, Osuka, Ace. Chem. Res., 37 (2004) 735. [38]S. B. Powless, Ann Rev. Plant Physiol., 35 (1984) 1544. [39] A. Ishida, Proceeding of The 1st International Congress on Bio-Nanointerface, May 19, 2003, Tokyo, 222.
184 184
A. Fujii and A. Ishida
[40]I. Taniguchi, Bioelectrochemical Conversion of Molecules by Enzyme-Mediator Systems, "New Challenge in Organic Electrochemistry", T. Osa (ed), Gordon & Breach Publ., Tokyo, 1998, pp237-252. [41]A. Trebst, Methods EnzymoL, 24 (1972) 146; M. Okano, T Iida, H. Shinohara, H. Kobayashi, and T. Mitamura, Agric. Biol. Chem., 48 (1984) 1977. [42]U. Schreiber, Photosynth. Res., 9 (1986) 261. [43]A. Marshall and J. Hodgson, Nat. Biotechnol, 16 (1998) 27; G. Ramsay, ibid., 16 (1998) 40. [44]R. E. Oosterbroek and A. van der Berg (eds), Lab-on-a-Chip, Elsevier, Amsterdam, 2003; A. Manz and H. Becker (eds), Microsystem Technology in Chemistry and Life Science, Topics in Current Chemistry 194, Springer, Berlin, 1998. [45]U. C. Fisher, J. Opt. Soc. Am., B 3 (1986) 1239. [46]The Handbook — A Guide to Fluorescent Probes and Labeling Technologies, Tenth Edition, Molecular Probes Inc. Eugene, 2005. [47]A. Fujii and A. Ishida, Proceeding of The 1st International Congress on BioNanointerface, May 19, 2003, Tokyo, 222. [48]E.P. Medyantseva, E. V. Khaldeeva, and G. K. Budnikov, J. Anal. Chem., 56 (2001) 886. [49]B. Foultier, L. Moreno-Hagelsieb, D. Flandre, and J. Remade, IEE Proc.-Nanobiotechnol, 152 (2005) 3.
Handai Nanophotonics, Volume 2 Kawata and H. Masuhara Masuhara (Editors) (Editors) S. Kawata © 2006 2006 Elsevier B.V. All All rights reserved.
185 185
Chapter 9
Localized surface plasmon resonance enhanced secondharmonic generation K. Kajikawa,"-" S. Abe, a Y. Sotokawa,* and K. Tsuboi" interdisciplinary Graduate School of Science and Engineering, Tokyo Institute of Technology, 4259 Nagatsuta, Midori-ku, Yokohama 226-8502, Japan ^PRESTO, Japan Science and Technology Agency, 4-1-8 Honcho Kawaguchi, Saitama Japan 1. INTRODUCTION A number of reports have been appeared on localized surface plasmon resonance [1], which occurs in metallic nanostructures such as rough surfaces and nanoparticles. Considerably intense electric fields of light are localized as an evanescent field around the nanostructures at the localized surface plasmon resonance condition. Even single molecular Raman spectroscopy has been realized using this phenomenon [2,3]. Among various systems in which localized surface plasmon resonance occurs, spherical metallic nanoparticles or their aggregates are most important from the viewpoint of the application to nanophotonics and nanobiology, because the systems are well defined. This article concerns experimental measurements of the electric field around the spherical gold nanoparticles. Although the analytical expression to evaluate the localized electric field is given as a solution of Maxwell's equations with the appropriate boundary conditions [4,5], little experimental evaluation has been carried out on the localized electric field, to our knowledge. Thus the actual enhancement factors originating from the localized surface plasmon resonance are still obscure. One may think of the ways to measure the enhancement factors: (1) the scattering intensity measurement of the evanescent field around the nanostructures using a tip for near-field optical microscopy (2) the fluorescence intensity measurement from fluorescent dyes that cover the metallic nanostructures (3) the Raman scattering measurement from the molecules over-coating the metallic nanostructure. However these methods will not provide us with the enhancement factors in a qualitative manner, because neither fluorescence nor Raman processes are coherent so that the radiation is
186 186
Kqjikawa et al. K. Kajikawa
omnidirectional. In addition, fluorescence is frequently quenched, when chromophore is located at a metallic surface, owing to the charge transfer process. Second-harmonic generation (SHG) is one of the second-order nonlinear optical effects. It is a coherent process and is allowed in a system without any inversion centre, so that it is known as a powerful surface probe [6]. Accordingly SHG from the nonlinear optical dyes that covers the metallic nanostructures will enable us the quantitative argument of the enhancement factors. As a result of our study, it was found that the enhancement factor for spherical gold nanoparticles of about 20 nm in diameter is about 26, which is by 50% smaller than the theoretically predicted one. This article also gives another topic on the SHG response from the surface-immobilized nanoparticles on a gold surface with a gap distance of a few nanometers. According to the electromagnetic theory, it is predicted that a large electric field is produced in the gap, so that the SHG will be intense compared with a gold surface without any gold nanoparticles. The present experimental results support this expectation. Both findings will give us experimental evaluation of the field enhancement originating from the localized surface plasmon resonance of metallic nanostructures. 2. ENHANCEMENT FACTOR OF LOCALIZED FIELD AT GOLD NANOPARTICLES Gold is stable and hardly oxidizes, and is compatible with biological molecules. Spherical nanoparticles of gold are, therefore, of importance from the viewpoints of application such as biological sensing and electrochemistry. Although a number of reports have appeared on biosensing with the gold nanoparticles [7-10], there are few works that use the intense electric field around the gold nanoparticles. Also little is known about the enhancement in practice. To clarify this, we adopted SHG from the surface-immobilized gold nanoparticles on a dielectric substrate, in which the surface of the particles are covered with SHG-active chromophoric molecules. First, we consider the electric field localized around metallic nanoparticles. Let us assume a geometry of a metallic nanoparticle whose dielectric constant is £\{O) in a medium with a dielectric constant £*2(jf35 a t a frequency Q. When the electric field of light E{Q) is incident to this system, we can consider that a polarization of the nanoparticle p{Q) appears at the center of the spherical particle [4,5]. Since the diameter of the metallic nanoparticles, R, is much smaller than the wavelength of the light X, quasi-static approximation is held. Hence the polarizability of the nanoparticles p(Q) can be described as p(Q) = e2(Q)a{Q)E (O) = 4nR3E2(Q)A(O)E(S)
(1)
Localized surface plasmon resonance enhanced second-harmonic generation
187 187
«(i2) refers to the polarizability of the nanoparticles and is expressed as
(2)
A{£2) is corresponding to an amplitude factor of the local electric field, and the localized surface plasmon resonance condition occurs when the absolute value of the denominator of A(£2) is minimum. To satisfy this condition, the real part of £\{£2) should be negative, namely, the medium 1 is metallic, since E2(i2)^ 1 in general. In order to describe the electric filed at a position Q around the nanoparticles, it is convenient to use the spherical polar coordinate, in which the position Q is defined as (r, 0, @) as descried in Fig. 1. The local field at Q, E^iQ), can be expressed as
(3) L{Q) is a tensor called a local field factor and is expressed as 3A'(Q)sin20cos1 -A L(O) = 3A'(Q)sin2 0sin0cos0 3A% {Q) sin 0 cos 6 cos
3A'(Q)sia1 ( 3A'(Q)sw2 0sin 2 -A\Q) + \ 3A'(Q)sin0cos0sin 3Al{a)sYa9cos6sux$ l-A'(,Q) + 3A'(Q)cos20
(4)
z
Q
ε 2(Ω)
θ r
polarization
S E(Ω)
ε 1(Ω)
R y
incident light
φ x Fig. 1. Spherical coordinates used for the analysis.
188 188
Kqjikawa et al. K. Kajikawa
where (5)
A\£ty=-A(Q) at the surface of the nanoparticles. Figure 2 shows the calculated electric field intensity around the nanoparticle at the resonance condition when the light is incident to the nanoparticles. The electric field is about 5 times more than that of incident light at limited parts. Therefore one can investigate the surface of the nanoparticles in detail by optical measurements with various combinations of the angle of incidence and polarizations. When the light with a large electric field E(a)) from a high power laser is incident to a medium, the induced polarization P(a>) is expressed as (6)
•).
P(a>) = ,
%m is a linear susceptibility and is the origin of a linear dielectric constant and a refractive index. Other susceptibilities are called nonlinear susceptibilities. For instance, %{1) is a second-order nonlinear susceptibility and is a third-rank tensor. SHG originates from the second-order term of Eq. (6), and is a coherent phenomenon that light at a frequency fijis converted to 2»light. Thus SHG is prohibited in a system with inversion symmetry. In case of a molecular system, microscopic polarization for each molecule ^(aj)can be defined as (7)
P(of) =
30 E 0 z -30
30
0
30 nm
yy
Fig. 2. Enhancement of the electric field around the gold nanoparticles due to localized surface plasmon resonance. The field amplitude is indicated with brightness.
189 189
Localized surface plasmon resonance enhanced second-harmonic generation
where a's are the molecular polarizabilities. £te(
(a)
(b)
gold nanoparticle
L
hemicyanine SAM
θa
C4H9
silane coupling agent
C4H9
N+
S 2
glass substrate Transfer Charge Transfer
Fig. 3. Schematics of the sample (a) and the chemical structure of hemicyanine disulfide (b).
190 190
Kqjikawa et al. K. Kajikawa
Nd:YAG laser A,=1064nm analyzer
polarizer
photomultiplier
Fig. 4, Optical setup used in the study.
First we show SHG intensity profiles as a function of angle of incidence as shown in Fig. 5. Clear interference profiles were observed from the sample in which both front and rear surfaces were covered with the gold nanoparticles with the chromophore [Fig. 5(a)], whereas the fringe was absent in the sample in which only front side was covered [Fig. 5(b)]. This means that the SHG lights from this system are highly coherent. It should be also noted that the gold nanoparticles were not fully covered with the chromophore, because the SHG will be absent in the case that the nanoparticles is completely covered with the hemicyanine monolayer, because of the presence of the inversion center. In order to evaluate the enhancement factors, we should estimate the polar angle of adsorption of the chromophore 0a [Fig. 3(a)]. This value can be obtained from the intensity ratio of Ipp(2ai} and Isp(2ai), where /sp(2<w) stands for s-polarized SHG field generated by p-polarized fundamental light for instance. Here we briefly introduce the procedure [12].
(a)
(b) 0.2 Θ
SHG intensity(a.u.)
SHG intensity(a.u.)
0.8 0.6 0.4
0.2
•
1
•
Θ
if
0.1 o.i -
C3
0.2 0
1
35
-40 -20
0
20 40
Angle of IncidenceΘ Incidence Incidence© (deg.)
0o t -40 -20
0
20 40
Angle of IncidenceΘ Incidence Incidence© (deg.)
Fig. 5. SHG intensity profiles of (a) both sides of the slide were covered with the gold nanoparticles and (b) single side of the slide was covered covered with the gold nanoparticles. The solid line and broken line show /pp and igp, respectively.
Localized surface plasmon resonance enhanced second-harmonic generation
191 191
The molecular second-order polarization p(2eo,&,)&t a position the surface of the nanoparticle can be described as {a, 0,#)£te (a), 9, #) = &{2) (Ltttt, 6, m(m))(L(a>, 0,#)£(a>))
(8)
where E(m) is the fundamental field and L(<»,0,$)is the local field factor at a position Q(#,$) at a frequency ax The observed SHG field J?(2o)is the sum of the field from each molecular polarization. Therefore E(2eo) is expressed as (9)
where B is a constant independent of 0 and . The sample consists of surface immobilized gold nanoparticles, where the surface of the gold nanoparticles is covered with the hemicyanine monolayer, with a surface number density Np, The molecular surface number density of hemicyanine is defined as N&. Suppose that the hemicyanine molecule has a dominant diagonal second-order polarizability af^ along the long axis <§, and is aligned in the direction normal to the surface, both Jpp(2fi^ and 7^(2 ai) can be obtained through the integration of the area where the molecules are adsorbed. Therefore they can be described with an angle of incidence @as follows: ^^
2
Epp(2&>) = E
1
(w)BN&Np •
4
1l
4
%
2^
(3cos $sin€?sin # a - 2 s i n e£eos 6a +2sin @)dttfa\&£&
(10a) 2
Eip(2ai) = E (0})BNdNp
sin@sin 0aAto^a^Z
(10b)
^totai is the total enhancement factor and is expressed as A M = (1 + 2A{2a>))(l + 2A(a)f
(11)
In this evaluation, we assume that the gold nanoparticles are free from the substrate contribution, since the nanoparticles are located by 1.5 nm from the dielectric substrate. The ratio of Epp(2o^ and Esp(2e$, V, at an angle of incidence Q-riA is reduced to
192 192
V =
K. al. A;. Kajikawa Kqjikawa et etal
Epp(2w)
1
Ev{2oi)
sin 4
1 tan 4 0 a
3 2
(12)
Figure 6 shows the relation between V and #,. V is minimum at 6& = 90 deg, at which the curve is folded back. The ratio V = ^Iw{2co)ll^{2a) obtained from the experiments was 2.51. With this value, the theoretical simulation provides with the polar angle of 0a- 90 deg, namely, the upper hemisphere of the nanoparticles is covered with the chromophore whereas the lower hemisphere is not. Probably the solution does not soak into the gap due to the surface tension of the ethanol solution. With the absolute intensity of the SHG from this sample, the second-order surface susceptibility is evaluated to be 1.2xlO"13 esu by comparing the reflected SHG intensity from a quartz crystal at an angle of incidence 0—nlA. The crystal was 10 mm-thick, with which we can separate the SHG at the front surface of the crystal from that at the rear surface. From this value, the enhancement factor was found to be Atots&=Q.6. This value was somewhat smaller than the value theoretically predicted one ^4totai=51. This may be due to the following assumptions: (1) the contribution of the Fresnel factor was ignored; (2) the dielectric constant of gold nanoparticles may slightly differ from that of bulk gold; (3) the calculation was based on the assumption that gold nanoparticles were ideally spherical, and (4) we used the model that the hemicyanine was assumed to be ordered along the surface normal direction.
20
•
V
15 10
- -
0
&'i
10 10 5
θaa
1i
m. 2.5
•
i !
0
30 60 θ0,a (deg)
90
Fig. 6. The ratio Fas a function of the polar angle of adsorption of the chromospheres ft.
Localized surface plasmon resonance enhanced second-harmonic generation
3.
A LARGE ENHANCEMENT AT NANOPARTICLE AND SUBSTRATE
THE
GAP
193 193
BETWEEN
The gold surface shows SHG activity because the surfaces and interfaces are intrinsically noncentrosymmetric, while the SHG intensity from the surface is weaker than that from a dye monolayer such as a hemicyanine self-assembled monolayer. The mechanism of SHG from a metal is rather complicated. The SHG originates not only from lack of inversion center at a surface but also from a higher order contribution such as electric quadruple and magnetic dipole. Such higher order contribution has been extensively studied from mid 60's, both classically and quantum-mechanically. The mechanism of SHG from gold is rather complicated because of the interband transition in the visible region, so that the second-order nonlinear optical properties of a gold are still ambiguous. However gold is an important metal because it hardly oxidizes and is compatible with biological applications. Presence of the metallic nanoparticles on the surface may change the SHG response because of the electric field enhancement in the gap between the nanoparticles and the surface, which is sometimes called "gap mode", and of the change in the shape of the gold surface. We have formed such structures using gold nanoparticles immobilized above a gold surface, and found that this system shows intense SHG radiation. The sample used is illustrated in Fig. 7, which was fabricated with the following manner. A 300 nm-thick gold film (~300 nm) was vacuum-evaporated on a silicon wafer substrate, followed by deposition of the alkanethiol SAM that is terminated with an amino group (NH2-(CH2)n-SH: n=2, 6, 11). The gap distance from the gold surface to the nanoparticles can be controlled with the number of the methylene unit of the alkylchain. After rinsing the substrate with pure ethanol, the substrate was immersed in an aqueous solution of gold nanoparticles of 40 nm in diameter, followed by a rinse with pure water. The optical setup for the SHG measurements was the same as that used in the SHG study described above except for the detection of the SHG light in the reflection direction. gold nanoparticle amino alkanethiol n=2,6,11) (NH (NH22(CH (CH22))nnSH, SH,n=2,6,ll) NH2 NH2 NH2 NH2 NH2 NH2
gold S
S
S
S
S
S
200nm Fig. 7. Schematics of the sample and the scanning electron microscope image of gold nanoparticles deposited on a silicon wafer.
194 194
Kqjikawa et al. K. Kajikawa
0.8 AET(n=2)
0.6
AHT (n=6)
0.4 0.20.2
400
AUT (n=11)
500 600 700 wavelength/nm
Fig. 8. Absorption reflection spectra in water from a gold surface where the gold nanoparticles are immobilized. The spacer monolayers are AET (n =2), AHT (« =6) and AUT(«=11).
Figure 8 shows the reflection absorption spectra for the p-polarized light at an oblique incidence with an angle of 45 degrees from the surface normal. In addition to the first peak at around 540 nm, which is usually observed in spherical gold nanoparticles dispersed in solution, another peak appears in the range of 650 nm to 700 nm. Let us call it the second peak. The second peak cannot be observed at normal incidence or for s-polarized light, so that this band originates from the surface normal component of the electric field of the ppolarized light. This is because the normal component of the electric field produces quadrapolar interaction between the nanoparticles and their mirror images in the gold substrate, whereas the in-plain component does not. The position of the second peak is blue-shifted with increasing the distance owing to the difference in the strength of the interaction. These results support the previous theoretical [13,14] and experimental [15] studies. Therefore it can be concluded that the second peak originates from the localized surface plasmon resonance band coming from the quadruple contribution, indicating the strong interaction between the gold nanoparticles and the gold substrate surface. In order to evaluate the enhancement factors, the SHG spectra were measured with a fs TkSaphire laser at ranging from 730 nm to 850 nm. Figures 9(a) and 9(b) respectively show the spectra from the sample with a gap of AET (H=2) and AUT («=11). The SHG spectra clearly show frequency dispersion: the SHG is larger with the shorter wavelength excitation. It is likely that the increase of the SHG intensity mainly comes from the resonance effect of the second peak. Although the second peak of the sample with AET is not distinct as shown in Fig. 8, the SHG spectrum clearly indicates the dispersion.
Localized surface plasmon resonance enhanced second-harmonic generation
(a)
195
(b)
iS
700
750 800 wavelength /nm
850
700
750 800 wavelength /nm
850
Fig. 9. SHG spectra from a gold surface where the gold nanoparticles are immobilized. The spacer monolayers are (a) AET (n =2) and (b) AUT {n=\ 1).
The enhancement factor was evaluated with the SHG intensity taken with a Nd:YAG laser (/l=1064 nm). Since the SHG intensity observed from the samples was more intense than that from the gold surface, we can regard the observed SHG as being contributed from the surface areas where the gold nanoparticles are immobilized. Figure 10 summarizes the SHG intensity divided by the cross section of the nanoparticles. The value is normalized by the SHG intensity from the gold surface. More than 102-fold SHG intensity was observed from these samples in which the nanoparticles were located above the gold surface. The possible reasons for the intense SHG response is (a) the electric field enhancement in the gap between the gold nanoparticles and the substrate and (b) the change in the shape of the gold surface provides additional noncentrosymmetic character. The clear SHG dispersion in Figs. 9 (a) and 9(b) will lead to conclude that reason (a) will be dominant, since the SHG intensity will be hardly affected by the fundamental frequency on reason (b). 4. SUMMARY We have experimentally evaluated the enhancement factor originating from localized plasmon resonance by use of SHG, which is a powerful probe to evaluate the intensity of electric field because it is coherent and is free from quenching even at the metallic surface. The major findings are (a) the actual enhancement factor for spherical gold nanoparticles in diameter is smaller than the theoretically predicted one, and (b) A large enhancement was observed in the SHG response from the surface-immobilized nanoparticles on a gold surface with a gap distance of a few nanometers.
196 196
K. Kajikawa et al. Normarized Normalized SHG intensity (a.u.) 11
10 10
100 100 1000 1000
bare bare
AET
AHT
AUT
Fig. 10. SHG intensity at aj=1064 nm from a gold surface where the gold nanoparticles are immobilized. The spacer monolayers are AET (« =2), AHT («=6) and AUT (n=ll). The SHG intensity is normalized by the cross sectional are of the gold nanoparticles to evaluate the enhancement factor easily.
5. ACKNOWLEDGEMENTS We thank Prof. H. Okawa and Prof. K. Hashimoto of Kogakuin University for providing the hemicyanine disulfide.
REFERENCES [1] [2] [3] [4]
T. Okamoto and Kogaku, 33 (2004) 152 and references therein [in Japanese]. S. Nie and S. E. Emory, Science, 275 (1997) 1102. H. Xu, E. J. Bjemeld, M. Kaell, and L. Boerjesson, Phys. Rev. Lett., 83 (1999) 4357. C. F. Bohren and D. R. Huffman, Absorption and Scattering of Light by Small Particles, Wiley, New York, 1983, pp.130-165. [5] S. Kawata (ed.) Near Field Optics and Surface Plasmon Polaritons, Springer, Berlin, 2001, pp. 97-122. [6] Y.R. Shen, Nature, 337(1989)519. [7] M. Himmelhaus and H. Takei, Sens. Actuators B (Chemical), 63 (2000) 24. [8] N. Nath and A. Chilkoti, Anal. Chem., 74 (2000) 504. [9] K. Mitsui, Y. Handa, and K. Kajikawa, Appl. Phys. Lett., 85 (2004) 4231. [10] T. Okamoto, I. Yamaguchi, and T. Kobayashi, Opt. Lett., 25 (2000) 372. [11]R. Naraoka, G. Kaise, K. Kajikawa, H. Okawa, H. Ikezawa, and K. Hashimoto, Chem. Phys. Lett., 362 (2002) 26. [12] K. Kajikawa and Y. Sotokawa, in preparation. [13] P. K. Aravind, A. Nitzan, and H. Metiu, Surf. ScL, 110 (1981) 189. [14] M. M. Wind, J. Vlieger, and D. Bedeaux, Physica, A 141 (1987) 33. [15] T. Okamoto and I. Yamaguchi, J. Phys. Chem. B, 107 (2003) 10321.
Nanophotonics, Volume 2 Handai Nanophotonics, S. Kawata and H. Masuhara (Editors) © 2006 Elsevier B.V. All rights reserved. reserved.
197 197
Chapter 10
Localized surface plasmon resonance-coupled photo-induced luminescence and surface enhanced Raman scattering from isolated single Ag nano-aggregates T. Itoh, K. Hashimoto, Y. Kikkawa, A. Ikehata, and Y. Ozaki* Department of Chemistry, Kwansei Gakuin University, Sanda, Hyogo 669-1337, JAPAN 1. INTRODUCTION Surface-enhanced Raman scattering (SERS) has recently been a matter of great interest from the view points of its applications as well as basic science because it allows one even single molecule detection [1,2]. According to recent theoretical studies of SERS, optical near-field at both an Ag nanoparticle junction and a sharp edge has enough amplitude corresponding to the Raman enhancement of 108 to 1010 order by localized surface plasmon (LSP) resonance [3]. Therefore, such metal particle positions are now expected as SERS active sites [1-3]. We have constructed new light scattering micro-spectroscopy system [4] and explored the following three major new findings concerning the mechanism of SERS in our previous studies: (1) The same optical anisotropy of SERS and the LSP resonance [5], (2) The changes in the LSP resonance bands during SERS active-to-inactive processes [6], and (3) The changes of SERS excitation profiles induced by shifts of band energy positions of the LSP resonances [7]. SERS has the important characteristic that a continuum always overlaps with a SERS spectrum [1,3,8-10]. The continuum is not present in normal free space molecular Raman scattering and cannot be explained by the above SERS electromagnetic theory. Thus, the importance of developing the SERS mechanism including the emergence of the continuum has recently been pointed out [1,8-10]. The continuum is conjectured to be luminescence or electronic Raman scattering from a roughened metal surface with adsorbed molecules. Brus et al. proposed that the continuum is due to Ag electronic Raman scattering caused by just one surface defect-an adsorbed molecule exchanging electrons with the metal [1]. Otto et al. proposed that the continuum emerges by a
198 198
T.T.hohetal. Itoh et al.
radiative charge transfer (CT) process [9]. The CT state is formed by mixing of a molecular LUMO state and an appropriate metallic state just above Fermi surface. Hildebrant and Stockburger demonstrated a CT band in a SERS excitation spectrum [11], and Weiss and Haran discussed SERS as CT resonance Raman scattering [12]. The luminescence from a rough Ag surface without an adsorbed molecule was assigned to the radiative recombination of holes in the d band with electrons in the conduction band coupling to the radiative decay of LSPs [13]. In this stem, it is reasonable that the luminescence from SERS active nano-aggregates is not due to electronic Raman scattering but due to molecular electronic transition coupling to the radiative decay of the LSP [14]. Figure 1 shows an energy diagram for the CT transition of a metal/adsorbed molecule system [1,8-10,15]. An electron excited by an incident photon (h(Bj) tunnels from the metal to the molecule and returns to the metal through the CT state. ECT indicates the energy level of the CT state, and is determined by the mixing of the molecular non-bonding state and the appropriate metal state. The transition dipole emerged by returning of the electron from the molecule (CT transition dipole) can couple with the LSP having the energy level of ELsp. Persson and Baratoff s calculation revealed that the yield of luminescence (hat) by coupling to the radiative decay of the LSP reaches up to 10"3 photons per a tunneling electron [16]. Figure 1 also shows the enhancement of CT luminescence by coupling to radiative decay of the LSP. An incident photon excites the LSP, and the energy stored in the Ag nano-aggregate is transferred to an adsorbed molecule via the LSP-CT dipole interaction. A portion of the energy activates the vibrational mode of the molecule and back to the Ag nano-aggregate via the interaction, and is scattered by coupling to the radiative decay of the LSP. The scattered photon loses energy by a molecular vibration. This process of enhancement of Raman scattering is called electromagnetic SERS model [17]. A portion of the energy stored in the molecule is back to the Ag nano-aggregate via the CT dipole-the LSP coupling. Against quenching due to Forster resonant energy transfer [18], this process simultaneously may enhance the molecular electronic transition of the CT luminescence mediated by the radiative decay of the LSP. Both the peak intensity and wavelength of a LSP resonance band strongly depend on the size and shape of an Ag nano-aggregate, and reflect clearly its local surrounding dielectric property [4,19]. Therefore, if the luminescence dipole couples to radiative decay of the LSP, it is anticipated that the continuum should reflect the optical properties of the LSP resonance and the strong dipole-LSP interaction between the adsorbed molecules and the SERS active Ag nano-aggregate may induce large changes in both the LSP resonance and luminescence bands during the SERS active-to-inactive process [6,20]. However, SERS active substrates always have quite inhomogeneous nano-structures and conventional absorption
Localized surface plasmon plasmon resonance-coupled photo-induced photo-induced luminescence and surface enhanced Raman scattering from from isolated single Ag nano-aggregates
199
and luminescence spectroscopy always give averaged spectra for an ensemble of many SERS-active and -inactive Ag nano-aggregates and their local surroundings. Therefore, these changes in the LSP resonance and luminescence bands cannot be directly correlated to each other. In order to exclude this inhomogeneity and examine the clear correlation among the LSP resonance, the continuum and SERS, we selectively measure isolated Ag nano-aggregates having the simple LSP resonance band structures with light scattering microspectroscopy [14,20]. metal surface
ehemisorbed molecule Si state
-LSP
ω hui,
ω
Dipole-dipole coupling
So state
Fig. 1. An possible energy diagram for the enhancement of the CT luminescence of a metal/molecule system [14,20]. hcDj, incident photon energy; hat, luminescent photon energy; EF, Fermi level energy; ELSP, energy of a LSP; CT, charge transfer state; Si and So, the lowest excited singlet state and the ground state, respectively.
The first purpose of this review paper is to discuss the novel SERS mechanism that we have recently proposed [14]. The main point of this review lies in the description about new findings on the strong optical correlations among the LSP resonance, SERS and the continuum. In particular, (1) we find that the continuum has clear a luminescence band structure, (2) the polarization-angle dependences of the LSP resonance, SERS and luminescence are the same as each other, (3) the changes in the band energy positions of the LSP resonance induce the same band changes in the luminescence, and (4) the LSP resonance having a higher quality factor yields stronger intensities of both SERS and luminescence. By combination of these new findings, we could, for the first time,
200
T. Itoh et al.
succeed in the direct demonstration of the SERS mechanism as follows; SERS is CT resonance Raman scattering coupling to radiative decay of the LSP [14]. The second purpose of this review is the direct demonstration of the mechanism of both the luminescence band variations and the temporal changes in the luminescence, LSP resonance and SERS for isolated Ag nano-aggregates during the SERS active-to-inactive processes [20], 2. EXPERIMENTAL For the SERS sample preparation, we followed the general procedure by Nie and Emory [2]. The mixture of a R6G aqueous solution (6.0>
Localized surface plasmon resonance-coupled photo-induced photo-induced luminescence and surface enhanced Raman scattering from from isolated single Ag nano-aggregates
201 201
LSP resonance bands [4,5,19]. Bright spots in Fig. 2(b) correspond to SERS active Ag nano-aggregates.
P
I White light lighL
Laser light
T
C
Ag nano-aggregate O1 P N L
Polychromator + CCD
O2
Pin
(a)
30x× 30 30UMII 30 µm (b)
30x× 30 30p.m 30 µm (c)
Fig. 2. (a) The experimental setup that we recently developed to measure LSP resonance and Raman scattering spectra for individual isolated Ag nano-aggregates [5-7,14,20]. C, dark or bright-field condenser; 01 and 2, objective lens; P, polarizer; N, notch filter; L, tube lens; Pin, pinhole; CCD, charge-coupled-device, (a) and (b) Dark-field and SERS images of selected sample surfaces, respectively [5].
The laser power for the sample irradiation was set to be 4 W/cm2, which was several tens time weaker value than those of other studies on single-molecule SERS [1,2,10,15]. Therefore, the luminescence was never observed for SERS inactive Ag nano-aggregates in the present measurements. It would be difficult to determine which is origin of the luminescence, molecular
202
T. Itoh et al.
electronic transition or inter metallic band transition (the recombination of holes in d band and electrons in sp band [13]) with the high laser power irradiation of several hundreds W/em2, because both SERS active and inactive Ag nano-aggregates emerge the luminescence. Another reason for using the lower laser power is to avoid the carbonization of adsorbed molecules and the destruction of Ag nano-aggregates. Under this irradiation condition, we did not observe a broad Raman band around 1560 - 1580 cm'1, which is attributed to the in-plain bond-stretching vibration of sp2 C-C band of graphite [21]. We also did not observe changes in the LSP resonance bands of SERS inactive Ag nano-aggregates under this irradiation condition [6]. Many SERS active Ag nano-aggregates showed "blinking" properties as reported by previous many articles [1,2,10,12,15,22], therefore we expects that the amount of adsorbed molecules on the Ag nano-aggregates is near single molecule level.
3. RESULTS AND DISCUSSION 3.1. Demonstration of quality factor sensitive SERS and luminescence intensities Figures 3 (a), (b), and (c) show SERS and luminescence bands from the same isolated Ag nano-aggregates measured with the excitation wavelength of 457, 532, and 633 nm, respectively. Narrow bands observed in the 470 - 510 nm region in trace (a) and those in the 545 - 560 nm region in trace (b) are due to SERS bands of R6G [11]. A clear luminescence band structure appears with peak energy at 620 nm instead of the well known continuum. The intensities and band energy positions on the Raman scattering have excitation photon energy dependences, however, the luminescence band does not vary its peak wavelength with the change of the excitation wavelength from 457 to 532 nm and disappears with the excitation wavelength of 633 nm. These results demonstrate that the luminescence is not due to Raman scattering but due to the radiative transition having excitation wavelength threshold between 532 and 633 nm. It is expected that an inverse transition band appears in a spectrum of the LSP resonance band. We show the simple LSP resonance bands of three SERS active Ag nano-aggregates in Fig. 3(d) - (f). These Ag nano-aggregates clearly show an additional weak band around 510 - 550 nm, and these bands disappear after SERS quenching [6]. These bands have resonance energy that is close to the absorption peak energy of free R6G molecule, and thus we assigned the bands to the enhanced HOMO to LUMO transition by coupling to the radiative decay of the LSPs [6]. However, From SERS excitation spectra it was clearly shown that a CT resonance band also appears around the energy position (around 570 nm (2.2 eV)) [11,12], and the present results support their study and
Localized surface plasmon resonance-coupled photo-induced photo-induced luminescence and surface enhanced Raman scattering from from isolated single Ag nano-aggregates
203
do not conflict with them. Thus, now we consider that the additional band and the luminescence band arise from the enhanced CT transition by coupling to the radiative decay of the LSP. These two bands cannot be observed for any SERS inactive Ag nano-aggregates and disappear after SERS quenching. This suggests that the huge enhancement of CT transition up and down is emerged at SERS active sites. Plasmon Plasmon resonance resonance band band (d) CT resonance band
e
'I
Scattering efficiency (a.u.)
Luminescence intensity (a.u.)
(a)
(b)
o o a o o
a
(c)
(f)
55 00 00 55 55 00 66 00 00 66 55 00 77 00 00
Wavelength/nm
450
500
550
600
650
700
450
500
550
600
650
700
44 5 00
55 0 00
55 5 00
66 0 00
66 5 00
77 00 0 0
(e)
Wavelength/nm
Fig. 3. (a), (b), and (c) The luminescence spectra from the same isolated Ag nano-aggregates measured with the excitation photon energy of 2.71,2.33, and 1.96 eV, respectively [14]. (d), (e), and (f) Dipolar LSP resonance and CT resonance bands for three SERS active Ag nano-aggregates [6,14].
We investigated polarizer-angle (0) dependences of the LSP resonance, CT luminescence, and SERS bands for an isolated Ag nano-aggregates and compared their anisotropy [5,6,7,14]. The polarization data were taken by rotating a polarizer-angle #from 0° to 150° with an interval of 30° as shown in Fig. 2. Fig. 4(a) and (b) show polarizer-angle dependences series of the white light scattering spectra and the Stokes shifted emission spectra, respectively. The white light scattering spectra clearly show two LSP resonance bands having peak wavelength at 680 and 430 nm. These LSP resonance bands are
204
T. Itoh et al.
accompanied by additional bands around at 520 nm assumed to be a CT resonance band. The Stokes shifted emission spectra show luminescence and (b)
(a) Short-axis
550 650 450 Wavelength/nm Wavelength/ntn (c) /
n C D 3
§
|
J
\
In
J H.
[
•<•*
•
r
3
1
o ~
\
9 0
' A
9
6
I
550
%
.
•.
\
iN
i
-
."X3
40 80 120 120 Polarization angle / degree
LSP resonance, SERRS, and luminescence intensity (a.u.)
LSP resonance intensity (a.u. )
Scattering efficiency (a.u.)
Luminescence intensity (a.u.)
Long-axis
1^
600
650
700
Wavelength/nm (d) i
•
1 Q. 1
1
i
• _
co
e co B » e o •-* O
o
OH
"3
- I -
e oc S e o o o
al
y j
\ D
-
x-
5 , 1 , 1
0
40 80 120 120 Polarization angle/degree
Fig. 4. (a) Polarization-angle dependences of LSP resonance and CT resonance bands [5,6,14]. (b) Polarization-angle dependences of CT luminescence and SERS bands [14]. (c) Plot of the intensities of the long-axis (open circles) and short-axis (closed circles) LSP resonance bands vs. the polarization-angles [5,6,14]. (d) Plot of the intensities of the long-axis LSP resonance (open circles), CT luminescence (open triangles), and SERS bands (open squares) vs. the polarization-angles [14].
SERS bands having peak wavelength at 630 and around 570 nm, respectively. When the intensity of the LSP resonance band having the longer resonance peak wavelength becomes the maximum at 8 of 60°, the CT resonance band, the luminescence band, and all the SERS bands also become maximum. These bands disappear when the LSP resonance band having the shorter resonance peak wavelength becomes the maximum at 6 of 150°. The open and closed
Localized surface plasmon resonance-coupled photo-induced photo-induced luminescence and surface enhanced Raman scattering from from isolated single Ag nano-aggregates
205 205
circles in Fig. 4 (c) indicate the peak intensities of the LSP resonance bands with the shorter and longer resonance peak wavelength, respectively, at each polarizer-angle. These polarizer-angle dependences are well fitted by cos(2#) curves. These polarizer-angle dependences show that these resonances are dipoles. The phase difference of each cos(2#) curve of the LSP resonance is 90°. This phase difference suggests that these resonance bands having the longer and shorter resonance peak wavelength correspond to the long-axis and short-axis LSP resonances, respectively [19]. The polarizer-angle dependences of the long-axis LSP resonance, CT resonance, SERS, and luminescence band intensities are shown in Fig. 4 (d) and are well fitted with the same eos(26>) curve. These fittings directly show that the SERS and luminescence couple to the long-axis LSP. The high field enhancement of the junctions between Ag nanoparticles may come from higher-polar LSP modes [1-3,8]. Therefore, the dipolar LSP coupling to the luminescence and SERS suggests that the luminescence and SERS observed in our experiment come from the sharp edge in the Ag nano-aggregate. The correlation among the SERS intensity, the luminescence intensity, and the quality factor of the LSP resonance is examined for the isolated Ag nano-aggregates to avoid the mhomogeneous broadening of the LSP resonance band that occurs for the collective measurement of many Ag nano-aggregates. The excitation wavelength is fixed at 532 nm. Quality factor Q {Q = Eres IF, Eres resonance peak energy, Ein incident laser photon energy (2.33 eV (532nm)), IIP dephasing time {F line width of LSP resonance band)) is a dominant parameter for the coupling of SERS and luminescence to the radiative decay of the LSP [19]. To achieve higher enhancement of the luminescence and SERS, the quality factor of the LSP resonance must be as large as possible. In particular, it is advantageous to minimize the nonradiative decay. In other words, it is useful to maximize the quantum efficiencies for the SERS and luminescence [19]. Now, we can identify the LSP coupling to the luminescence and SERS. Thus, it is possible to directly examine the correlation between the quality factor and the luminescence and SERS intensities. We show that the 4th power of the quality factor of long and short-axis LSP resonances vs. their band peak energy position is plotted in Fig. 5 (a). SERS intensity is believed to be proportional to the 4th power of quality factor [3,8,19]. It is clear from Fig. 5(a) that the 4th power of the quality factor of long-axis LSP resonance is always larger than that of short-axis one. This is the reason why we always observe coupling of the luminescence and SERS to the long-axis LSP. Figure 5 (a) also shows that the distribution of data points is quite inhomogenous comparing with that of nanorods [19]. This means that the shape and size of these Ag nano-aggregates are quite inhomogenous. The distribution of peak energy of the LSP resonance is considerably inhomogenous
206
T. Itoh et al.
Quality factor
20
(a)
15 10 5 0
Luminescence intensity(arb .u.)
thus the peak energy dependence of SERS and luminescence may be averaged. Therefore, we plot the 4th power of the quality factor of the LSP resonances vs. the luminescence and SERS intensities in Fig. 5(b) and (c), respectively. The same positive correlations are clear for both plots; this may be due to the consequence of the fact that the increment of radiative decay and the surpression of nonradiative decay increase the SERS and luminescence intensities.
2 1 0.8 0.6 0.4 0.2 0.1
(c)
9 4 22 33 4 5 6789 5 V 110 10 4th power of Q factor factor
Luminescence intensity(arb .u.)
SERS intensity(arb .u.)
400 450 500 550 600 650 Peak wavelength of LSP (nm)
2 (b) 1 0.8 0.6 0.4 0.2 0.1
2 3 4 556789 6 7891()55 9? Q 44 10 th 10 th 4 power of Q factor factor
2.5 2.0 1.5 1.0 0.5 0.0
(d)
1.0 1.5 1.5 2.0 0.0 0.5 1.0
SERS intensity (arb .u.)
Fig. 5. (a) Plot of the quality factors of long and short-axis LSP bands vs. their resonance peak energy positions [14]. (b) Plot of the peak intensities of SERS bands vs. the 4th power of the quality factors [14]. (c) Plot of the peak intensities of CT luminescence bands vs. the 4th power of the quality factors [14]. (d) Plot of the peak intensities of CT luminescence bands vs. the peak intensities of SERS bands [14].
The SERS intensities are proportional to the luminescence intensities as shown in Fig. 5 (d) and we find that their intensities also show the same temporal changes. These results strongly suggest that the CT resonance coupling to the radiative decay of the LSP is the origin of the luminescence and SERS. These results mean that the shape and size of these Ag nano-aggregates are largely inhomogenous and the quality factor of the LSP resonance is an important parameter to achieve the higher intensities of both the luminescence and SERS [14].
Localized surface plasmon plasmon resonance-coupled photo-induced photo-induced luminescence and surface enhanced Raman scattering from from isolated single Ag nano-aggregates
207
3.2. Variations of the luminescence spectra of SERS active isolated Ag nano-aggregates We found that the observed luminescence spectra largely differ from each other [20], Fig. 6(a) shows the fluorescence spectrum of an aqueous solution of R6G and Fig. 6(b) to (g) display the luminescence spectra for SERS active Ag nano-aggregates. All the luminescence spectra have the vibronic structures and these structures can be reproduced by the combinations of four types of bands pointed by perpendicular dotted lines at (1) 555, (2) 590 ±20, (3) 620 ±20, (4) 680 ±20 nm. Every band is well fitted by Lorentzian shape except the band (1) as shown in Fig. 6. (1)
(2)
(3)
(4)
(a)
Luminescence intensity (a.u.)
(b)
(c)
(d)
(e)
(f)
(g)
55 5500
6 00 600
65 6500
7 00 700
Wavelength / nm
Fig. 6. (a) Fluorescence spectrum of R6G, (b) to (g) Observed typical luminescence spectra from 6 isolated Ag nano-aggregates [14, 20]. A dotted line in (b) A fluorescence spectrum of R6G [20].
208 208
T. Itoh et et al. T.
SERS bands in Fig. 6 (b) are overlapped with the band (1) with the peak wavelength of 555 nm. This type of the luminescence spectrum usually appears just before SERS quenching and shows "blinking" properties simultaneously with the SERS spectrum in the present experiment. The dotted-line band in Fig. 6 (b) represents the fluorescence modulated by a notch filter of the spectroscopy system used, and clearly shows that the SERS spectrum is overlapped with the fluorescence spectrum. The field enhancement of the fluorescence by the LSP resonance is exponential to the distance between an adsorbed molecule and an Ag nano-aggregate, and it is possible that this enhancement overcomes the quenching by Forster resonant energy transfer [18] within a certain distance. Therefore, the observed luminescence in Fig. 6 (b) is thought to be the enhanced fluorescence just after desorption of R6G molecules from SERS active sites. SERS bands in Fig. 6 (c) are overlapped with the band (2) with the peak wavelength of 595 nm. This type of the luminescence spectrum is the most common in the present measurements. SERS spectra in Fig. 6 (d) and (e) are overlapped with the bands (2) and (3) with peak wavelengths of 590 and 630 nm, respectively. SERS bands in Fig. 6 (f) are overlapped mainly with the band (3) with the peak wavelength of 625 nm. SERS spectrum in Fig. 6 (g) is overlapped with three luminescence bands with peak wavelengths of 590, 625 and 695 nm. These luminescence spectra (c) to (g) clearly differ from the fluorescence, and therefore, the luminescence is not due to the molecular So +—Sj transition of free R6G molecule. These spectra are discretely decomposed into only four kinds of band, and thus the luminescence is not continuum the spectrum due to the metallic d-sp recombination modulated by the LSP [13]. Wess and Haran [12] noted that the CT transition band emerges at 570 nm in a SERS excitation spectrum of R6G, and they and Brus et al. [1] showed the validity of their CT-resonance Raman model by comparing their results with other studies by scanning tunneling microscopy (STM), electron energy loss spectroscopy, and ultraviolet photoemission spectroscopy. According to their discussion, the luminescence is due to the electron-hole recombination (an inverse transition of the CT). The time scale of this recombination process is within several to several-tens femtosecond by electron-electron scattering [23]. All observed luminescence bands are well fitted by Lorenzian shape and these homogeneous width; /"(100 to 150 meV) corresponding to the decay time; IIP of 30 to 40 fs. This value is consistent to the dephasing of the observed LSP by electron-electron scattering [1,23]. We summarize above discussions as that the varieties of the vibronic structures in the luminescence spectra correspond to variations in the combinations of four kinds of bands coupling to the radiative decay of the LSPs. By using ultraviolet photoemission spectroscopy Kim et al. [24] demonstrated that the mixing between a nitrogen lone pair electronic orbital and the appropriate metallic electronic orbital largely reduces its work function from
Localized surface plasmon plasmon resonance-coupled photo-induced photo-induced luminescence and surface enhanced Raman scattering from from isolated single Ag nano-aggregates
209
4.5 to 2.5 eV in the case of pyridine on an Ag surface. Therefore, the evidence for the chemical bonding yielding Ag-N stretching bands in the SERS spectra of R6G reveals that the CT luminescence can be emerged by this large reduction of the barrier height for CT [1,11,12]. Then, the variations of the luminescence bands may result from those of the orientation of chemisorbed R6G molecules on the Ag surface because their orientation determines the reduction of the barrier height for the CT. This orientation phase is fixed by the electrostatic interactions on the Ag surface, such as hydrogen bonding and van der Waals interaction. Wang et al. [25] proposed three possible orientations of rhodamine B molecules on an Au(l 11) surface, and demonstrated by using STM that the molecule in its adlayer is adsorbed on the surface by its two diethylamino function groups with an edge-on orientation. The present R6G concentration (10"10 M) was much lower than that of their STM measurement (10"s M). Therefore, the Ag surface is not completely covered by the molecules and all possible orientations are available for SERS active Ag nano-aggregates. In the stem of Smoluehowski electron smoothing effect [26], the reduction of the local barrier height for CT, h<j> depends on a tilted angle, 8 between the direction of the CT dipole and the normal of Ag surface (&$=0Ar-($Ar-
210
T. Itoh et al.
by an increment of the aspect ratio of the nano-aggregate shape [28]. The similarity of the shifts of both the peak wavelengths confirms that these CT dipoles couple with the LSPs, and reveals that the observed the slight inhomogeneity of each luminescence band position (±20 nm in Fig. 6) is originated in the spectral modulation by the LSP [20]. (a)
Scattering efficiency (a.u.)
Luminescence intensity (a.u.)
(b)
(c)
(d)
(e)
(f)
550
600 650 Wavelength / nm
700
Fig. 7. (a) to (f) The luminescence (black lines) and the corresponding LSP resonance (dotted lines) bands of 6 isolated Ag nano-aggregates [14,20].
We focus on the second and third luminescence bands pointed by dotted perpendicular lines (2) and (3) in Fig. 6, and the peak wavelengths vs. the peak wavelengths of the LSP resonance bands are plotted for 39 Ag nano-aggregates in Fig. 8. The increments of both the peak wavelengths show that the change in the LSP resonance band position induces a shift in the luminescence band. This trend also supports that the conclusion of the above paragraph. The
Localized surface plasmon resonance-coupled photo-induced photo-induced luminescence and surface enhanced Raman scattering from from isolated single Ag nano-aggregates
211
Peak wavelength of luminescence band / nm
inhomogeneity in data points around 610 nm of the horizontal axis arises from the overlapping of the SERS and luminescence spectra. 660 660 640 640
o o a u o e |
620 620 600
-
e o
580
-
560
-
1 O
OH
600
620 620
640 640
660 660 of LSP resonance band //nm nm Peak wavelength of
680
Fig. 8. A plot of peak wavelength of the luminescence band vs. that of the LSP band for 39 isolated Ag nano-aggregates [14,20].
3.3. Temporal changes in the luminescence spectra of SERS active isolated Ag nano-aggregates We found that the luminescence sometimes largely varies its spectral shape and intensity [20], The observed temporal profiles of the luminescence spectra can be roughly classified into two types; the luminescence band with a large peak shift and that with a small peak shift. The luminescence bands that temporarily change their peak wavelengths are shown in Fig. 9 (a), (b), and (c) for the three major cases shown in Fig. 6 (c), (f), and (g), respectively. The luminescence band with almost no peak wavelength shift is shown in Fig. 9 (d). This type of temporal change is the most common in the present experiments. The luminescence signals usually show blinking properties reported as single-molecule SERS signals (10'14 M R6G concentration) [1,2,10,12,15,22] even the concentration up to 10"7 to 10"10 M. Therefore, this blinking may be suspected as fluctuation of Ag nano-aggregates. To exclude this complexity of the discussion of the spectral changes, stable temporal ones (without the blinking) in the luminescence are selectively measured.
212
T. Itoh et al.
Luminescence intensity (a.u.)
(a)
(b)
a
•3
(c)
—
(d)
550
600
650
700
Wavelength / nm Fig. 9. (a) to (d) Temporal changes in the luminescence spectra of 6 isolated Ag nano-aggregates [20].
The dotted arrows in Fig. 9 indicate the direction of time change. A peak at 600 nm in Fig. 9 (a) shifts to 580 nm with the decrements of the peak intensities of both the SERS and luminescence bands. A peak at 630 nm in Fig. 9 (b) shifts to a shorter wavelength side with the intensity decrease in the SERS and luminescence bands. Then, the peak shifts to almost the same position as that of the luminescence band in Fig. 9 (a) with the intensity increase. The following temporal band change is almost the same as that in Fig. 9 (a). Two peaks of the luminescence bands in Fig. 9 (c) shift to shorter wavelength sides with the intensity decrease until the band at the shorter wavelength shifts to about 590 nm. Then, the band with the longer peak wavelength disappears when the other band with the shorter peak wavelength becomes almost the same as that in Fig. 9
Localized surface plasmon plasmon resonance-coupled photo-induced photo-induced luminescence and surface enhanced Raman scattering from from isolated single Ag nano-aggregates
213
(a). Note that the same band appears in the inactive process in Fig. 9 (a) and (b). A peak at 630 nm in Fig. 9 (d) does not shift and the peak intensities of both the SERS and luminescence band vertically decrease. The CT dipole of SERS yielding molecule couples strongly with the LSP because electromagnetic field at the SERS active site estimated to several thousand times larger than that at the normal surface of Ag nano-aggregate [3]. Therefore, the weakening of this CT dipole-LSP coupling is the same as the decrement of local surrounding dielectric constant and may emerge the change in the luminescence band. It was reported that the chemisorbed molecules on Ag surface was showed the orientation phase transition [29]. Thus, we consider that two mechanisms are included in the observed temporal changes as follows; (A) The destabilisation of the adsorbed molecule by the laser irradiation induces the transition of the molecular orientation phase corresponding to the higher energy level of CT state, and consequently the CT luminescence is blue-shifted; and (B) the decrement of the local dielectric constant by the weakening of the CT dipole-LSP coupling makes the LSP resonance blue-shifted and thereby the luminescence also blue-shifted. The mechanism (A) predicts only the synchronous decrement and increment of the peak intensities of band (2) and band (3) in Fig. 6 without their blue-shifts. On the other hand, by adding the mechanism (B), the observed continual changes from the band (3) to (2) in Fig. 9 will be accountable. Only using the mechanism (B), we cannot explain the most common decay profiles without band blue-shift as shown in Fig. 9 (d). Thus, the combination of the above mechanism (A) and (B) is important to describe all the experimental band changes. The decay of peak intensities of both SERS and the luminescence without band-shift in Fig. 9 (d) indicates that the luminescence yielding molecules desorb from or bleach at the SERS active site without orientation phase transition. The plot of the band intensity vs. it peak wavelength in Fig. 9 (c) is also shown in the upper panel of Fig. 10 (b). The dependence of the SERS intensity on the longer and shorter luminescence wavelengths also show the saddle points around 610 and 650 nm, and both their temporal changes are similar to that in Fig. 10 (a). This rapid decrement of the peak intensity restarting from 580 nm is shown in the upper panel of Fig. 10 (a). This decrement suggests the desorption of the molecules yielding the luminescence. The peak intensity of the SERS band (closed circles) vs. the peak wavelength of the luminescence is plotted in the lower panel of Fig. 10 (a) (the peak intensity of the SERS band at 584 nm is background-corrected).
214
T. Itoh et al.
Luminescence (a.u.)
intensity
(a)
+•
(b)
1
1
1 .- .1
-
f
•
•
• f *»l
1
1
1 - +• • i
SERRS intensity (a.u.)
•3 -5-
e o e oo
y« f **i 560
i 600
i
i - f 640
560
•
i
/
\ ^ 1
°'
+
QD
-
O
» \
O
$
W
•
• ^
a a
i
~
° i
cP|
' • ' It ' / ••• \ • • ••• • 9 •
'
t
i
4
'o h Q)°
0
: 600
640
680
Peak wavelength of luminescence band / nm
Fig. 10. (a) The plot of the peak wavelength vs. the peak intensity of the luminescence band (upper panel) and the plot of the peak wavelength of the luminescence band vs. the SERS intensity (lower panel) of the Ag nano-aggregate in Fig.. 5 (b) [20]. (b) The plot of the longer and shorter peak wavelengths of the luminescence bands vs. their peak intensities (upper panel) and the plot of the longer and shorter peak wavelengths vs. the SERS intensities (lower panel) of the Ag nano-aggregate in Fig.. 6 (c) [20].
The intensity of the Raman band at 1650 cm"1 due to the aromatic C-C stretching mode is used as the SERS intensity [11,12]. The temporal change in the SERS intensity differs from that of the luminescence in the upper panel. The SERS intensity monotonously increases in the region from 630 to 580 nm, and coincidentally decreases together with the luminescence from 580 nm. These coincident decrements suggest that the SERS yielding molecules are the same as the luminescence yielding ones. The monotonous increments of the SERS intensities in the region of 630 to 580 nm can be explained as the increments of the CT resonance effect by an approach of the CT transition energy gap to the laser photon energy. These experimental and calculated results show that the coupling of the LSP and CT dipole enhances both the Raman and luminescence signal from the same molecule. We also make the same plot as Fig. 10 (b) in the
Localized surface plasmon plasmon resonance-coupled photo-induced photo-induced luminescence and surface enhanced Raman scattering from from isolated single Ag nano-aggregates
215
Scattering efficiency (a.u.)
case shown in Fig. 9 (c). The SERS intensity vs. the longer peak and shorter wavelengths in Fig. 9 (c) are shown in the lower panel of Fig. 10 (b). The dependence of SERS intensity on the longer peak wavelength and that on the shorter one show the maximum points around 590 and 660 nm, respectively, and their temporal properties are the same as that of Fig. 10 (a). If the observed temporal changes in the luminescence are induced by the weakening of the CT dipole-LSP coupling (the decrement of local surrounding dielectric constant), the LSP resonance also should show the coincidental temporal band changes. We selectively monitored the LSP resonance and the luminescence classified as band (2) in Fig. 6. Fig. 11 (a) and (b) show the changes in the LSP resonance and luminescence bands during SERS active-to-inactive process. The dotted arrows in Fig.. 11 indicate the directions of time course. The LSP resonance bands shift to a longer wavelength side by 32 nm and their peak intensities increase, and the luminescence bands coincidentally shift by 29 nm and their peak intensities decrease. (a)
o
a u
u e o
Luminescence intensity (a.u.)
450
500
550 600 Wavelength // nm nm Wavelength
650
700
(b)
u o o
au
s 550
600
650 Wavelength / nm
700
Fig. 11. (a) and (b) Temporal changes in the LSP resonance and luminescence band of the same isolated Ag nano-aggregate. Dotted lines in (a) A variation of the calculated LSP resonance bands of an Ag ellipsoid coated with R6G films having different thickness [6,20]
216 216
T.hohetal. T. Itoh et al.
We tried to estimate the strength of the CT dipole-LSP coupling as the amount of adsorbed molecules. We calculated the changes in the LSP resonance bands of Ag nano-ellipsoids coated with and without R6G films using point dipole approximation [30] because the SERS active Ag nanopaticle always forms an aggregate whose shape is always deviated from a spherical shape [1,2,5]. We supposed that the dielectric constant of adsorbed R6G is the same as that of R6G in the solid state. The changes in the LSP resonance bands in the calculated aspect ratios of the ellipsoids before and after the addition of the R6G film are determined to be (21.2 (long-axis), 12.0 (short-axis), 4.1 (height)) to (20.0,12.0,4.1), respectively, and these calculated bands are shown as the dotted lines in Fig.. 11 (a). The height of SERS active nano-aggregates were usually about 40 nm by atomic force microscopy measurements [5]. Thus, it is expected that the length of long, short, and height axes of the ellipsoid are about 200, 120, 40 nm, respectively, and the film thickness is estimated to be 12 nm. It means that the amount of adsorbed R6G molecule is about 210,000. However, in the experimental section, we assumed that the number of molecules adsorbed at a SERS active site is near single-molecule level if the "SERS blinking" property is one of the proofs for the single-molecule detection [1,2,10,12,15,22]. This inconsistency show that the CT dipole-LSP interaction between the nano-aggregate and the luminescence yielding molecules is several thousand times stronger than the normal electromagnetic interaction as predicted by theoretical works [3]. These results reveal that the change in the LSP resonance and luminescence band is induced by the weakening of the CT dipole-LSP interaction between the luminescence yielding molecules and the Ag nano-aggregate [20].
4. CONCLUSIONS To directly show the LSP contribution to the huge Raman enhancement, we have investigated SERS, luminescence and LSP resonances of isolated Ag nano-aggregates [14]. We have found that the luminescence is not due to Raman scattering but due to the CT transition that couples to the radiative decay of LSP at the SERS active site. The polarization measurements of the SERS and luminescence of the nano-aggregates identify LSP coupling to them [5,14,20]. It is demonstrated that the larger quality factor of the LSP resonance is important to achieve higher intensities of the both luminescence and SERS [14]. In particular, to directly show the CT contribution to the large field enhancement, we have investigated the origin of the variations and temporal changes in the CT luminescence bands [20]. The variations of luminescence spectra were reproduced by combination of four kinds of Lorenzian band and each peak wavelength and shape of the band clearly reflects the LSP resonance of the Ag
Localized surface plasmon plasmon resonance-coupled photo-induced photo-induced luminescence and and surface enhanced Raman scattering from from isolated single Ag nano-aggregates
217
nano-aggregate [20], The changes in the LSP resonance and luminescence bands were represented as the decrement of the local dielectric constant induced by the weakening of the CT dipole-LSP interaction between the luminescence yielding molecules and the Ag nano-aggregate [20]. The simultaneous decay of the luminescence and SERS bands in this process has been interpreted as the desorption of the luminescence and SERS yielding molecules, and it has been suggested that the luminescence yielding molecules are the same as the SERS yielding ones [20]. In this review paper, we have directly showed the importance of both the quality factor of the LSP resonance and the CT between the chemisorbed molecule and the Ag surface to achieve higher and more stable SERS signal, and concluded that SERS is the CT resonance Raman scattering enhanced by amplified electromagnetic field and coupling to the radiative decay ofLSP. ACKNOWLEDGMENTS The authors thank Dr. S. Sasic of our group and Prof. H. Tamaru of the University of Tokyo [19] for stimulating discussion on the SERS mechanism. The authors thank Prof. N. Tamai of Kwansei Gakuin University, Prof. T. Asahi of Osaka University, and Prof. N. Ichinose of Kyoto Institute of Technology for stimulating discussion on the mechanism of the photo-induced luminescence. This study was partly supported by Grant-in-Aid for Young Scientists "WAKATE (B)" (No: 16760042) from MEXT (Ministry of Education, Culture, Sports, Science and Technology). This work was also supported by "Open Research Center" project for private universities: matching fund subsidy from MEXT. REFERENCES [1] A. M. Michaels, M. Ninnal and L. E. Bras, J. Am. Chem. Soc., 121 (1999) 9932. J. Jiang, K. Bosnick, M. Maillard and L. E. Bras. J. Phys. Chem. B, 107 (2003) 9964. [2] K. Kneipp, Y. Wang, H. Kneipp, I. Itzkan, R. R. Dasari and M. S. Feld, Phys. Rev. Lett., 76 (1996) 2444, S. Nie, Emory, S. R. Science, 275 (1997) 1102, H. Xu, E. J. Bjemeld, M. Kill and L. Borjesson, Phys. Rev. Lett., 83 (1999) 4357. C.J.L. Constantino, T. Lemma, P. Antunez, R. F. Aroca, Anal. Chem., 73 (2001) 3674. [3] F. J. Garcia-Vidal and J. B. Pendry, Phys. Rev. Lett., 77 (1996) 1163, H. Xu, E. J. Bjemeld M. Kail and L. Borjesson, Phys. Rev. Lett., 83 (1999) 4357, P. Zhang, W. Huynh, L. Tay, T. L. Haslett and M. Moskovits, Phys. Rev. B., 59 (1999) 10903, M. Futamata, Y. Maruyama and M. Ishikawa, Vib. Spectrosc, 30 (2002) 1714. [4] T. Itoh, T. Asahi and H. Masuhara, Appl. Phys. Lett., 79 (2001) 1667. T. Itoh, T. Asahi and H. Masuhara, Jpn. J. Appl. Phys., 41 (2002) L76. [5] T. Itoh, K. Hashimoto and Y. Ozaki, Appl. Phys. Lett., 83 (2003) 2274. [6] T. Itoh, K. Hashimoto, A. Ikehata and Y. Ozaki, Appl. Phys. Lett., 83 (2003) 5557.
218
T. Itoh et al. T.Itohetal.
[7] T. Itoh, K. Hashimoto, A. Ikehata and Y. Ozaki, Chem. Phys. Lett., 389 (2004) 225. [8] M. Moskovits, Rev. Mod. Phys., 57 (1985) 783. [9] A. Otto, I. Mrozek, H. Grabhorm and W. J. Akemann, Phys.: Condens. Mater., 4 (1992) 1143. [10] P. C. Anderson, M. L. Jacobson and K. L. Rowlen. J. Phys. Chem. B, 108 (2004) 2148. [11] P. Hildebrant and M. Stockburger. J. Phys. Chem., 88 (1984) 5935. [12] A. Weiss and G. Haran. J. Phys. Chem. B, 105 (2001) 12348. and references therein. [13] J. Lambe and S. L. McCarthy. Phys. Rev. Lett, 37 (1976) 923, J. K. Gimzewski, J. K. Sass, R. R. Schlitter and J. Schott. Europhys. Lett., 8 (1989) 435, M. R. Bevershiis, A. Bouhelier and L. Novotny. Phys. Rev. B, 68 (2003) 115433. [14] T. Itoh, K. Hashimoto, Y. Kikkawa and A. Ikehata, Y. Ozaki, (submitted) [15] K. A. Bosnick, J. Jiang and L. E. Bras, J. Phys. Chem. B, 106 (2002) 8096. [16] B. N. J. Persson and A. Baratoff, Phys. Rev. Lett., 68 (1992) 3224. [17] B. J. Pettinger, Chem. Phys., 85 (1986) 7442. [18] T. Forster, Ann. Phys. 2 (1948) 55, E. Dulkeith, A. C. Morteani, T. Niedereichholz, T. A. Klar, J. Feldmann, S. A. Levi, F. C. J. M. van Veggel, D. N. Reinhoudt, M. Moller and D. I. Gittins, Phys. Rev. Lett., 89 (2002) 2030021. [19] J. Muller, C. Sonnichsen, H. von Poschinger, G. von Plessen, T. A. Klar, J. Feldmann, Appl. Phys. Lett. 81 (2002) 171, H. Kuwata, H. Tamara, K. Esumi, and K. Miyano, Appl. Phys. Lett., 83 (2003) 4625. [20] T. Itoh, K. Hashimoto, Y. Kikkawa and A. Ikehata, Y. Ozaki, (submitted). [21] A. Hie, W. Durkan, W. I. Milne and M. E. Welland, Phys. Rev. B, 66 (2002) 045412. [22] Y. Marayama, M. Ishikawa and M. Futamata, J. Phys. Chem. B, 108 (2004) 673. [23] S. L. Logunov, T. S. Ahmadi, M. A. El-Sayed, J. T. Khoury and R. L. Whetten, J. Phys. Chem., 101 (1997) 3713. [24] C. -W. Kim, J. C. Villagran, U. Even and J. C. Thompson, J. Phys. Chem., 94 (1991) 3974. [25] D. Wang, L. -J. Wan, C. Wang and C. -Li. Bai, J. Phys. Chem. B, 106 (2002) 4223. [26] J. F. Jia, K. Inoue, Y. Hasegawa, W. S. Yang and T. Sakurai, Phys. Rev. B, 58 (1998) 1193. [27] T. Nakano, M. Terakado, F. Kaneko, K. Shinbo, K, Kato, T, Kawakami and T. Wakamatu, Jpn. J. Appl. Phys., 41 (2002) 2774. [28] J. J. Mock, M. Barbie, D. R. Smith, D. A. Schultz and S. J. Sehultz, Chem. Phys., 116 (2002) 6755, C. Sonnichsen, T. Franzl, T. Wilk, G. von Plessen, J. Feldmann, O. Wilson and P. Mulvaney, Phys. Rev. Lett., 88 (2002) 0774021. [29] M. Bader, J. Haase, K. -H. Frank, A. Puschmann and A. Otto, Phys. Rev. Lett., 56 (1986) 1921. [30] C. F. Bohren and D. R. Huffman, Absorption and Scattering of Light by Small Particles. Wiley: New York, 1983; p. 148.; dielectric property of Ag of P. B. Johnson and R. W. Christy, Phys. Rev. B, 6 (1972) 4370.
Nanophotonics, Volume 2 Handai Nanophotonics, S. Kawata and H. Masuhara (Editors) © 2006 Elsevier B.V. All rights reserved. reserved.
219 219
Chapter 11
Single particle spectroscopic study on surface plasmon resonance probing local environmental conditions T. Asahi, T. Uwada, and H. Masuhara Department of Applied Physics and Handai Frontier Research Center, Osaka University, 2-1 Yamada-oka, Suita, Osaka 565-0871, Japan 1. INTRODUCTION Gold nanoparticles exhibit strong optical absorption and scattering extending from the visible to near IR spectral range, which is known as the surface plasmon resonance (SPR) band owing to the collective oscillations of the conduction electrons coupled with incident light [1,2], The peak wavelength and the linewidth of the SPR band depend on the size and shape of the nanoparticle as well as dielectric properties of their local environment including solvent, substrate, and adsorbates [3-17]. Thus, their optical properties have received much attention and systematic studies have been performed for the potential applications to chemical and biological sensors [13-17]. The optical properties of gold nanoparticles were investigated using conventional ensemble measurements until several years ago. Recently, due to progress in optical microscopes and light detection equipments, the studies on single nanoparticles have been made possible, in which scattering light from an isolated particle is detected in a dark-field illumination setup of an optical microscope [17, 18]. Investigations of single nanoparticles yield new insight to understanding of their optical properties because of the following advantages. First, inhomogeneity is suppressed. Even careful preparation of gold nanoparticles provides scatter in the size and shape and also their surface defects and chemical composition. Second, no electronic and optical interactions between particles are involved in experimental data. We can directly compare the experimental SPR spectrum to the calculated one using Mie theory and address the details on the effects of the size, shape, and aggregation on the optical properties. Thus, single nanoparticles can probe local properties and
220
Uwada and H. Masuhara T. Asahi, T. Uwada
chemical environments without considering the above scatters and interparticle interactions, which is indispensable to understand the nature of environmental effect on SPR. There are considerable numbers of reports on SPR responses of gold nanoparticles adsorbing molecules. In most case, adsorbed molecules are limited to large ones such as dye and protein, and a red-shift of the SPR band peak upon molecular adsorption is observed [13-17]. The mechanism was eventually explained by increasing of the local refractive index at or near the surface of the nanoparticle, while molecular adsorption effects on the electronic properties of gold nanoparticles have not been addressed as far as we know. In this paper, we present some details on single particle spectroscopy of gold nanoparticles adsorbing Cl anion, and discuss how the SPR band peak and its spectral shape are affected by environmental solvent and adsorbed anion. 2. EXPERIMENT 2.1 Spectroscopic Setup Fig. 1 illustrates our experimental setup [8, 18]. Collimated unpolarized white light from a 100 W halogen lamp was introduced into an inverted microscope (1X70, Olympus) with an incident angle of 60° to the surface of a sample substrate, on which gold nanoparticles were dispersed. b)
CCD camera
Fig. 1. (a) Schematic illustration of our microspectroscopic system for measuring Rayleigh light scattering spectra of individual nanoparticles and b) a dark-field microscope image of gold nanoparticles dispersed on a glass substrate.
The scattered light from a single bright spot was collected with an objective lens (60*, numerical aperture; 0.7) and led to a polychromator (77480, ORIEL) with an image-intensified charge-coupled device (PS 150, Andor) through an optical fiber. In order to collect selectively the scattered light from one
Single particle spectroscopic study on surface plasmon resonance probing probing local environmental conditions
221
nanoparticle and to reduce the background stray light, we placed a pinhole (300 Hm in diameter) at an image plane of the microscope. The scattering efficiency spectra were obtained by dividing the detected scattered light intensity of gold nanoparticle by that from a frost plate (DFQ1-30C02-240, SIGMA), whose light scattering efficiency is uniform in the spectral range from 400 to 800 nm. We defined the spectrum of the scattering efficiency as the scattering spectrum. 2.2 Sample and measurement procedure For single particle measurements, we used gold nanoparticles with a diameter of 100 ± 20 nm dispersed on a glass substrate covered with immersion oil (the refractive index; 1.52), distilled water, or aqueous solution of CaCl2 (1 M). The sample substrate was prepared by spin-coating a colloidal gold particles (EMGC100, British BioCell) on a glass substrate (Micro Slide Glass, MATSUNAMI) (refractive index; 1.52), then baking it at 250 °C for 2 h in vacuum. To study anion adsorption effect on the SPR spectrum, we measured firstly the scattering spectra of many single gold particles in pure water and marked the position of each nanoparticle on the substrate in its dark-field microscope image. After replacing pure water to a CaCl2 aqueous solution, we measured SPR spectra of the marked nanoparticles, and then compared the spectral shapes in water and in the solution for the same particle. 3. RESULTS and DISCUSSION 3.1 Size and shape dependence of the scattering spectral shape The colloidal sample consisted of gold nanoparticles having somewhat large distribution in their size and shape, which was evaluated by scanning electron microscope (SEM) observation. Figure 2 shows a SEM image of the gold nanoparticles on an ITO coated glass plate. From many SEM images, the distribution of spherical and non-spherical particles was confirmed to be 65 % and 35 %, respectively, and the diameters of spherical ones ranged from 80 to 120 nm. To examine the effect of this scatter in their size and shape on their SPR spectra, we measured the scattering spectra of about 30 gold nanoparticles on a glass substrate covered with immersion oil, and compared them to those calculated according to Mie theory. Three representative scattering spectra are shown in Fig. 3. Here, the gold nanoparticles are considered to be in an optically homogeneous medium, since the oil has the same refractive index to the glass substrate's one (1.52). Hence, we calculated the scattering efficiency of spherical gold nanoparticles in a dielectrically homogenous environment as a function of probe light wavelength, particle diameter, scattering angle, complex refractive index of particle, and the refractive index of surrounding medium (Nenv), using the algorithm described in the literature [2]. Here, we use 1.52 as
222
Uwada and H. Masuhara T. Asahi, T. Uwada
Nenv and the refractive index of bulk gold [19] as complex retractive index of the particle, and vary the diameter as an adjustable parameter.
Fig. 2. A SEM image of the sample gold nanoparticles with the mean diameter of 100 mn.
550
600
650
Wavelength / mn
500
550
600
Wavelength / nm
550
600
650
Wavelength / nm
Fig. 3. The scattering spectra of individual gold nanoparticles on a glass substrate covered with an immersion oil. Smooth curves in each figure show the calculated spectra of gold nanosphere with the diameter of 90, 100, and 110 nm (from bottom to top) embedded in a medium with the refractive index of 1.52.
Smooth curves in Fig. 3 represent calculated scattering spectra of individual particles with diameters of 90, 100, and 110 nm. The shape of the observed spectra in Figs. 3 (a) and (b) was reproduced well by one of the calculated spectra of gold nanospheres with 100 and 90 nm diameters, respectively. On the other hand, no reasonable agreement was found between the observed spectrum in Fig. 3(c) and calculated one as far as the particle diameter is used as the adjustable parameter. This disagreement is considered to be due to the deviation of the particle shape from sphere. Indeed, statistical analysis of the spectral data demonstrated that the spectral shape of 60 % nanoparticles were in good agreement with the spectra calculated by varying the diameter, but 40 % particles were not. The % values are almost the same to those of spherical and non-spherical particles in morphological observation using a SEM. Therefore, it is concluded that the spectral shape of SPR scattering
on surface plasmon resonance probing local Single particle spectroscopic study on environmental conditions
223
of gold nanosphere in an optically homogenous medium can be evaluated quantitatively by the simple calculation on the basis of the Mie theory. 3.2 Scattering spectrum of gold nanoparticle in dielectriccally inhomogeneous environment Fig. 4 shows the scattering spectra of the same single nanoparticle on the glass substrate in air and covered with immersion oil or pure water. This particle diameter was estimated to be 100 nm by comparing the spectrum in the oil with the spectrum calculation as described above. It is obvious that the spectral peak in oil locates at a longer wavelength compared to those in air and water. This peak shift can be ascribed to the increment of the Nmv, since the refractive index (1.52) of the oil is larger than that of water (1.33). The increasing of 2Venv results in a red-shift of SPR band, which can be explained by the increment of effective SPR dipole volume [1,2] .However, the spectrum in water of course did not agree with the calculated spectrum using the refractive index of bulk water. This is because the nanoparticle was on the glass plate whose refractive index is different from water.
1.60
Scattering efficiency / a.u.
Nenv=1.00
500
550 550
600 600
650 650
700 700
750
Wavelength / nm
Fig. 4. The scattering spectra of a single gold nanoparticle on a glass substrate whose environment was changed from air to an immersion oil or water (from left to right). Smooth curves show the SPR spectra of gold nanosphere (110 nm in diameter) calculated by changing the environmental refractive index from 1.0 to 1.6 by 0.1 step.
SPR scattering spectra of gold nanoparticle is strongly affected by refractive of its local environment. The effective refractive index (Neff) should be redefined instead of a bulk refractive index of the environment when a gold nanoparticle is located in dielectrically inhomogeneous environments [8, 9, 11]. Smooth curves in Fig. 3 are the calculated scattering spectra of a 100-nm spherical nanoparticle for several Nenv. The measured spectrum is in good agreement with the simulated spectrum using Nenv of 1.40, which is almost an average of the refractive indices of water and glass substrate. It is noteworthy
224
Uwada and H. Masuhara T. Asahi, T. Uwada
that the SPR scattering spectrum of single gold nanoparticle in a dielectrically inhomogeneous environment can be calculated by introducing the effective refractive index to the simple model where the nanoparticle is in a homogeneous matrix. In other words, it is possible to evaluate a local Neg for the gold nanoparticles in heterogeneous surrounding media such as a solid/liquid interface. The results are important for spectral analysis of molecular adsorption effects on SPR response, and also for the applications to chemical and biological sensors, because gold nanoparticles fixed on a glass substrate were widely used in many of such sensors.
a)
450
500
550
600
650
Wavelength / nm
700
750
Scattering efficiency (normarilzed)
Scattering efficiency (normalized)
3.3 Anion adsorption effect As a representative example, the scattering spectra of an individual gold nanoparticle (100 nm in diameter) in pure water and CaC^ aqueous solution (1 M) are shown in Fig. 5. The peak wavelength of SPR scattering in the solution shifted to a long wavelength by 11 nm from that in pure water, and the shift was reversible upon a cyclic change of pure water and the solution. The peak shift of many single nanoparticles is summarized in Fig. 6 as a function of the peak of SPR band in pure water. The shift value seems not depend clearly on the peak wavelength, although it scatters from particle to particle. We estimate a mean value of the shift to 11.5 nm, which is much larger than that expected from the difference between the bulk refractive indices of water and 1 M CaCl2 solution [20] (see table 1). Anion adsorption at a metal/liquid interface has been a widely studied subject of applied electrochemistry. It is well established that Cl anion is adsorbed on a metallic gold surface, and the surface density of the adsorbed Cl anion on the polycrystalline gold electrode was evaluated to be an order of 1014 atoms/cm2, depending on the voltage of the electrode [21, 22]. A large scatter in the peak shift shown in Fig. 6 will probably arise from a distribution in the number of Cl anion adsorption sites of nanoparticles.
b)
560
580
600
620
640
660
680
Wavelength / nm
Fig. 5. The SPR scattering spectra of a gold nanoparticle (100 nm) on a glass substrate whose environment was changed from pure water to aqueous solution of 1 M CaCl2 (see text).
Peak shift / nm
Single particle spectroscopic study on surface plasmon resonance probing probing local environmental conditions
22 20 18 16 14 12 10 8 6 4 2 0
580
590
600
22 225
610
Peak wavelength (in pure water) / nm
Fig. 6, The peak shift of the SPR scattering spectra of individual gold nanoparticles adsorbing Cl anion is plotted against the peak of corresponding nanoparticles in water. Table 1 The averaged value of the shift in SPR scattering peak of gold nanoparticles on glass substrate for changing the covered solution from pure water to CaClj (1M) solution, and the calculated peak shift by taking account of the effective refractive index of each medium, and the bulk refractive indices of water and 1 M CaC^ solution Peak shift Peak shi Refractive index8' (experimental) (calculated) Pure water 1.33 CaC12 aq. solution (1 M) 11.5 nm 3.8 nm 1.36 a) from ref. 20
An idea to explain the SPR peak shift upon chloride anion adsorption is formation of an electric double layer (with a few-nm thickness) at the interface of the particle surface and the solution. The refractive index of the double layer will be larger than that of the bulk solution because of higher local concentration of counter cation (Ca*) and of higher polarization of solvent in the double layer. We have tried to simulate the spectral shift by modelling double layer formation at the interface as a spherical gold core (100 nm in diameter) coated with a dielectric shell (1 to 10 nm in thickness) as shown Fig. 7(a). The scattering spectra of the coated spherical nanoparticles in a dielectric homogenous environment as a function of the shell thickness and its refractive index were calculated using the algorithm described in the literature [2]. We used 1.40 as Nenv and changed the refractive index of the shell from 1.4 to 1.8. The simulated peak shift illustrated in Fig. 7(b) show that the shift value increases with the thickness and the refractive index. When a 10-nm thickness shell with the refractive index of 1.48 is assumed as example, the observed peak can be explained qualitatively. It is noteworthy, however, that the thickness of the
226
Uwada and H. Masuhara T. Asahi, T. Uwada
electric double layer is less than 1 nm for a metal/solution interface in a high concentration electrolysis solution [23], In the case of the 1-nm thickness shell, a very large value of its refractive index (about 2.5), which is comparable to that of solid CaCl2, is necessary to explain the experimental result. Therefore, we consider that the SPR peak shift can not be ascribed to the formation of the electronic double layer.
35
shell ; (b-a)
G
gold core ttntttn incident light I I
30
Peak shift / nm
a)
1
1
b)
1
'
• •
lnm 1nm 5nm 5nm A 10nm
25
yS
^/ -
20 20 • 15 15 10 10 5 0
__ 1.4 1.4
1.5 1.5
^
#
1.6 1.6
1.7 1.7
1.8 1.8
Refractive index of the shell Refractive
Fig. 7. (a) A schematic illustration of a core-shell model in SPR calculation using Mie theory and (b) the effect of the layer refractive index on the peak shift in the SPR scattering spectra which was calculated for a gold nanoparticle (100 nm) coated with a dielectric shell having the thickness of 1,5, and 10 nm.
The above results demonstrate clearly that the refractive index change of the local environment is not enough for explaining the shift in the SPR peak upon Cl anion absorption. It is necessary to take into account the anion adsorption effect on the dielectric property of gold nanoparticle. It is known that addition and removal of a number of electrons from nanoparticles produces a shift in SPR band position. This shift in SPR band peak position due to the change in electron density was demonstrated for colloidal gold nanoparticles (10 nm order in diameter) [24] and silver nanoparticles (10 nm in diameter) in a polymer film [25] by controlling their charge with elctrochemical techniques. The SPR peak of the gold colloid shifts to a longer wavelength as the number of free electrons in the nanoparticles decreases. We calculated the peak shift for a 100 nm gold nanoparticles using Mie theory, and obtained 5 nm as the shift value for reduction of 1 % free electrons in the nanoparticles. This change in the free electrons seems too large as the effect of anion adsorption on the large nanoparticle. It would be required to consider local and specific effects on the dielectrically properties of gold nanoparticle; for example, the negative charge on the surface may scatter the free electrons in the nanoparticle, and will change the electron density near the particle surface.
on surface plasmon resonance probing Single particle spectroscopic study on probing local environmental conditions
227
4, CONCLUSION We have presented some advances in single nanopartiele spectroscopy of gold nanoparticles, and discussed its importance for investigating nature of SPR band and surface environmental effect on it. Not only the size and shape but also local environmental conditions are very critical for assigning SPR bands. By comparing the measured spectra with those calculated using Mie theory, we evaluated the effective local refractive index for a single gold nanopartiele in a heterogeneous surrounding medium such as solid/liquid interface. Also, the effect of adsorbed Cl anion on SPR provides a new possible insight that the dielectric property of gold nanopartiele is modified by the adsorbed ions.
REFERENCES [I] [2] [3] [4] [5] [6] [7] [8] [9] [10] [II] [12] [13] [14] [15] [16] [17] [18] [19] [20] [21]
U. Kreibig and M. Volmer, Optical Properties of Metal Clusters, Springer Series in Material Science Vol. 25, Springer-Verlag, Berlin, 1995. C. F, Bohren and D. R. Huffman, Absorption and Scattering of Light by Small Particles, Wiley, New York, 1983. C. L. Haynes and R. P. Van Duyne, J. Phys. Chem. B, 105 (2001) 5599. R. Elghanian, J. J. Stochoff, R. C. Muck, R. L. Letsinger, and C. A. Mirkin, Science, 277 (1997) 1078. C. A. Mirkin, R. L. Letsinger, R. C. Mucicm, and J. J. Stochoff, Nature, 382 (1996) 607. J. Zhu, F. Xu, S. J. Schofer, and C. A. Mirkin, J. Am. Chem. Soc, 119 (1996) 235. H. Sato, K. Uehara, T. Ishii, and Y. Ozaki, Photochem. Photobiol., 62 (1995) 509. T. Itoh, T. Asahi, and H. Masuhara, Appl. Phys. Lett., 79 (2001) 1667. T. Itoh, T. Asahi, and H. Masuhara, Jpn. J. Appl. Phys., 41 (2001) L76. S. Underwood and P. Mulvaney, Langmuir, 10 (1994) 3427. C. Sonnichsen, S. Geier, N.E. Hecker, G. von Plessen, J. Feldmann, H. Ditlbacher, B. Lamprecht, J.R. Krenn, F.R. Aussenegg, V.Z. Chan, J.P. Spatz, and M. Moller, Appl. Phys. Lett., 77 (2000) 2949. S. Schultz, D. R. Smith, J. J. Mock, and D. A. Schultz, Proc. Natl. Acad. Sci. U.S.A., 97 (2000) 996. S. Franzen, J. C. W. Folmer, W. R. Glomm, and R. O'Neal, J. Phys. Chem. A, 106 (2002) 6533. C. R. Yonzon, E. Jeoung, S. Zou, G. C. Senate, M. Mrksich, and R. P. Van Duyne, J. Am. Chem. Soc, 126 (2004) 12669. N. Nath and A. Chilkoti, Anal. Chem., 74 (2002) 504. G. Raschke, S. Kowarik, T. Franzl, C. Sonnichsen, T. A. Klar, J. Feldmann, A. Nichtl and K. Kulrzinger, Nano Lett., 3 (2003) 935. M. A. Van Dijk, M. Lippitz, and M. Orrit, Ace. Chem. Res., 38 (2005) 594. T. Asahi and H. Mashuara in Single Organic Nanoparticles, H. Masuhara et al (eds.), Springer, 2003, pp. 94-108 P. B. Johnson and R. W. Christy, Phys. Rev. B, 6 (1972) 4370. D. R. Lide (ed.), Handbook of Chemistry and Physics 74th Edision, CRC Press, Florida, 1998. T. Pajkossy, T. Wandlowsk, and D. M. Kolb, J. of Electroanal. Chem., 414 (1996) 209.
228
T. Asahi, T. Uwada Uwada and H. Masuhara T.
[22] M. J, Walters, C. M, Pettit, F. X. Bock, D. P. Biss, and D. Roy, Surf. Interface Anal. 27 (1999) 1027. [23] B. V. Enustun and J. Turkevich, J. Am. Chem. Soc., 3317 (1963) 85. [24] M. J. Walters, C. M. Pettit, F. X. Bock, D. P. Biss, and D. Roy, Surf. Interface Anal. 27 (1999) 1027. [25] T. Pajkossy, T. Wandlowski, and D. M. Kolb, J. Electroanal. Chem., 414 (1996) 209.
PART III: MATERIALS AND DEVICES FOR NANOPLASMONICS
This Page is Intentionally Left Blank
Handai Nanophotonics, Nanophotonics, Volume 2 (Editors) S. Kawata and H. Masuhara (Editors) © 2006 Elsevier B.V. All All rights rights reserved. reserved.
231 231
Chapter 12
Enhancement of luminescence in plasmonic crystal devices T. Okamoto*, F. H'DMIf, J. Feng", J. Simonenab, and S. Kawata**11 a
RIKEN, Nanophotonics Laboratory, Hirosawa 2-1, Wako, Saitama 351-0198, Japan ''Department of Physics, University of Joensuu, P.O. Box 111, 80101 Joensuu, Finland department of Applied Physics, Osaka University, Siiita, Osaka 565-0871, Japan 1. INTRODUCTION When light is incident upon a metal surface, an unusually strong electromagnetic field may form on the surface. It is in the state where surface plasmons are excited on the metal surface. The enhanced electromagnetic field results from the long propagation length of the surface plasmon, and it enhances fluorescence of dye molecules near the metal surface [1]. Moreover, the electric field can be further enhanced by modulating the metal surface profile with a random corrugation [2] or with a 2-D relief grating [3-5]. Indeed, a 2-D grating surface carries out Bragg reflection of the surface plasmon, and propagation is forbidden in all in-plane directions in certain frequency regions. We call such a structure and such a forbidden frequency range a plasmonic crystal and a plasmonic band gap, respectively. At the edge frequency of the plasmonic band gap, the surface plasmons serve as standing waves, therefore, photons are confined in three dimensions resulting in a strong enhancement. In this paper, we demonstrate band gap behavior of surface plasmons propagating on a periodically corrugated metal surface. We report on the observation of fluorescence enhancement of a dye film deposited onto such a surface. This strong fluorescence may yield lasing action that involves a loss coefficient of the surface plasmon lower than the gain coefficient of a gain medium (dye molecules). We discuss how one can accomplish this requirement. We also describe enhancement of electroluminescence of organic light emitting diodes (OLEDs) using plasmonic crystals.
232
T. Okamoto et al.
2. PLASMONIC CRYSTALS Three-dimensional photonic crystals, which are composed of structures with a periodic modulation of the refractive index, prohibit light propagation in some frequency domains called photonic band gaps [7]. Since it is very difficult to fabricate 3-D photonic crystals, the focus of research has been recently moved towards 2-D photonic crystals with the use of a core-clad structured waveguide, which confines photon energy in its normal direction [8]. The plasmonic crystal that we propose here can be regarded as a two dimensional photonic crystal in which photon confinement in the normal direction is performed by surface plasmon polariton instead of waveguide modes. Figure 1 shows the dispersion relation of surface plasmon at the interface between a metal and a dielectric medium represented by (1)
lin
e
where ksp is the wave vector of the surface plasmon, cois the angular frequency, and £^, and e& are the dielectric constants of the metal and the dielectric, respectively. If the interface is smooth, the dispersion relation has no discontinuity. However, if a periodic corrugation is introduced into the interface, the dispersion relation shows band structure and energy gaps. The energy width of the gap depends on the amplitude of the introduced corrugation.
Lig
ht
Surface plasmon
bandgap: Δ Aft) Smooth metal surface bandgap:
0
Λ • |
71/A π/Λ
k
plasmon Surface plasmon
Corrugated metal surface
Fig. 1. Dispersion relation of surface plasmon traveling on a flat metal surface and a corrugated metal surface.
Kitson et al. [4] have confirmed that a full energy band gap opens when a two dimensional corrugation is introduced onto metal surfaces. A full band gap
Enhancement of luminescence in plasmonic crystal devices
233
means that a gap opens in all in-plane directions, so that in this energy region no surface plasmon can propagate. At the edges of the band gap, called band edges, the group velocity of surface plasmon, which is given by the slope of the dispersion relation, tends to zero and surface plasmons behave as standing waves. Therefore, the density of photons becomes extremely high at the band edges. The plasmonic crystal used here was made as follows: first, we fabricated a two dimensional corrugated substrate on which we then deposited a silver film by thermal evaporation. The patterned surface was made using a holographic system as shown in Fig. 2. It consists of a collimated He-Cd laser beam (325-nm wavelength) that is divided into two beams by a beam splitter. These beams are superposed on each other to form interference fringes onto the photoresist layer spun on a glass substrate. To obtain a hexagonal grating we rotated the substrate by 60 degree around its normal axis followed by a second exposure [6]. Finally, the photoresist was developed and dried and the topography was analyzed. Figure 3(a) shows a scanning electron micrograph of a typical fabricated 2-D grating using Shipley photoresist S1400. Figure 3(b) is a low magnification image, which shows that we obtained a very homogeneous two-dimensional corrugation in a wide area.
a
He-Cd laser (325 nm, 50 mW) 60° rotation Sample
θ Optical shutter
Beam splitter Beam splitter
f expander Beam Beam expander
Fig. 2. Holographic exposure system for 2-D corrugated surface fabrication.
234
T. Okamoto et al.
Fig. 3. Scanning electron micrographs of a 2-D corrugated photoresist surface fabricated by a holographic exposure system as shown in Fig. 2, (a) high resolution image and (b) low magnification image.
Figure 4 shows an atomic force microscope (AFM) image of an example of a 2-D grating using another photoresist THMR-iP33OO (Tokyo Ohka). Because of its high resolution, the obtained grating has a large modulation depth. The AFM image also shows a local defect. This defect has been randomly obtained, probably during the drying process using a dry nitrogen gun on the grating after development. Note that a defect in a plasmonic crystal is also a candidate for photon confinement and field enhancement.
Fig. 4. Scanning atomic force micrograph of an example of 2-D corrugated photoresist surface with a defect.
235
Enhancement of luminescence in plasmonic plasmonic crystal devices
3. FLUORESCENCE ENHANCEMENT We investigate fluorescence enhancement using the expected field enhancement related to the surface plasmon confinement close to the plasmonic crystal surface. In this experiment we deposited a 150-nm silver thin film onto the developed photoresist (Shipley, SHOO) directly by thermal evaporation to obtain plasmonic crystals. We first observed the fluorescence of 4-dicyanomethylene-2-methyl6-p-dimethyl-aminostyryl-4H-pyran (DCM), which is a well-known laser dye. The DCM film was evaporated onto a plasmonic crystal with a 550-nm grating pitch and a 150-nm thick silver film. The thickness of DCM was ~50 nm, which does not support waveguide modes. A layout of the complete device is displayed in the inset of Fig. 5(a). We used an incident laser beam at 532 nm (the second harmonic of a YAG laser) to excite dye molecules which in turn excite surface plasmons at the interface by both grating coupling and near-field coupling. Then, the surface plasmons radiate by the grating coupling mechanism.
• for 1-D and 2-D plasmonic crystal only for 2-D plasmonic crystal 550
600
550
700
750
Wavelength (nm)
BOO
850 Wave number (2n/A)
Fig. 5. (a) Fluorescence spectra of DCM films deposited onto a planar silver surface and a plasmonic crystal, (b) Dispersion relation of the obtained fluorescence peaks.
Figure 5 (a) shows the fluorescence spectra observed from the direction normal to the surface. The fluorescence intensity of DCM on the plasmonic crystal was enhanced by a factor of three compared with that on a planar silver surface. The peak wavelength corresponds to the lower band gap edge of the plasmonic crystal, while the shoulder appearing left of the peak corresponds to the upper band gap edge. This fact was confirmed by observing the angular dependence of fluorescence spectrum. Figure 5(b) shows the dispersion relation of both 1-D and 2-D plasmonic crystals, which was obtained by plotting the fluorescence peak energy as a function of its wave number. The dispersion relation clearly exhibits a plasmonic band gap. Fluorescence from the 1-D plasmonic crystal was strongly polarized perpendicular to the grating grooves,
236 236
T. T. Okamoto Okamoto et et al. al.
confirming that the enhancement is not caused by waveguide modes but by surface plasmons. A film of methyl-red doped PMMA was spun onto a 507-nm pitch plasmonic crystal from its chloroform solution. The resultant thickness was ~25 nm, which also does not support waveguide modes. Figure 6 shows the obtained fluorescence spectrum of methyl red observed from the direction normal to the surface. For comparison, we also measured the fluorescence spectra of the same film coated on a planar silver surface and on a bare glass substrate. Clearly, the fluorescence intensity in the case of plasmonic crystal is much higher than those corresponding to the planar silver and the bare glass substrate. The obtained enhancement factor is 150 at wavelength of 560 nm in comparison with that on bare glass. Note that the increase of the fluorescence intensity from the dye film on the planar silver surface compared with bare glass surface is caused by surface roughness, which exists even on the planar silver surface. The obtained enhancement factor for methyl-red doped PMMA is much higher than that for DCM. We attribute this result to the quenching process. In fact, it has been reported that for molecules with a very low fluorescence quantum yield, the excited energy is transferred to a surface plasmon and then radiates before nonradiative damping takes place in the molecule. Consequently, the fluorescence is much enhanced [12]. The enhanced fluorescence spectrum shown in Fig. 6 has more peaks than that in Fig. 5. This might be caused by additional waveguide modes excited in the PMMA spun film. The thickness of the PMMA film of ~25 nm was measured from a flat region. However, in the grating region the thickness might be larger, resulting in supported waveguide modes. 1600 1600 Intensity (arb. units)
1400 1400
S" 1200 1200
on plasmonic plasmonic crystal crystal methyl methyl red red
= 1000 1000 -9 -2- 800 »
600 600
c
400 400 200 0 500
silver
.» X i i photoresist photoresist
-I 25 25 nm nm
U ^150 O nm
glass substrate substrate on planar silver yon silver' ' I on on bare bare glass glass^ 550 600 650 Wavelength (nm)
700
Fig. 6. Fluorescence spectra of methyl red films deposited onto a planar silver surface and a plasmonic crystal.
Enhancement of luminescence in plasmonic plasmonic crystal devices
237
4. PLASMONIC BAND GAP ENGINEERING FOR LASING One of our targets is the creation of laser using a plasmonic crystal. For lasing action, the loss coefficient must be lower than the gain coefficient, so that, in addition to increasing the gain of the dye film, we have to decrease the loss due to the plasmonic crystal. We will discuss how one can satisfy these requirements. 4.1 Optimization of Gain Medium Before we discuss loss of plasmonic crystals, we first discuss a gain medium giving high gain coefficient. It is clear that the evaporated DCM film used in the experiment above is not the appropriate dye material for lasing because the quantum efficiency of DCM molecules decreases as its concentration increases, due to concentration quenching. One candidate for the laser material is DCM doped tris-(8-hydroxyquinoline) aluminum (Alq3), which can be also deposited by co-evaporation. The reported gain coefficient of DCM doped Alq3 is more than 600 cm"1 [13, 14]. The excitation energy absorbed by the host Alq3 molecules is nonradiatively transferred to the guest DCM molecules. The Alo^ emission spectrum matches the DCM absorption spectrum and results in an efficient Forster energy transfer between the two molecules, so that only a few percent of DCM molecules is required for a high quantum efficiency. This also prevents the concentration quenching of DCM. 4.2Plasmonic Band Structure on Semi-infinite Metal There are two kinds of loss in the propagation of a surface plasmon. One is radiation loss and the other is absorption loss. First we discuss the radiation loss. Figure 7 shows the band structure of a plasmonic crystal, in which two band gaps are drawn. The first band gap, in which the energy is the lowest, lies below the light line. Thus, surface plasmons at the first band gap edges do not couple to free-space radiation, i.e. no radiation loss. However, if we employ these band gap edges for lasing, we cannot extract the laser power from the device. Therefore, the first band gap edges are not appropriate for lasing. Next we consider the second lowest band gap. Because this band gap is above the light line, usually the surface plasmons radiate, resulting in large loss. However, we can reduce the radiation loss at the band gap edges. The dispersion relation of the surface plasmons in photonic crystals highly depends on the surface profile of the metal. We theoretically analyzed the behavior of the dispersion using numerical calculations. We used Fourier modal method, which is equivalent to the rigorous coupled wave analysis (RCWA). In this paper we calculated only one-dimensional gratings. To obtain the dispersion relation of surface plasmon, first we calculated reflectance or absorbance for incident plane waves, then plotted them as a two dimensional function of frequency and angle of incidence.
238
T. Okamoto et al.
CD power distribution distribution E power
ikikkkkkkkil
LAAAAAA
radiative region
plasmonic plasmonic crystal crystal
2+ ^2+
E power distribution
}2–
WA A A A A
"-
1+
plasmonic crystal
gh
tl
in e
1–
Li
Vy!
ionradiative nonradiative region
kx 71/A π/Λ
Fig. 7. Plasmonic crystal band structure and the electric power distribution corresponding to the band gap edges.
Figure 8 shows four examples of obtained dispersion relations for different grating profiles, which exhibit different band structures. The dispersion relation of the surface plasmons is expressed by reflectance minima. The surface profile h(x) can be developed by Fourier series and presented by the sum of harmonic components as h{x) = h0 + hj cos Kx + h2
cos(2Kx+A0)+•
(2)
where /*, is the amplitude of the z-th harmonic component, K is the grating wave vector, and A0is the phase between the K and 2K harmonic components. In Fig. 8 we listed possible grating geometries on a semi-infinite silver layer. These gratings have different amplitude of the second harmonic component hi and a different phase difference A0. We note that the K component of the surface modulation is used for the photon coupling to surface plasmon. Let us explain the requirement to have a plasmonic band gap. To this end, we only treat the second lowest band gap as mentioned before. When surface plasmons are generated, they propagate in both forward (x) and backward (-x) directions. To open a band gap these surface plasmons should interact with each other giving a difference in the wave vector equal to 2ksp, so that surface plasmons can gain or lose momentum in integer multiples of 2K. Thus, it is important that the surface corrugation contains a second harmonic component. This explains the fact that the profile shown in Fig. 8(b) does not produce any gap while those shown in Figs. 8(a), (c), and (d) provide clear energy gaps.
239
Enhancement of luminescence in plasmonic crystal devices
Angle (deg.)
1.5
0
Angle (deg.)
A A AA No 2K component
A<|> = 71 / 3
1.5
0
Angle (deg.)
1.5
0
Angle (deg.>
1.5
AAMWW
Fig. 8. Surface plasmon dispersion curves with the associated plasmonic crystal geometries and their K and 2K harmonic components, respectively from the top to the bottom.
Next we discuss the coupling between surface plasmons and free-space radiation. In the grating shown in Fig. 8(c) the upper mode disappears around normal incidence angle, while the lower mode disappears in the grating shown in Fig. 8(d). In other words, the surface plasmon does not radiate in these regions, resulting in no radiation loss. In the grating in which the width of the groove is narrower than that of ridge (Fig. 9(a)), the phase difference between the fundamental component and the second harmonic component is A0= n. In order to excite surface plasmon, electric field component normal to the surface is required. If the surface is flat, no normal components exist. However, when the surface is corrugated, a normal field component arises. Usually, the amplitude of the second harmonic component is much smaller than the fundamental one, so that the electric field normal to the surface of the fundamental component excites surface plasmons. At the second lowest plasmonic band gap edges, where the surface plasmons exist as standing waves, there are two standing waves of surface plasmons, one has field intensity maximum at the peak of the second harmonic component of corrugation correspond to ca. as shown in Fig. 9(a) and the other has that at the 9(b). They have different energies (frequencies), the frequency of the former is lower than the later.
240
T. Okamoto et al.
K component
2K component
CO.
Ac{) = 0
Fig. 9. The relation between grating profiles and surface plasmon modes.
In order to excite these modes at the normal incidence, the incident electric field component normal to the fundamental surface component must match the field distribution of them. Therefore, in the grating shown in Fig. 9(a) only the oi mode can be excited by normal incident waves, while in the grating shown in Fig. 9(b) only the ca, mode can be excited. Here we consider which mode should be used for lasing; the upper band gap edge (fflO or the lower band gap edge (ai). In the Qh mode the energy of plasmon, i.e. the photon, is localized in the dielectric region, while that in the (O. mode is localized in the metal region [15]. Thus, it is better to use the a\ mode to minimize the absorption losses by the metal. 4.3 Reduction of Absorption Loss In the following we propose the use of a thin metallic film in order to reduce the absorption loss coefficient. In fact, it is found that when a metal film is thin enough and is sandwiched with dielectric media of identical refractive indices, surface plasmons at both metal-dielectric interfaces are coupled with each other and split into two modes, the long-range surface plasmon (LRSP) and the short range surface plasmon (SRSP) [16]. LRSP mode has a smaller fraction of its field inside the metal than that SRSP has, so that the loss of surface plasmon due to the dissipation of power inside the metal is lower. In this case, the excited surface plasmons can achieve propagation length that exceeds one order of magnitude larger than standard modes (non-coupled surface plasmons). Figure 10 shows the calculated loss coefficients of both modes supported by a thin silver film embedded in a dielectric medium of refractive index 1.7, which corresponds to that of Alq3 at 620-nm wavelength. We observe a clear decrease of the loss coefficient of LRSP as the thickness of the metal film decreases.
Enhancement of luminescence in plasmonic plasmonic crystal devices
241
3000 dielectric netal film
1080 cm'
LRSP (asymmetric mode): or
650 cm" Gain region 0
50
100
150
Thickness of silver (nm)
200
SRSP (sym
Fig. 10. Loss coefficient of the long range surface plasmon and short range surface plasmon as a function of the thickness of the metal film.
When the thickness of the silver film is semi-infinite, the loss coefficient is as large as 1080 cm"1. However, when the film thickness is as thin as 20 nm, the loss coefficient of the LRSP mode (of mode) dramatically decreases to 71 cm'1. If we employ the gain coefficient of 650 cm"1 corresponding to Alq3: DCM gain medium [13], we can expect lasing for the silver film thinner than 60 nm, where the gain coefficient is larger than the loss coefficient. 4.4Plasmonic band structure in metallic films In order to reduce the absorption loss, we have to employ long-range surface piasmons as described above. Here we investigate the behavior of plasmonic band gaps in thin metallic films. Figure 11 shows the dispersion relation of surface piasmons in thin metallic films whose surfaces are corrugated in various profiles. Upper two branches correspond to LRSPs, while lower two branches correspond to SRSPs. In Figs. 1 l(a) and (b) the dispersion relations show no band gaps at kx = 0. However, there is a gap between LRSP and SRSP at around 1.5 degree in both cases. We call these gaps intermode gaps. On the other hand, in the film structure shown in Fig. ll(c) gaps open at kx = 0. We call this gap intramode gap. The position of the band gaps depends on the thickness distribution of the metallic films. The structures shown in Figs. 11 (a) and (b) have no thickness modulation, while the structure shown in Fig. 1 l(c) has thickness modulation. After further calculations we conclude that, in the thin metallic film cases, opening gaps at kx = 0 requires a second harmonic component in the film thickness modulation for small amplitude gratings (<10 nm).
242
T. Okamoto et al.
590
Fig. 11. Dispersion relation of surface plasmons supported by tfain metal films.
In the film structures shown in Fig. 11, all the dispersion curves appear at all angles. This indicates that both LRSPs and SRSPs couple to free-space radiation, thus these films have a large radiation loss. Coupling between surface plasmons and free-space radiation shows the same behavior as in the case of semi-infinite metals. Additionally, in the film cases, one has to consider the film thickness modulation, which must have a second harmonic component necessary to open band gaps. 590
Oblique vacuum vapor deposition
Fig. 12. Band structure of obliquely deposited thin metal film.
Figure 12 shows our proposed film structure to suppress the radiation loss, which can be easily fabricated by oblique vacuum vapor deposition as shown in the same figure. This structure has the second harmonic component in the film
Enhancement of luminescence in plasmonic crystal devices
243
thickness distribution and the surface profile has also the second harmonic component whose phase difference from the fundamental component is zero. In this structure both upper branches of LRSPs and SRSPs vanish at normal incidence. 5. ORGANIC LIGHT EMITTING DIODES WITH PLASMONIC CRYSTALS We investigated the effect of plasmonic crystals on the increase of the external quantum efficiency of organic light emitting diodes. OLEDs based on organic thin films have attracted a great deal of attention for many applications. In conventional OLEDs photons generated in emissive layer are trapped in the layers and cannot be extracted into free space, because the refractive index of the emissive layer is relatively high [17]. Thereby more than 80% of the produced photons are trapped in guided modes within the device (Fig. 13 (a)), leading to a significant reduction in the external quantum efficiency. Much effort has been made to improve their external quantum efficiency. In this paper we demonstrate that the emission can be much increased by extracting light from OLEDs by using a plasmonic crystal (Fig. 13(b)) [18].
(a)
Surlace plasmon Waveguide mode \ Silver Alq3/NPB;iTO n = 1.7-2.0 Quartz substrate n = 1.46
Fig. 13. (a) Typical conventional OLED with plane interfaces, in which generated photons are trapped as waveguide modes and surface plasmons, which cannot radiate, (b) OLED with a plasmonic crystal, in which both substrate side emission and topside emission (through the silver cathode) increase very much. Photographs of the emission regions are also shown.
244
T. Okamoto et al.
(a)
silver (cathode)
y i f 50 nm
A!q 3 (emitting layer) NPB (hole trans, layer) ITO (anode)
, 80 nm 100 nm 150 nm
silica substrate
Fig. 14 (a) The structure of the fabricated OLED with a plasmonic crystal, (b) AFM image of a ID and a 2D device.
The corrugation was introduced into a quartz substrate. First we fabricated 1-D and 2-D corrugation in photoresist (Tokyo Ohka, THMR-iP3300) using the same method as described in section 2. After development, the quartz substrate was etched by reactive ion etching, and then the photoresist was removed. The obtained groove depth and the grating pitch were ~70 nm and ~550 nm, respectively for both 1-D and 2-D gratings. The substrate was then coated with a 150-nm thick indium-tin-oxide (ITO) layer by DC magnetron sputtering. This was followed by vacuum evaporating successively an 80-nm thick N,N'-diphenyl-N,N'-bis(l-naphthyl)-(l»r-biphenyl)-4,4'-diamine (NPB) layer, a 100-nm thick Alq3 layer, and a 50-nm thick silver layer. A complete OLED is schematically displayed in Fig. 14(a). The ITO and silver layers act as an anode and a cathode, respectively, while the NPB and Alq3 layers act as hole-transporting and emitting layer, respectively. AFM images of the top surface of the 1-D and 2-D corrugated device are given in Fig. 14(b) showing a similar topography as that of quartz substrate. Figure 15 shows the electroluminescence (EL) spectra observed at normal direction of flat, 1-D, and 2-D corrugated devices driven by the same applied voltage (12V). The emission from quartz side of the 2-D corrugated device increases by a factor of four in comparison with that from the flat device, because the trapped guided modes in the active layers are effectively coupled out into free-space radiation by the plasmonic crystal. The silver side emission intensity of the 2-D corrugated device is four times higher than that of the 1-D corrugated device, while a very weak emission was observed from silver side of the flat device as shown by the corresponding photograph in Fig. 13(a). Since the thickness of the silver layer was 50 nm, the light from the active layer does not
in plasmonic crystal devices Enhancement of luminescence in
245
EL intensity (arb. units)
transmit the silver layer. However, when the wave vector of the surface plasmon matches with that of the other side of silver layer via the grating vector, photons pass through the silver layer and then are coupled out into free space. Thereby, the external quantum efficiency of OLEDs with a plasmonic crystal increases also for the silver side. 2
-SS.
>, 11 to
2-D quartz side 1-D quartz side 2-D silver side 1-D silver side Flat
CD
LU
0 400
500 600 700 (nm) Wavelength (nm)
800
Fig. 15. Electroluminescence (EL) spectra observed from normal direction of 1-D, 2-D corrugated and flat OLEDs.
SUMMARY We fabricated plasmonic crystals by a holographic technique and observed enhanced narrow fluorescence peaks from a DCM film deposited on them. We discussed how we could realize lasing from dye films on plasmonic crystals. We proposed that nonradiative modes for low radiation loss, long-range surface-plasmons for low absorption loss, and DCM doped Alq3 for a high gain medium should be used. We also fabricated organic light emitting diodes with use of plasmonic crystals, which increase the external quantum efficiency by a factor of four compared with conventional flat devices. The OLEDs also showed high efficiency of the emission through the silver cathode, which was very weakly observed from flat devices. We believe plasmonic crystals will surely contribute to the development of organic current injection laser, which is our goal.
REFERENCES [I] T. Liebermann and W. Knoll, Colloids Surf. A, 171 (2000) 115. [2] V. M. Shalaev ed. Optical Properties of Nanostructured Random Media, Springer, Berlin, 2002. [3] T. W. Ebbesen, H. J. Lezec, H. F. Ghaemi, T. Thio, and P. A. Wolff, Nature, 391 (1998) 667. [4] S. C. Kitson, W. L. Barnes, and J. R. Sambles, Phys. Rev. Lett., 77 (1996) 2670. [5] W. L. Barnes, S. C. Kitson, T. W. Preist, and J. R. Sambles, J. Opt. Soc. Am. A, 14 (1997) 1654.
246
T. Okamoto et al.
[6] S. C. Kitson, W. L. Barnes, and J. R. Sambles, IEEE Photonics Technol. Lett., 8 (1996) 1662. [7] J. D. Joannopoulos, R. D. Meade, J. N. Winn, Photonic Crystals, Princeton University Press, Princeton, 1995. [8] S. Nie and S. R. Emory, Science, 275 (1997) 1102. [9] H. Xu, E. J. Bjerneld, M. Kail, and L. Borjesson, Phys. Rev. Lett., 83 (1999) 4357. [10] T. Okamoto and I. Yamaguchi, J. Phys. Chem. B, 107 (2003) 10321. [11] S. C. Kitson, W. L. Barnes, and J. R. Sambles; IEEE Photon. Technol. Lett., 8 (1996) 1662. [12] J. R. Lakowicz, Anal. Biochem., 298 (2001) 1. [13] V. Bulovic, V. G. Kozlov, V. B. Khalfin, and S. R. Forrest, Science, 279 (1998) 553. [14] S. Riechel, U. Lemmer, J. Feldmann, S. Berleb, A. G. Miickl, and W. Bruiting, A. Gombert, and V. Wittwer, Opt. Lett., 26 (2001) 593. [15] W. L. Barnes, T. W. Preist, S. C. Kitson, J. R. Sambles, N. P. K. Cotter, and D. J. Nash, Phys. Rev. B, 51 (1995) 11164. [16] D. Sarid, Phys. Rev. Lett. 47 (1981) 1927. [17] N. K. Patel, S. Cina, and J. H. Burroughes, IEEE J. Sel. Top. Quant. 8 (2002) 346. [18] J. Feng, T. Okamoto, and S. Kawata, Opt. Lett., 30 (2005) 2302.
Handai Nanophotonics, Volume 2 (Editors) S. Kawata and H. Masuhara (Editors) © 2006 Elsevier B.V. All All rights rights reserved. reserved.
247 247
Chapter 13
Intrinsic properties due to self-organization of 5nm silver nanocrystals M. P. Pileni Laboratoire LM2N, URA CNRS 7070, Universite P. et M. Curie (Paris VI), BP 52,4 place Jussieu, 75252 Paris cedex 05, France.
1. INTRODUCTION During the last decade, due to the emergence of a new generation of high technology materials, the number of groups involved in nanomaterials increases exponentially, Mesoscopic structures of nanocrystals are nowadays a rapidly growing field of science where the efforts of chemists, physicists, material scientists and biologist merged. A new field of research has recently emerged in the use of individual nanocrystals for growing 2D and 3D superstructures and investigation of collective properties of these artificial quantum dots solids [1]. Fabrication of nm-order in mesoscopic scale is considered as the key for applications in data storage, functional devices, communications and technology. The nanostructures, which are randomly distributed, fluctuate in size and have an unchanged periodicity give significant limitations for applications. Hence, an ultimate challenge in materials research is now the creation of perfect nanometer-scale crystallites, identically replicated in unlimited quantities, in a state that can be manipulated and that behave as pure macromolecular substances. Thus the ability to systematically manipulate these is an important goal in modern materials chemistry. We first demonstrated self-organization of nanocrystals with formation, on a mesoscopic scale, of a monolayer in a compact hexagonal network and in 3D superlattices [1-5]. Crystallization followed by the unambiguous determination of the exact position of each nanocrystal in the superlattice structures is the most suitable way of characterization [6-7]. The physical properties of theses superstructures [1] (optic, magnetic, transport) markedly differ from those of isolated nanocrystals and bulk phase. They are mainly due to the close vicinity of nanocrystals that is to say to dipolar interactions.
248
M. P. Pileni
In this review we first demonstrate the intrinsic property of 5nm silver nanocrystals self organized in 3D FCC. 2. RESULTS AND DISCUSSION To study this property 5 nm silver nanocrystals coated with dodecyl alkyl chains are produced from reverse micelles [8]. The nanocrystals, characterized by a very low size distribution (10%), are then dispersed in hexane. The resulting colloidal solution is characterized by an UV-visible absorption spectrum attributed to the well-known Mie resonance centered around 2.9eV and interband transitions at larger energy [9]. This is in good agreement to that deduced from numerical calculations [10-11] on particles of isolated silver spheres taking into account the surrounding media and the presence of dodecyl alkyl chains. The influence of the chemical environment on the optical response of nanocrystals has also been demonstrated for others systems [12-13]. The theory used is an extension of Mie theory which is the solution of Maxwell's equation for an isotropic sphere. Surrounded by an infinite external medium (refractive index 1.3914), there are two concentric regions of dielectric material of specified thickness. By deposition of drop of the solution on HOPG substrate, the silver nanocrystals self-organize in compact hexagonal network [9] (Fig. 1A). The average distance between two adjacent nanocrystals is around 2 nm. The electron diffraction pattern shows concentric rings, which can be indexed (111), (200), (220) and (311) reflectance characteristic of a FCC structure [16]. Inside the concentric rings appear spots corresponding to various orientations of the nanocrystals. This clearly shows a good crystallinity of the nanocrystals. High resolution transmission electron microscopy (HRTEM) images show three structures identified as multiply twinned particles (MTP) such as decahedra (Fig.IB), icosahedra (Fig.lC) and cuboctahedra (Fig.ID) From this structural investigation [16], Ag nanocrystals appear to be highly crystallized. However, it can be seen that the crystal structure of the nanoparticles does not always correspond to that of the bulk solid. This is in good agreement with what has been shown in the literature for such very small crystals [17]. In addition, it is known that the energies of these different types of crystals are so close that in a given sample like here, it is expected that a statistical distribution of structures can be observed specially for the case of smaller sizes [18].
toself-organization of of5nm silver nanocrystals Intrinsic properties due to
' • •: " jtif
s*fc
2 nm
P* rt r
sjfk*f
'
•
249
:~f • • ? + '
*• •••:# Si?.**: • i . v . . r
2 nm>y
^
Fig. 1. A-TEM of silver nanocrystals having 5nm as a mean diameter B,C and D. HTEM images: with (B) Decahedron viewed along the fivefold axis; (C) Icosahedron viewed along the threefold axis ; (D) Cuboctahedron in the [110] orientation.
On increasing nanocrystal concentration small aggregates characterized by a four-fold symmetry are observed by TEM indicating an FCC structure [9]. On controlling substrate temperature (22°C) and evaporation rate, large aggregates [19-20] are formed with heights and widths of several hundreds of um respectively (Fig. 2A). The diffraction pattern shows two strong reflections normal to the substrate [21]. The absence of further diffraction orders is due to a decrease in the structure factor for spherical nanocrystals. The width of the first order reflection is found to be nearly resolution limited, indicating long range ordering of the silver nanocrystals perpendicular to the surface. Other reflections less intense are also observed. Furthermore, the diffraction pattern shows a weak ring with an intrinsic width nearly resolution-limited. This confirms a long range ordering of the silver nanocrystals. Tilting the sample by 10°, the diffraction pattern (insert Fig.2A) shows numerous additional diffraction spots revealing a 3D long-range ordering within supracrystalline domains. From these results, it appears that the ordered domains share a common crystallographic axis normal
250
M. P. Pileni
to the substrate and that their in-plane orientation is random. A comparison of observed and calculated diffraction spot coordinates leads to a face centered cubic packing. At low substrate temperature (10°C), compact islands formed by the stacking of several layers of nanocrystals (Fig.2B) with appearance of defects and a very rough surface are observed. The diffraction pattern (insert Fig. 2B) shows a broad diffusing ring typical of a disordered arrangement. Hence with the same nanocrystals, it is possible to produce either "supra" crystals in FCC structure or disordered aggregates. We study the Stokes-antiStokes Raman spectrum of nanocrystals forming these two assemblies (either disordered aggregate or "supra" crystals).
Fig. 2. SEM images and inserted X ray diffraction patterns of 5nm silver nanocrystals: A- self organized in FCC structure and forming a "supra" crystal; B. forming a disordered assembly of nanocrystals
Let first consider disordered aggregates of highly crystallized nanocrystals. Figure 3A shows the Stokes-antiStokes Raman spectrum of this assembly. The vibrations are characterized by the quantum numbers n and /, like for spherical harmonics [22], and the vibrational frequencies are given by:
Intrinsic properties due to self-organization of 5nm silver nanocrystals
-30
-20
-10 0 10 Raman shift (cm1)
-20
-10 0 Raman shift (cm")
20
251
30
Fig. 3. Stokes-antiStokes Raman spectrum of silver nanocrystals deposited on a substrate. A. at 10°C and forming a disordered assembly; B. at 22°C and forming an FCC "supra" crystals.
D
(1)
where S[n depends on the ratio v Ivp The quadrapolar modes [23-25] appear as sharp intense lines. Because the nanocrystals assembly is characterized by a size distribution [9] we would expect also a vibrational frequency distribution. The good agreement of the Stokes lineshape with the inverse size distribution demonstrates the zwfr-a-nanoparticle coherence, i.e. nanocrystallinity. This agrees
252
M. P. Pileni
with data published previously by Duval et al [24] for silver nanocrystals dispersed in a polymeric matrix. The Stokes-antiStokes Raman spectrum of small "supra" crystals (dashed line) in FCC structure (Fig. 3B) is shifted toward low frequency while the quadrapolar lines is narrowed compared to that obtained with disordered assembly. From calculation described in reference 21 the Raman vibration intensity ISD(v) for the scattering by the nanocrystals vibrating coherently in a small supracrystal is: (2) This equation shows that in the case of small supracrystals, the coherence effect shows up as a narrowing of the Raman peak. Figure 4 compares the profile obtained with the "supra" crystals (solid line) with the square intensity of the profile of disordered assembly multiplied by the square of the frequency (v2) (dotted line). The good matching in Fig. 4 means mat the average size of the supra crystals is smaller than the light wavelength.
-1D
0
10
20
1
Raman shift (cm )
Fig. 4. Comparison of the Raman scattered intensity I(v) from silver nanocrystals a "supra" crystals) with the product I(v)v2 where this I(v)is the Raman scattered intensity of a disordered assembly of nanocrystals
The shift to lower frequency is due to the effect of the local electric field in the FCC "supra" structure. The scattered intensity is proportional to the square of the local field. Such situation differs from a uniformly disordered arrangement of nanocrystals having different sizes, where the local electric field has the same mean value for all nanocrystals. Therefore, in a set of supracrystals where each of them are built with nanocrystals of one same size which differs from one
due to self-organization of 5nm silver nanocrystals Intrinsic properties properties due
253
supracrystal to another, the relatively strong Lorentz field E\ at the plasmon resonance enhances the intensity of Raman scattering and induces a shift of the quadrapolar mode as observed in Figure 3B. 3. CONCLUSION We first demonstrate that self organization of silver nanocrystals in compact hexagonal network induces intrinsic properties of the self-organization. This is in addition to the collective properties with appearance of coupled plasmon modes due to induced dipole-dipole interactions [9, 26-29]. REFERENCES [I] [2] [3] [4]
M. P. Pileni, J. Phys. Chem., 105 (2001) 3358. L. Motte, F. Billoudet, and M. P. Pileni, J. Phys. Chem., 99 (1995) 16425. M. Brust, D. Bethell, D. J. Schiffrin, and C. Kiely, Adv. Mater., 9 (1995) 797. S. A. Harfenist, Z. L. Wang, M. M. Alvarez, I. Vezmar, and R. L. Whetten, J. Phys. Chem., 100 (1996) 13904. [5] L. Motte, F. Billoudet, E. Laeaze, J. Douin, and M. P. Pileni, J. Phys.Chem. B, 101 (1997) 138. [6] A. Courty, C. Fermon, and M. P. Pileni, Adv. Mater., 13 (2003) 58. [7] I. Lisiecki, P. A. Albouy, and M.P. Pileni, Adv. Mater., 15 (2003) 712. [8] M. P. Pileni, J. Phys. Chem., 97 (1993) 6961. [9] A. Taleb, C. Petit, and M. P. Pileni, J. Phys. Chem. B, 102 (1998) 2214. [10] U. Rreibig and M. Vollmer (Eds.) J. Peter Toennies, Optical Properties of metal clusters, Springer-Verlag, Berlin, Springer Series in Material Science Vol,25,1993. [II] B. N. J. Persson, Surf. Sei., 281 (1993) 153. [12] P. Mulvaney, Langmuir, 12 (1996) 788. [13] M. D. Malinsky, K. L. Kelly, G. C. Schatz, and R. P. Van Duyne, J. Am. Chem. Soc., 123 (2001) 1471. [14] A. L. Aden and M.J. Kerker, J. Appl. Phys., 22 (1951) 1242. [15] C. F. Bohren, and D. R. Huffman, Absorption and Scattering of Light by Small Particles 1983, Wiley-Interscience. [16] A. Courty, I. Lisiecki, and M. P. Pileni, J. Chem. Phys., 116 (2002) 8074. [17] J. Urban, H.Sack-Kongehl, and K. Weiss, Zeitchrift Fur Physik D,, 28 (1993) 247. [18] M. Jose Yacaman, J. A. Ascencio, H. B. Liu, and J. Gardea-Torresdey, J. Vac. Sci. Technol. B, 19 (2001) 4. [19] A. Courty, C. Fermon, andM. P. Pileni, Adv. Mater., 13 (2001) 254. [20] A. Courty, O. Araspin, C. Fermon, and M. P. Pileni, Langmuir, 17 (2001) 1372. [21] A.Courty, A. Mermet, P.A. Albouy, E. Duval, and M.P. Pileni, Submitted for publication. [22] Lamb, J. Proc. London Math, Soc, 13 (1882) 187. [23] Palpant, H. Portales, L. Saviot, J. Lerme, B. Prevel, M. Pellarin, E. Duval, A. Perez, and M. Broyer, Phys. Rev. B, 60 (1999)17107. [24] H. Portales, L. Saviot, E. Duval, M. Fujii, S. Hayashi, N. Del Fatti, and F.Vallee, J. Chem. Phys., 115 (2001) 3444. [25] E. Duval, H. Portales, L. Saviot, M. Fujii, K. Sumitomo, and S. Hayashi, Phys. Rev. B, 63 (2001) 075405.
254
M. P. Pileni
[26] A. Taleb, V. Russier, A. Courty, and M. P. Pileni, Phys. Rev. B, 59 (1999) 13350. [27] N. Pinna, M. Maillard, A. Courty, V. Russier, and M. P. Pileni, Phys. Rev. B, 66 (2002) 45415. [28] M. Maillard, P. Montchicourt, andM. P. Pileni, Chem. Phys. Lett., 107 (2003) 7492. [29] F. Silly, A. O. Gusev, A. Taleb, F. Charra, and M. P. Pileni, Phys. Rev. Lett., 84 (2000) 5840.
Nanophotonics, Volume 2 Handai Nanophotonics, S. Kawata and H. Masuhara (Editors) © 2006 Elsevier B.V. All rights reserved. reserved.
255 255
Chapter 14
Gold nanorods: preparation, characterization, and applications to sensing and photonics S. Yamada and Y. Niidome Department of Applied Chemistry, Graduate School of Engineering, Kyushu University, Hakozaki, Higashi-ku, Fukuoka, 812-8581, Japan 1. INTRODUCTION Rod-like gold nanoparticles (gold nanorods: AuNRs) show unique optical properties depending on the size and the aspect ratio (the ratio of longitudinal-to-transverse length) [1]. Although the spherical gold nanoparticle (nanosphere) has only one surface plasmon (SP) band in the visible region, the NR has a couple of SP bands. As shown in Fig. 1, one SP band corresponding to the transverse oscillation mode locates in the visible region at around 520 nm, while the other corresponding to the longitudinal oscillation mode between far-red and near-infrared (near-IR) region. This is the distinctive optical characteristic of the NR as compared with the nanosphere. In spite of such unique optical property, the NR had not been utilized for scientific and technological fields, because of the absence of high-yield synthetic methods with desirable size and aspect ratio. Recently, however, owing to the recent reports on high-yield preparation methods of tailor-made AuNRs, extensive applications of the NRs have just started in various fields. In this article, we will describe recent progress about the preparation, characterization, and optical and technological applications of the NRs with mainly focusing on our studies. 2. SYNTHESIS Torigoe and Esumi first reported the synthetic methods of rod-like gold nano-particles [2,3]. Recently, several approaches for high-yield preparation of the NRs have been reported, as summarized in Fig. 2: template using nonporous alumna (a) [4-7], electrochemical (controlled current electrolysis) (b) [8,9],
256 256
Yamada and and Y. Y. Niidome Niidome S. Yamada
(b) transverse mode
S0.4
)
|^ i longitudinal , j mode M
V
\ \ \
^ ^ 0
400
600 800 Wavelength / nm
1000
Fig. 1, Transmission electron microscopic (TEM) image (a) and absorption spectrum (b) of AuNRs.
of alumina alumina dissolution of
deposition of of gold
gold electrode electrode
rod
JUI1
rod
O
nanoporous alumina (a) Alumina template Au33++ s
(b) Electrolysis UV light
Au3+
gold seed rod (c) Seed-mediated growth Fig, 2. Preparation methods of AuNRs.
rod
k
Au33++ (d) Photoreduction
Gold nanorods: preparation, preparation, characterization, and applications to sensing and photonics
257 257
seeding (seed-mediated growth) (c) [10-13], and photochemical (d) [14] processes. The electrochemical and photochemical methods can provide fairly uniform NRs. In the case of the electrochemical method, however, considerable amount of spherical nanoparticles are formed together with the NRs. Since the electrochemical method involves complicated conditions such as sonication and constant current electrolysis, it is difficult to improve the uniformity of NRs. The photochemical method, on the other hand, hardly produces the spherical nanoparticles, and gives high yields of NRs. This method is relatively simple and the shape (aspect ratio) of NRs can be controlled by the amount of silver ions present in the reaction solution. However, it requires very long reaction time of about 30 hours. The seeding method is rather complicated because it still requires many kinds of reagents and more than two steps to generate the NRs: preparation of gold nanoparticles as seeds and subsequent growth reaction using a different solution. Until now, however, transition states for the growing processes of NRs have not been verified yet. Quite recently, we have succeeded in the rapid synthesis of AuNRs by the combination of chemical reduction and photo irradiation, as shown in Fig. 3(a) [15]. During the reaction, we also found reproducible shape changes in the NRs from rectangular, I-shape (dumbbell-like), and capsule-like with the proceeding the reaction time. The reagent conditions are similar to the previous photochemical method, except the absence of tetradodecylammonium bromide (TDAB). As a typical example, the reaction solution containing hexadecyl-tri-methylammonium bromide (CTAB), hydrogen tetrachloroaurate, small amounts of acetone and cyclohexane, and certain amounts of silver nitrate solution were added to the reaction solution. As a next step, ascorbic acid (AS) was added to the solution to reduce Au3+ ion. Subsequently, an aliquot of the reaction solution was put into a thin quartz cell (optical path length: 1 mm), and was irradiated by ultraviolet (UV) light from an ultrahigh-pressure mercury lamp through a band pass filter to eliminate the visible light. Fig. 3(b) shows absorption spectra of the reaction solutions before and after addition of AS, and those of the UV-irradiated solutions. The absorption peaks of AuCLf at around 380 nm (curve (A)) disappeared instantaneously by the addition of AS (curve (B)), indicating the reduction of AuCl4". Since the solution showed no absorption peak in the region from 300 nm to 1000 nm, it is clear that the addition of AS formed no larger gold particles having the SP band. As shown in (C) of Fig. 3(b), a couple of plasmon peaks characteristic of NRs appeared at about 520 nm and 600 - 800 nm by photoirradiation and they grew with proceeding photo irradiation (Fig. 3(c)). Typical transmission electron micrograph (TEM) images of the reaction solution after 5 and 30 min of photo irradiation are shown in Fig. 3(d). After 5 minutes of irradiation, cylindrical nanoparticles are formed almost quantitatively, and their shapes are highly uniform. In the case of 30-minute irradiation, on the other hand, both ends of the NRs are dumbbell-like (I-shaped)
258
S. Yamada and Y. Niidome
with high uniformity. The reaction is very rapid but its mechanism is very complicated. As an example, we have investigated the effects of silver ion concentration on the generation of AuNRs, as shown in Fig. 4. The longitudinal SP band showed red-shift with increasing the concentration of silver nitrate. This indicates that the longer (that is, higher-aspect ratio) NRs are generated at higher concentrations of silver ions.
(a)
A: yellow
B: colorless
HAuCl4, CTAB, AgNO3, acetone, cyclohexane
400 600 800 Wavelength / nm
(d)
5 min
50 nm
C: purple
TEM image
uVlight several min-
1000
400
600 800 1000 Wavelength / nm
i 30 min fjt
/** •<
50 nm
Fig. 3. Preparation procedure of AuNRs by the combination of chemical reduction and photo-irradiation.
Gold nanorods: preparation, preparation, characterization, and applications to sensing and photonics
400
600 800 Wavelength / nm
259
1000
Fig, 4. Effects of silver ion concentration on the formation of AuNRs.
A noteworthy difference of the present method as compared with the previous photochemical method is that the absence of TDAB in the reaction solution. In order to investigate the effects of other surfactants, we have added cationic or anionic co-surfactant with CTAB in the reaction solution [16]. However, addition of co-surfactant tended to hinder the formation of AuNRs. Probably the micelle (or assembling) structure of CTAB is essentially important for the generation of AuNRs [17].
3. STRUCTURES In order to elucidate the crystal structure of AuNRs, high-resolution TEM observations have been carried out. As a typical example, Wang et al. obtained high-resolution TEM images of AuNRs prepared by the electrochemical method [18]. Fig. 5(a) shows the TEM image of the NR oriented with its [010] parallel to the incident electron beam. In this case, (001) and (100) faces are images edge-on. An enlargement of the (100) surface is shown in the inset, which clearly shows the presence of 1 atom-height surface steps as indicated by arrows. The TEM image oriented along [1T0] is shown in Fig. 5(b). The (110), (001), (111), and (iil) surface are imaged edge-on. According the previous studies by Johnson et al. [19], it is suggested that CTAB molecules, in higher concentrations, bind more strongly to the {100} edges than the {111} faces. As a consequence, the crystal grows preferentially along the [110] direction as the side edges/faces becomes stabilized [19-21].
260
Yamada and Y. Niidome S. Yamada
(a)1;
001 ;, .. (b) , Hit-
001 111
110
100 2 nm
111
1 nm
Fig. 5. High-resolution TEM images of AuNRs. Reproduced with pennission from reference 18.
4. SPECTRAL SIMULATIONS Simulation of the optical spectra of AuNRs has been carried out by Link et al [22]. Using an extension of the Mie theory, they attempted to explore the relationship between the aspect ratio of AuNRs and the position of the SP band maximum. The extinction coefficient 6 of randomly oriented particles in the dipole approximation is expressed by the following equations:
s=2xNVsm
3/2
(1)
(2) B (3)
(4)
where N the number of particle per unit volume, V the volume of each particle, £„, the dielectric constant of the interacting light (assumed to be a constant), and Ei and 62 are the real and imaginary part of the gold dielectric function and the latter one is frequency dependent. Then, the depolarization factors for the
Gold nanorods: preparation, preparation, characterization, and applications to sensing and photonics
261
longitudinal (A) and transverse (B) axes of NR, represented as PA and respectively, are included in the above equations ((1) ~ (4)).
R-3.6 2.5
3.0 3.5 ratio R—2.6
R
400
500
600 700 Wavelength / nni
800
900
800
900
(b) 400 300
100
0 400
500
600
700
Wavelength / nni Fig. 6. Calculated spectra of AuNRs. Reproduced with permission from reference 22.
Fig. 6(a) shows the calculated spectra of the AuNRs with different aspect ratios from 2.6 to 3.6, where the £„, value is assumed to be 4. A couple of SP bands are seen in the spectra. The band maximum of the transverse oscillation mode shows blue-shift with increasing the aspect ratio, while the longitudinal oscillation mode shows large red-shift. The effect of longitudinal oscillation mode is much more pronounced. Calculated spectra with different Em values under the constant aspect ratio of 3.3 are shown in Fig. 6(b). In this case, both maxima show red-shift and increase intensity with increasing the £„, value. As in
262
Yamada and Y. Y. Niidome S. Yamada
the case of Fig. 6(a), the longitudinal oscillation mode is more sensitive than the transverse one. In any case, these theoretical treatments are very helpful for understanding the optical and physical properties of AuNRs. 5. APPLICATIONS 5.1. LPR Sensing The method of surface plasmon resonance (SPR) sensing has been extensively applied as a powerful tool for chemical and dispensing. The basic principle of the popular SPR method is to detect the resonant angle of the reflected light from the planar metal (gold or silver) film; thus the optical alignment is rather complicated and the incident angle of light should be strictly adjusted. As to gold nanoparticles, SP waves are localized at the particle surface, and therefore the use of these localized SP waves for sensing as in the case of the planar metal system is possible. In addition, the strict adjustment of the incident angle of the light is not necessary in the nanosphere from the geometrical point of view. Accordingly, the SPR method using non-planar metal surfaces (such as gold nanoparticle surfaces) can be called as the localized SPR or the localized plasmon resonance (LPR) method. In order to increase the photometric sensitivity using gold nanoparticles, the SP bands should be as strong and sensitive as possible to the dielectric constant of the surrounding medium. As described in the previous section, the plasmon peak of longitudinal SP band is quite sensitive to the surrounding medium. From these viewpoints, AuNRs are promising as compared with spherical gold nano-particles. AuNRs have already been applied as some sensing tools such as DNA-sensing [23] as in the case of gold nanoparticles and for surface-enhanced Raman scattering [24,25]. However, the deposition of AuNRs on the substrate was limited only by a precipitation procedure [25], and useful methods for depositing the AuNRs without aggregation have not yet been reported. Quite recently, we have succeeded in well-dispersed fixing of the AuNRs onto a glass substrate by the layer-by-layer approach [26]. Before performing the layer-by-layer deposition process, a part of surface CTAB molecules as the capping molecules were replaced with thionicotineamide (TNA), to facilitate electrostatic deposition; this procedure was inevitable to achieve layer-by-layer deposition of NRs. The degree of the direct electrostatic deposition of NRs on this glass substrate was quite low, even on immersion into the aqueous colloidal solution of NRs. Thus, the surface of glass substrate was modified with polyions as follows. First, the glass substrate was immersed into an aqueous solution of poly (allylamine hydrochloride) (PAH) for 20 min, to generate cationic charges on the surface. Then, the PAH-modified substrate was immersed into an aqueous solution of poly
Gold nanorods: preparation, preparation, characterization, and applications to sensing and photonics
263
(sodiumstyrene sulfonate) (PSS) for 20 min, so that the outermost layer of the substrate was negative. Gold licinorod
(b)
500 nm
Fig. 7. Schematic illustration of AuNR-polyion assembly (a), and its scanning electron microscopic (SEM) image (b).
This PSS-PAH-modified precursor substrate was immersed into the aqueous colloidal solution of NRs for 30 min. However, when this NR-modified substrate was immersed into organic solvents, substantial aggregation of NRs was observed. In order to avoid the aggregation of the NRs, the NR-modified substrate was again immersed into an aqueous solution of poly (diallyldimethyl-ammonium chloride) (PDDA) for 20 min, to obtain the PDDA-NR-modified substrate. By treating with PDDA, the deposited NRs showed no appreciable aggregation even on immersion into the organic solvents. The proposed structure of PDDA-NR-modified substrate is shown in Fig. 7(a). As can be recognized from the scanning electron micrograph (SEM) image of the modified substrate (Fig. 7(b)), most NRs are isolated each other and distributed randomly on the substrate. The outmost PDDA layer may fill the spaces among NRs, avoiding lateral diffusion on PSS layer.
264
S. Yamada and Y. Niidome
Absorbance
0.03
0.02 Air 0.01 400
600 800 Wavelength / nm
1000 1000
Fig. 8. Spectra of AuNR-polyion assembly at various solvents.
Figure 8 shows the extinction spectra of the PDDA-NR-modifled substrate in several solvents. The spectrum of the substrate in air shows clear plasmon bands around 550 (transverse mode) and ~760 nm (longitudinal mode), respectively. On immersion into the solvent, the longitudinal plasmon band shows substantial red-shift, while the transverse SP band shows no appreciable shift. The positions of the longitudinal plasmon bands and the refractive indices of the solvents are summarized in Table 1. Table 1 Band maxima of longitudinal SP band in various solvents. Refractive Solvent Index 1.00 Air 762 Water 1.33 805 1.36 Ethanol 806 1.36 Acetone 807 1.43 Cyclohexane 818 Chloroform 841 1.45
The degree of red-shift is in the order: water > ethanol > acetone > cyclohexane > chloroform. This is roughly correlated with the order of refractive index (dielectric constant) of the solvent. However, the tendency of spectral shift is contradictory to the theoretical simulation as shown in Fig. 6. Thus, more quantitative investigations are needed for the optical properties of AuNRs. Anyhow, the longitudinal SP band of AuNR is experimentally found to be quite sensitive to the surrounding medium. Since the longitudinal band locates in the near-infrared region, applications to biosensing are fascinating.
Gold nanorods: preparation, preparation, characterization, and applications to sensing and photonics
265
5.2. SERS sensing The SERS method has been a powerful tool for microanalysis obtaining vibrational information on local areas. The enhancement factor of SERS using silver and gold nanoparticles is experimentally and theoretically estimated to be 106 ~ 1015 [27]. Because of its extraordinary sensitivity, SERS is expected to be a practical technique for single molecular detection [28], The enhancement factor of SERS depends on the condition of the metal surface (domain size, roughness, and so on). This is due to existence of "hot spots", for example a junction of the adjacent two particles or protrusion of a particle, having intense electromagnetic fields in which highly efficient Raman scattering can be obtained. acetonitrile hexan e
drying aqueous aqueous colloidal colloidal solution solution
nanoparticle film
Fig. 9. Schematic illustration for the preparation of gold nanoparticle films at the liquid/liquid interface.
As to optical spectrometry, the use of near-IR light offers practical advantages over the ultraviolet-visible region especially in biosensing, because most of biomolecules have no absorption in the near-IR region and thus do not cause serious interferences so long as the near-IR light is used. For example, near-IR lasers are suitable for the SERS detection using gold nanoparticles in order to decrease the Rayleigh scattering as compared with visible lasers, and to avoid possible melting of the gold nanoparticles due to the strong interaction between the visible light and the SP oscillation around 520 ran. Moreover, the molecules do not always have the absorption at the excitation wavelength of the Raman scattering, namely resonance Raman scattering condition. Therefore, the SERS detection using gold nanoparticles is important under the non-resonant condition of the molecules. Theoretical calculations suggest that the AuNRs are expected to be much more SERS-active than gold nanospheres. However, no
266
S. Yamada and Y. Niidome
convenient methods for preparing well-aligned AuNR films for SERS substrates had been reported so far. Quite recently, we have discovered a simple method of preparing monoparticular films of NRs and nanospheres (abbreviated hereafter as NSs) at liquid-liquid interfaces, as schematically shown in Fig. 9 [29]. For example, the liquid-liquid interface of the aqueous colloidal solution of gold nanoparticles and hexane is formed. Then, acetonitrile or methanol is vigorously injected into the colloidal solution. Addition of the above polar solvent changes the reddish color of the colloidal solution into pale pink, and produces the film of the gold particles at the liquid-liquid interface. The resultant film is transferred onto a glass plate by vertical lifting from the bottom of the vial through the film. We have prepared three kinds of gold nanoparticle films to compare the SERS activities: NRs with aspect ratios of 5.5 (H-NR) and 3.2 (L-RN), and NSs with mean diameter of ~ 15 nm.
Absorbance
1.0
8
3
(b)
0.6 0.4
(c)
(a)
0.2 0.0 400
600
800 1000 1000 Wavelength / nm
Absorbance
(f)
0.0 400
1400 1400
(d)
0.4
0.2
1200 1200
(e)
600
800 1000 1000 Wavelength / nm
1200 1200
1400 1400
Fig. 10. Absorption spectra of H-NRs (a), L-NRs (b), and NSs (c) in water, and of H-NR (d), L-NR (e), and NS (f) films.
Fig. 10 shows the absorption spectra of the colloidal solutions and films of three nanoparticles. In the colloidal solutions, the SP bands are clearly seen and the L-NRs give the largest peak in the near IR region. However, the extinction spectra of the films are considerably broadened due to interparticle plasmon coupling. The SEM images of gold nanoparticle films are shown in the right
Gold nanorods: preparation, preparation, characterization, and applications to sensing and photonics
267
column of Fig. 11. The long axes of the both NRs are almost the same, but the short axis is longer in L-NRs. All films have well packed monoparticle-level domains. Raman scattering spectra of Rhodamine 6G (R6G) molecules cast on the films are shown in the left column of Fig. 11, together with the spectrum of R6G powder having characteristic peaks at 1311, 1363, and 1508 cm'1, respectively. It is clear that the Raman scattering intensity of the H-NR film is larger than that of the L-NR film. On the other hand, no appreciable Raman scattering peaks of R6G are detected from the NS film and glass slide; the broad scattering band around 1450 cm'1 may be ascribed to the glass slide. Since R6G has no absorption at the excitation wavelength (785 nm), no resonance effects for R6G molecules are included in the Raman scattering spectra in this case. Thus, the enhancement of the Raman scattering arises only from the surface effect excluding the resonance effect of the R6G molecules. Accordingly, it is clear that the H-NR film enhances the Raman scattering most effectively. A 42 nm 7.7 nm B
41 nm 13 nm C
~ 12 nm 800
1200
1600 2000
Raman Shift / cm 1 Fig. 11. Non-resonance Raman scattering spectra of Rhodamine 6G on the AuNRs (A, B) and nanosphere (C) and on the glass plate. The spectrum of powder is also shown at the bottom.
The quantity of R6G molecule in the observed area is estimated to be the order of 10"H mol, on assumption that the diameter of the focused laser light on
268
S. Yamada and Y. Niidome
the sample is about 0.5 mm. From the comparison with the results of gold nanospheres [30], the enhancement factor is estimated to be higher than 107. In any case, the H-NR film showed the higher sensitivity, indicating that the AuNRs with higher aspect ratios are more effective in electromagnetic field enhancement. 5 J . Optical storage (Laser writing) Morphological control of metal nanoparticles will open-up new fields in future optics and photonics. Pulsed-laser irradiation of metal nanoparticles can induce some morphological and/or size changes [9, 31-33], Particularly, the gold nanoparticles can be transformed into very large particles by repeated coagulation-fusion cycles in the case of near-IR excitation [34]. All of these shape transformation induce appreciable changes in the SP bands. As to AuNRs, the basic principle of laser-induced shape transformation is schematically shown in Fig. 12 [33]. Upon irradiation of very-short laser pulse, the free electrons absorb photons with order of fs, and then thermalize with the order of ps. Then the thermalized free electron and the holes combine and into the initial state though electron-phonon relaxation processes within sub-ns. These relaxation processes are exothermic, so that the NR is heated and causes shape change into ellipsoidal or spherical, dependent on the degree of photon absorption. Of course some of inner-core electrons are excited, though the degree of excitation depends on laser wavelength. Accordingly, pulsed-laser induced shape transformation of AuNRs induces significant spectral changes in the near-IR region, due to the substantial reduction or disappearance of the longitudinal SP band.
(a)
(b)
Fig. 12. Schematic illustration of laser-induced shape change: (a) thermal heating of electrons by phonon absorption; (b) lattice heating resulting from electron-photon relaxation; (c) shape transformation.
In order to make the most of this laser-induced reshaping phenomenon for optical data storage materials, NRs should be fixed (or aligned more desirably) in thin supports or films [35], We have incorporated AuNRs into a
Gold nanorods: preparation, preparation, characterization, and applications to sensing and photonics
269
poly(vinylalcohol) (PVA) film, and have carried out the laser-induced transformation of the AuNRs by using linearly-polarized 1064- or 532-nm laser light, as schematically shown in Fig. 13(a); circularly-polarized laser writing is also carried out for comparison (Fig. 13(b)). Although polarized laser writing has been achieved previously, the writing condition has not been optimized yet; especially the longitudinal SP band of AuNRs was not satisfactorily matched with 1064 nm [36]. Recent progress on the synthetic methods has made it possible to adjust the SP band with the laser wavelength for optical recording, as can be recognized from the following results [37]. The TEM image of AuNRs and the absorption spectrum of the NR-PVA film are shown in Fig. 13(c) and (d). The SP peaks are observed at 980 and 530 nm, overlapping well with 1064and 532-nm laser lights.
(a)
(b)
Linear (horizontal)
Circular
1064 nm: 5-7 ns, 50 mJ/pulse ⇔ <=> longitudinal 1064 <=>transverse 532 nm: 5-7 ns, 50 mJ/pulse ⇔
(c)
(d) Absorbance
1.2 1.2
100 nm
•t
0.8
532 nm
/
1064 nm nn 1064
0.4 0n 400
1000 1200 1200 600 800 1000 Wavelength / nm
Fig, 13. Schematic illustration for lineariy-(a) and circularly-(b) polarized laser irradiation to the AuNR-PVA film.
Schematic illustration of polarized-laser writing and polarized read-out system is shown in Fig. 14(a). One assumes the simplest case, where only two NRs, aligned horizontally (lying) and vertically (standing), are taken into consideration. In this geometry, the longitudinal SP band of the lying NR is
270
S. Yamada and Y. Niidome
resonant while the transverse SP band of the standing NR is not resonant with the horizontally-polarized 1064-nm laser light. Therefore, only the lying NR absorbs 1064-nm photons and causes shape transformation into ellipsoidal and/or spherical nanoparticles. (a)
1064 nm( f->) excitation -520 nrn
400
600 800 1000 1200 Wavclenffl.li /nm
-530 nm
600 800 1000 1200 Wavclcnfflh / nm
Fig. 14. Schematic illustration for the shape transformation of AuNRs by linearly-polarized 1064-nm laser irradiation (a) and the polarized absorption spectra of AuNR-PVA film before and after one shot of laser irradiation: (b), linearly-polarized; (c) circularly-polarized 1064-nm laser irradiation.
Fig. 14(b) shows absorption spectra of the film before and after one shot of 1064-nm laser irradiation (50 mJ/pulse). It is clear that the longitudinal SP band is remarkably reduced in the horizontally-polarized absorption spectrum, while the transverse SP band is somewhat increased (Fig. 14(b)); the reduction of baseline is not clear at this stage. This is clearly due to morphological changes of the NRs into ellipsoidal and/or spherical particles. Before laser irradiation, the directions of the NRs are distributed random as to the polarization direction of the laser light. The horizontally-polarized 1064-nm laser light can selectively photoexcite the longitudinal SP band of the lying NR and induces shape change. While, the standing NR does not absorb horizontally-polarized 1064-nm photons effectively and does not cause substantial shape transformation. Figures 14(c) shows the polarized absorption spectra of the NR-PVA film irradiated by the
Gold nanorods: preparation, preparation, characterization, and applications to sensing and photonics
271 271
circularly-polarized 1064-nm laser light (50 mJ/pulse), where the polarization direction of the monitor light is horizontal or vertical. Clearly, there is no appreciable dichroic property in the polarized absorption spectra, since the circularly-polarized light is isotropic in all polarization angles and thus the transformation of the NRs occurs randomly and independently to the direction of the NRs in the film. (a)
532 nm( f-») excilaiion -530 nm
400
600 800 1000 1200 Wavelength / nm
400
600 800 1000 1200 Wavelength / nm
Fig. 15. Schematic illustration for the shape transformation of AuNRs by linearly-polarized 532-nm laser irradiation (a) and the polarized absorption spectra of AuNR-PVC film before and after one shot of laser irradiation: (b), linearly-polarized; (c) circularly-polarized 532-nm laser irradiation.
The polarized photoexcitation geometry of NR is inversed in the case of horizontally-polarized 532-nm laser writing experiment, as shown in Fig. 15(a). In this case, the longitudinal SP band of the lying NR is not resonant while the transverse SP band of the standing NR is resonant with the horizontallypolarized 532-nm laser light. Therefore, ideally, the standing NR preferentially absorbs 532-nm photons and causes shape transformation into ellipsoidal and/or spherical nanoparticles. Figure 15(b) shows absorption spectra of the NR-PVA films irradiated by horizontally-polarized 532-nm laser light of the identical energy with the 1064-nm laser (50 mJ/pulse). The longitudinal SP band in parallel to the polarization direction of the 532-nm laser light is somewhat retained, while almost disappears in the vertical direction. Circularly-polarized
272
S. Yamada and Y. Niidome
writing induces the almost identical spectral changes as in the case of 1064-nm laser writing. In the case of 532-nm laser writing, the transverse SP band overlaps with the absorption based on interband transition (5d to 6sp), which induces the isotropic transformation of the NRs. Therefore, substantial shape transformation of the lying NR is also occurring, in different with the case of 1064-nm laser writing.
Fig. 16 (a) Schematic illustration of ideal shape transformations by polarized laser writing, (b) Schematic illustration of four spots generated by polarized laser irradiation conditions: left, 1064 nm; right, 532 nm; <-», horizontally-polarized laser light (top); o, circularly-polarized laser light (bottom), (c) Microscopic images of AuNR-PVA film observed by horizontally(left) and vertically- (right) polarized near-IR light at 910 nm. (d) Microscopic images of AuNR-PVA film observed by horizontally- (left) and vertically- (right) polarized visible light (530 nm).
Figure 16(a) shows the summary of the basic principle for polarized laser-induced shape transformation of NRs, and the writing conditions of the NR-PVA film is shown in Fig. 16(b). When the read-out is carried out by near-IR monitor light (910 nm), the contrast (change of color) produced by the polarized 1064-nm laser light is clearer in the horizontal read-out, while the situation is inversed in the case of horizontally-polarized 532-nm laser writing (Fig. 16(a)). Circularly-polarized laser writing induces similar degree of contrast. As can be recognized from Figs. 15 and 16, the read-out by using the visible monitor light (530 nm) causes no substantial differences in the contrast by laser writing. In the case of visible images at 530 nm (Fig. 16(c)), the contrast of color change is not so clear as in the case of near-IR images, because the absorption around the transverse SP bands (~ 530 nm) is not significantly changed as compared with the absorption around the longitudinal SP bands (Figs 15 and 16).
Gold nanorods: preparation, preparation, characterization, and applications to sensing and photonics photonics
273
Accordingly, linearly-polarized laser writing and polarized reed-out is realized by using the polarized laser light and the NR-PVA film. 6. CONCDUDING REMARKS In the present article, we have described recent progress on the synthetic methods and structural characterization of AuNRs, Control of their size and aspect ratio has been extensively improved and the strategy has also been extended to the synthesis of nanowires (more than -20 of aspect ratio). These improvements have materialized extensive applications of AuNRs. As shown in the present article, we have applied AuNRs to the substrate material for SERS sensing. Also, a new method of localized plasmon resonance (LPR) in the transmission geometry has been proposed. The laser-induced shape transformation of NRs in the polymer film has enabled optical storage with polarization geometry. Now, various kinds of metal NRs such as bi-metallic (for example, platinum and gold) NRs have been reported, and have already been applied for nonsocial functional materials such as fluorescent barcodes [38,39]. Also, alignment of metal NRs at the substrate has been applicable to nanoscale optical data storage systems [40]. At the same time, the gold nanoparticles can localize the incident light field at their near surfaces as the SP fields, and thus the gold nanowires as well as NRs can propagate the light wave along the long axes, functioning as nanoscale optical waveguides. Accordingly, the AuNRs are much more promising nanomaterials than the corresponding nanospheres in the future nanoscience and nanotechnology. REFERENCES [I] [2] [3] [4] [5]
S. Link and M. A. El-Sayed, J. Phys. Chem. B, 103 (1999) 8410. K. Torigoe and K. Esumi, Langmuir, 8 (1992) 59. K. Esumi, K. Matsuhisa, and K. Torigoe, Langmuir, 11 (1995) 3285. C. K. Preston and M. Moskovits, J. Phys. Chem., 97 (1993) 8495. C. A. Foss, Jr., Gt L. Hornyak, J. A. Stockert, and C. R. Martin, J. Phys. Chem., 98 (1994) 2963. [6] B. M. I. van der Zande, M. R. Bohmer, L. G. J. Fokkink, and C. Schonenberger, Langmuir, 16(2000)451. [7] B. R. Martin, D. J. Dermody, B. D. Reiss, M. Fang, L. A. Lyon, M. J. Natan, and T. E. Mallouk, Adv. Mater., 11 (1999) 1021. [8] Y. -Y. Yu, S.-S. Chang, C.-L. Lee, and C. R. C. Wang, J. Phys. Chem. B, 101 (1997) 6661. [9] S.-S. Chang, C.-W. Shih, C.-D. Chen, W.-C. Lai, and C. R. C. Wang, Langmuir, 15 (1999) 701. [10] N. R. Jam, L. Gearheart, and C. J. Murphy, Chem. Commun., (2001) 617. [II] N. R. Jana, L. Gearheart, and C. J. Murphy, Adv. Mater., 13 (2001) 1389. [12] C. J. Murphy and N. R. Jana, Adv. Mater., 14 (2002) 80.
274
S. Yamada and Y. Niidome
[13] T. K. Sau and C. J. Murphy, Langmuir, 20 (2004) 6414. [14] F. Kim, J. H. Song, and P. Yang, J. Am. Chem. Sac., 124 (2002) 14316. [15] Y. Niidome, K. Nishioka, H. Kawasaki, and S. Yamada, Chem. Commun., (2003) 2376. [16] Y. Niidome, K. Nishioka, H. Kawasaki, and S. Yamada, Colloids Surf. A, 257-258 (2005) 161. [17] B. Nikoobakht andM. A. El-Sayed, Langmuir, 17 (2001) 6368. [18] Z. L. Wang, R. P. Gao, B. Nikoobakht, and M. A. El-Sayed, J. Phys. Chem. B, 104 (2000) 5417. [19] C. J. Johnson, E. Dujardin, S. A. Davis, C. J. Murphy, and S. Mam, J. Mater. Chem., 12 (2002) 1765. [20] E. Leontidis, K. Kleitou, T. Kyprianidou-Leodidou, V. Bekiari, and P. Lianas, Langmuir, 18 (2002) 3659. [21] Y. Xia, P. Yang, Y. Sun, Y. Wu, B. Mayers, B. Gates, Y. Yin, F. Kim, and H. Yan, Adv. Mater., 15 (2003) 353. [22] S. Link, M. B. Mohamed, and M. A. El-Sayed, J. Phys. Chem. B, 103 (1999) 3073. [23] E. Dujardin, L. -B. Hsin, C. R. Chris, and S. Mann, Chem. Commun., (2001) 1264. [24] B. Nikoobakht, J. Wang, and M. A. El-Sayed, Chem. Phys. Lett., 366 (2002) 17. [25] B. Nikoobakht and M. A. El-Sayed, J. Phys. Chem. B, 107 (2003) 3372. [26] Y. Niidome, H. Takahashi, S. Urakawa, K. Nishioka, and S. Yamada, Chem. Lett,, 33 (2004) 454. [27] D. -S. Wang and M. Kerker, Phys. Rev. B, 24 (1981) 24. [28] S. Nie and S. R. Emory, Science, 275 (1997) 1102 [29] M. Suzuki, Y. Niidome, N. Terasaki, K. Inoue, Y. Kuwahara, and S. Yamada, Jpn. J. Appl. Phys., 43 (2004) L554. [30] M. Suzuki, Y. Niidome, Y. Kuwahara, N. Terasaki, K. Inoue, and S. Yamada, J. Phys. Chem. 8,108(2004)11660. [31] A. Takami, H. Kurita, and S. Koda, J. Phys. Chem. B, 103 (1999) 1226. [32] H. Fujiwara, S. Yanagida, and P. V. Kamat, J. Phys. Chem. B, 103 (1999) 2589. [33] S. Link, C. Burda, M. B. Mohamed, B. Nikoobakht, and M. A. El-Sayed, J. Phys. Chem. A, 103(1999)1165. [34] Y. Niidome, A. Hori, T. Sato, and S. Yamada, Chem. Lett., (2000) 310. [35] O. Wilson, G J. Wilson, and P. Mulvaney, Adv. Mater., 14 (2002) 1000. [36] Y. Niidome, S. Urakawa, M. Kawahara, and S. Yamada, Jpn. J. Appl. Phys., 42 (2003) 1749. [37] H. Takahashi, Y. Niidome, and S. Yamada, to be submitted. [38] S. R. Nicewarner-Pena, R. G Freeman, B. D. Reiss, L. He, D. J. Pefla, I. D. Walton, R. Cromer, C. D. Keating, and N. J. Natan, Science, 294 (2001) 137. [39] N. I. Kovtyukhova and T. E. Mallouk, Chem. Eur. J., 8 (2002) 4354. [40] H. Ditlbacher, J. R. Krenn, B. Lamprecht, A. Leitner, and F. R. Aussenegg, Opt. Lett, 25 (2000) 563.
Nanophotonics, Volume 2 Handai Nanophotonics, S. Kawata and H. Masuhara (Editors) © 2006 Elsevier B.V. All rights reserved. reserved.
275 275
Chapter 15
Optical trapping and assembling of nanoparticles H. Yoshikawa, C. Hosokawa, and H. Masuhara Department of Applied Physics and Frontier Research Center, Osaka University, Osaka 565-0871, Japan
1. INTRODUCTION Colloidal aggregates of silver or gold nanoparticles produced by addition of electrolyte show giant local-field enhancements and have been widely used for surface enhanced Raman scattering (SERS) measurement. However such colloidal aggregates have problems like instability and poor reproducibility. Fabrication and control of colloidal aggregates are crucial for progress in plasmonics. Intense electromagnetic field produced by a focused laser beam attracts microparticles dispersed in liquid into the focal spot. This is known as optical trapping and used to manipulate various microparticles dispersed in liquid. If the size of particles is much smaller than that of the focal spot, i.e. the order of wavelength, a number of particles are trapped in the focal spot. This means that the focal spot of the laser beam acts as a microcage for nanoparticles and confined nanoparticles form a microassembly. We are studying a formation mechanism and optical properties of the colloidal assembly produced by the optical force. This "optical assembly" of nanoparticles is expected to become a novel micro- and nanofabrication technique using optical force. Laser fabrications using ultrashort pulse lasers have recently been advanced. A highly nonlinear process like a multiple excitation induced by ultrashort pulse lasers is making novel fabrications possible. They are classified in "top-down" process, in which a bulk material is fabricated in an ultrashort moment without a thermal damage. Against such trends, optical force produced by a non-resonant laser beam is utilized for "bottom-up" fabrication in our approach. Bottom-up process is exemplified by self-assembly, which is crucial to obtain precise molecular assemblies over a wide area, but generally it is difficult to control the structure. Self-assembly is not accomplished during a blink of a short laser pulse, since it takes a time to complete physical motion and rearrangement of molecules or nanoparticles.
276
H. Yoshikawa, C. Hosokawa and H. Masuhara
Namely an ultrashort pulse laser is inapplicable to such bottom-up fabrications and a continuous-wave (cw) laser is more suitable. In this chapter, the fundamental studies to produce and control assembling structures of nanoparticles by using the optical trapping technique are introduced. Nanoparticles are attractive materials for characteristic chemical, physical, and optical properties and show unique behavior for optical assembly, since their trapping potential is comparable to thermal energy, which is described in section 4 in detail. Gold-nanoparticles are one of the most representative nanoparticles, as they show characteristic optical properties due to surface plasmon resonance. In section 5 it is demonstrated that assembling structure and optical properties of gold-nanoparticles can be controlled by laser power. 2. PRINCIPLE OF OPTICAL ASSEMBLY When a laser beam is focused in nanoparticle suspension, a nanoparticle in the focal area is polarized by electromagnetic field of the laser beam. The induced dipole is attracted into the position where the electromagnetic field is in maximum, i.e. the laser focus. This potential energy of a nanoparticle produced by electromagnetic field (optical trapping potential) is described as follows [1], (n/n'f-l
(n/riff Thermal motion
ω
Optical trapping potential depth: Uo
Fig.l. A schematic diagram of an optical trapping potential
where a is a polarizability of a nanoparticle, E is electric field, n and n' are refractive indexes of nanoparticle and surrounding medium, / i s laser intensity, c is the velocity of light, and V is particle volume. This equation indicates that optical trapping potential (force) is proportional to the polarizability and laser
277
Optical trapping and assembling of nanoparticles
intensity. The spatial shape of the potential is similar with that of the laser focus, which is usually close to Gaussian shape. Figure 1 shows a schematically diagram of an optical trapping potential. The averaged duration rin which a nanoparticle is confined in the trapping potential (i.e. focal spot) is called escape time, that is approximately described as follows, (2) (3) where k is the Bolzmann's constant, Tis absolute temperature, aio is width of the laser spot (Gaussian beam parameter). The depth of the potential UQ is significant parameter to determine the trapping stability. Figure 2 shows the escape time t of a 20 nm-sized polystyrene particle plotted as a function of the laser power. The solid black circles represent experimental data measured by using fluorescence correlation spectroscopy (FCS) [2]. The solid line represents theoretical values calculated from eq.2. They are in good agreement with each other. 10 410
I
I
I
I
I
J-
τ / ms
8 6 4
-
2 0 t 0
I
I
i
i
200
400
600
800
i 1000 1000
r 1200
laser power / mW
Fig, 2. Escape time vs laser power. Solid black circles and solid line represent experimental data and the theoretical value calculated from Eq. (2), respectively.
If \UQ\ is larger than the energy of Brownian motion kT(k is the Bolzmann's constant and Tis the absolute temperature) sufficiently, the particle is confined in the potential for a long time, and vice versa. As the size of nanoparticles become smaller, the duration of the trapping becomes shorter, and the term of "trapping " seems to be no more suitable to express the situation. However, even if the trapped nanoparticles escape from trapping in a short time, a microassembly of
278
H. Yoshikawa, C. Hosokawa and H. Masuhara
trapped nanoparticles is formed as far as the frequency of trapping exceeds that of escape. Therefore high concentration is needed to make an optical assembly of nanoparticles. 3. EXPERIMENTAL SETUP FOR STUDY OF OPTICAL ASSEMBLY Our experimental setup for the study of optical assembly is described in Fig. 3.
Spectrometer
Colloidal Solution
Condenser Lens Halogen Lamp
Fig. 3 . An experimental setup for study of optical assembly of nanoparticles.
A 1064 nm fundamental beam from a cw Nd3+: YAG laser is introduced into an optical microscope. The laser beam is focused into a sample via an objective lens (xlOO magnification; numerical aperture, 1.25). Sample solution is dropped in a depression glass slide (1 mm depth) and covered by a cover slip (0.17 mm thickness). The spot shape is close to the Gaussian profile with 620 nm of the full width at half maximum (FWHM) in the focal plane. The focal point is set between 10 and 15 |im from bottom of the cover slip. Emission from nanoparticles trapped in the focal spot is detected by an avalanche photo diode (APD) operated in photon counting mode. For absorption (extinction) spectral measurement of trapped nanoparticles, a slightly collimated white light from halogen lamp is
Optical trapping and assembling of nanoparticles
279
illuminated to a sample from the opposite side of the objective. The transmission spectrum of the white light through the sample solution is measured by a polychromator with a CCD camera. A pinhole placed at the conjugate spot of the laser focus limits the detection area of the transmitted light in a spot 1 |J.m in diameter centered on the focal point. 4. OPTICAL ASSEMBLING DYNAMICS STUDIED BY THE SINGLE PARTICLE COUNTING Recently, in order to reveal the growth process of the molecular crystals or aggregates, the assembling dynamics of charged and hard nanoparticles has been extensively studied [3]. In general, crystallization or aggregation starts from somewhere in solution, so that it is difficult to catch the growth process of single crystal or aggregate from the initial stage. However optical assembly always emerges from a focal spot of laser beam. Of course similarities and differences between natural crystallization or aggregation and optical assembly must be understood. We have investigated the initial optical assembling process of polymer nanoparticles suspended in water [4]. Polystyrene nanoparticles including dyes were used as samples. When they are trapped in a focal spot of YAG laser (wavelength: 1064 nm), they fluoresce due to two-photon excitation. Samples were diluted to appropriate concentrations with distilled water. It was confirmed by using dynamic light scattering (DLS) apparatus that nanoparticles dispersed well without aggregates. When a 1064 nm cw-laser beam was focused in colloidal nanoparticles, the orange emission of the nanoparticles was observed and the intensity increased gradually. Fluorescence intensity is proportional to the square of excitation laser power, thus the effective excitation volume is restricted to the central portion of the focal spot, allowing only nanoparticles trapped at the focal spot to be detected as demonstrated in [5]. Therefore, this method is useful for investigation of optical assembling process of nanoparticles. Figure 4 shows temporal profiles of fluorescence intensity after the laser irradiation in the 100 nm-sized nanoparticles suspension. The increase in fluorescence intensity corresponds to that in the number of trapped nanoparticles. In high concentration, we confirmed that the number of nanoparticles increases continuously with decreasing a rate of the increment as shown in Fig. 4(a). This represents that nanoparticles are trapped one after another and filling up a focal spot. On the other hand, when the suspension was diluted (<109 particles/ml), a stepwise increase of fluorescence intensity was observed as shown in Fig. 4(b). Fluorescence intensity was converted into the particle number by measuring fluorescence of a same sized particle put on a slide glass. Because of fluctuation and photobleaching of the trapped nanoparticle, a step height in Fig. 4(b) is lower than the estimation (indicated as a bar).
280
H. Yoshikawa, C. Hosokawa and H. Masuhara
(a)
(b)
20 particles 40
60
Time / s
1 particle 50
100
Time/s
Fig. 4. Temporal trapping profiles of 100 nm-sized particles suspended in water. The concentration of the suspension is (a) 4.3*1010 and (b) 4.3*107 particles /ml. Vertical bars indicate fluorescence intensity corresponding to the particle number described beside each bar.
This assembling process can be described as Brownian motion of individual nanoparticles under the external potential energy: 14.^ = -o^E1\l2. The trapping potential shape is ascribed to the intensity distribution of the laser beam whose region is experimentally and numerically determined to be ~1 \im in the focal plane and ~3 pm on the optical axis, respectively. When the trapping potential energy is introduced to a diffusion equation, the probability flux [6] in unit area of nanoparticle suspension, namely the number of nanoparticles trapped in a focal spot in unit time is described as (4) where D is a diffusion coefficient, C is concentration (particles/ml) of nanoparticle suspension, and Ig is the laser power. This relation loses accuracy in high concentrated suspension because particle-particle interactions cannot be neglected in the assembling process, so that single particle counting in dilute suspension is necessary for this analysis. The validity of the relation (4) is evaluated by the statistical analysis of our experimental results. Figure 5(a) and (b) show the mean values of assembling rates (the number of trapping events in a unit time; particles/s) as a function of concentration and laser power, respectively. Figure 5 shows that the assembling rates were proportional to the concentration of nanoparticle suspension and the laser power, indicating that the assembling process of 100 nm-sized particles is consistent with the relation (4), i.e. diffusion limited process.
Optical trapping and assembling of nanoparticles
281 281
(b)
(a)
^0.01 t 1
10
1000
100
4
5
6 7 8 9
100 Laser power / mW
Concentration / lOparticles/ml
Fig. 5. Assembling rates of 100 im-sized particle suspensions as functions of concentration (a) and incident laser power (b). Solid lines represent linear relations. 2.0xl0:
(a)
(b)
1 particle
1 particle
(c)
5 particles 5
10
15
20
25
30
Time / s Fig. 6. Temporal trapping profiles of 40 nm-sized particles suspended in water. Concentration: 7 (a), (b) 8.8xlO and (c) 1 .8* 109 particles / ml. Laser power: (a), (c) 300 mW and (b) 900 mW.
In contrast to the 100 nm-sized particles, 40 nm-sized ones showed a characteristic assembling process depending on the laser power, as shown in Fig. 6. In the ease of 300 mW of laser power, packets of specified fluorescence intensity (500-1000 counts/0.05 s) were observed in a temporal profile, demonstrating that a couple of particles were trapped and escaped simultaneously. Furthermore it was observed that the number of nanoparticles increased stepwise
282
Yoshikawa, C. Hosokawa and H. Masuhara H. Yoshikawa,
one-by-one in the case of higher laser power, as shown in Fig. 6(b), which is a similar profile to that of 100 nm-sized particle suspensions. On the other hand, nanoparticles were assembled as a few-several particles are caught in one time in a higher concentration sample, as shown in Fig. 6(c). In the case of 100 nm-sized particles, laser power Io and concentration c contribute to the assembling process in the same way as represented by relation (4), and actually this is demonstrated in Fig. 4. These experimental results of 40 nm-sized particles can not be represented by a simple diffusion limited process represented by relation (3). This unexpected fluorescence signals shown in Fig. 6(a) are attributed to aggregates which have generated during optical trapping. In other words, clustering took place in advance of assembling in Fig. 6(a), while nanoparticles were assembled one-by-one in Fig. 4(b) and Fig. 6(b). We propose one model to explain this assembling dynamics. Nanoparticles form clusters with a certain possibility by getting over the repulsive potential of electric double layers. If the depth of trapping potential is larger than kinetic energy of Brownian motion and the repulsive interparticle potential is large enough to prevent clustering for a while, nanoparticles are trapped at the focal spot one-by-one, because trapping frequency exceeds escape one and clustering frequency around the focal spot is low. Namely nanoparticles are captured before cluster formation and this corresponds to assembling of 100 nm-sized particles [see Fig. 4(b)]. On the other hand, when trapping frequency is comparable low to escape one and clustering frequency around the focal spot is high, i.e. when the depth of trapping potential and repulsive interparticle interaction are small and weak, formed clusters possessing large polarizability are trapped and stay in the focal spot for a relatively long time. From the difference between Fig. 6(a) and (b), it is clear that above two cases are interchanged between two laser powers (300 mW and 900 mW) in the case of 40 nm-sized particles. In contrast, higher concentrated sample of 40 nm-sized particles gives higher trapping and clustering frequencies, so that aggregates are trapped and filling the focal spot [see Fig. 6(c)]. To evaluate the validity of this model, the assembling process of nanoparticles was investigated by Monte Carlo (MC) simulation [7], which calculates Brownian motion of nanoparticles under the optical gradient force in 2D system. Figure 7(a) and (b) exemplify the temporal profiles of 40 nm-sized particle suspensions in the simulation at the laser power of 300 mW and 900 mW, respectively. These results of MC simulation indicate that our proposed model gives a consistent explanation of the optical assembling process of nanoparticles.
Optical trapping and assembling of nanoparticles
283
Fig. 7. (a), (b) Numerical MC simulation of assembling process of 40 nm-sized particles. Concentration: 14 particles in 30 fim square region. Laser power: (a) 300 and (b) 900 mW. (c) Interparticle potential used in this simulation. The simulation protocol is appeared in Ref. 4.
5. OPTICAL ASSEMBLY OF GOLD-NANOPARTICLES In the former section, the mechanism and characteristic dynamics of optical assembly were studied. Next we demonstrate one representative examples of interesting and characteristic optical assembly in this section, which is a reversible control of the color of colloidal gold-nanoparticles by using optical assembling technique [8]. Gold-nanoparticles are interesting research target for optical assembly because their assemblies are reflected in the extinction spectrum [9]. The extinction spectral measurement of trapped nanoparticles was successful by using a quasi-confocal system developed for the present study. Optical trapping procedure is almost same as that in the former section. Nd:3+ YAG laser was focused in colloidal gold suspended in water (mean diameter: 40.5 nm, 9 xlO10 particles/ml) via a microscope objective. Extinction spectra of trapped gold-nanoparticles were obtained with a following procedure (see Fig. 8). A spectram of the transmitted white light through a sample, Ii(A), was measured without laser irradiation, then mechanical shutter open, and 10s later another spectrum, h(^), was measured. This 10 s is a waiting time for trapping of enough number of gold-nanoparticles. h(X) includes a little signal depression due to the laser absorption of trapped nanoparticles which have emerged on the light path. By calculating [IjfZj-fyZ)]/ h(A), extinction spectra of trapped nanoparticles were obtained.
284
H. Yoshikawa, C. Hosokawa and H. Masuhara
1
(λ)
I22(\)(λ) Tr an s in i s s i on s p e c t ruin
wavelength
wavelength
7,(A)-/ (A) (λ) 2(λ) (λ)
spec irome Ler
A
0 0
o
o
0 0
0
Gold colloidal solution
wavelength
AA Halogen lamp
Trapping laser OFF
Trapping laser ON
Fig. 8. A principle of an extinction spectral measurement of trapped gold-nanoparticles.
Figure 9 shows the extinction spectra of the trapped gold-nanoparticles measured for different laser powers ((a) 0 mW, (b) 100 mW, (c) 150 mW, (d) 200 mW, (e) 300 mW, (f) 450 mW, (g) 600 mW, (h) 750 mW, (i) 900 mW). Single extinction band corresponding to the SPR band is clearly observed in Fig. 2(c) and (d), demonstrating that gold-nanoparticles are trapped in the focal spot. As laser intensity increases further, another extinction band around 700 nm appears and grows. The spectral shape, especially intensity of the second SPR band in longer wavelength region, is strongly dependent on the laser power. In addition it is interesting that the spectral shape follows the laser power reversibly and repeatedly.
285 285
Optical trapping trapping and assembling assembling of of nanoparticles nanoparticles Optical
0.5W a
1
0
b
1
0
0 500
500
550
600
650
700
550 W 600 600 650 650 700 700 avelength/nm Wavelength Wavelength // nm nm
750
750 d
1
0
500 500
550 W 600 600 650 650 700 550 avelength/nm Wavelength Wavelength// nm nm
= =
e
1
600 650 650 700 700 550 W 600 avelength/nm nm Wavelength / nm
750 750
g
1
0 600 650 650 700 700 550 W 600 avelength/nm Wavelength Wavelength// nm nm
750
550 W 600 600 650 650 700 700 550 avelength/nm Wavelength / nm Wavelength/nm
750 750
f
11]
0 500 500
550 W 600 600 650 650 700 700 550 avelength/nm nm Wavelength / nm
1
500
750 750
h
550 W 600 600 650 650 700 700 550 avelength/nm nm Wavelength / nm
750 750
i
1
0
0 500
500 500
750 750
0 500
c
1
500 500
550 600 650 550 W avelength/nm 650 700 700 Wavelength Wavelength // nm nm
750 750
500 500
550 W 600 600 650 650 700 700 avelength/nm Wavelength / nm Wavelength/nm
750 750
Fig. 9. Extinction spectra of trapped gold-nanoparticles. All spectra measured at different laser powers [ (b) 100 mW, (c) 150 mW, (d) 200 mW, (e) 300 mW, (f) 450 mW, (g) 600 mW, (h) 750 mW, and (i) 900 mW ] are normalized.
Figure 9 shows the extinction spectra of the trapped gold-nanoparticles measured for different laser powers ((a) 0 mW, (b) 100 mW, (c) 150 mW, (d) 200 mW, (e) 300 mW, (f) 450 mW, (g) 600 mW, (h) 750 mW, (i) 900 mW). Single extinction band corresponding to the SPR band is clearly observed in Fig. 2(c) and (d), demonstrating that gold-nanoparticles are trapped in the focal spot. As laser intensity increases further, another extinction band around 700 nm appears and grows. The spectral shape, especially intensity of the second SPR band in longer wavelength region, is strongly dependent on the laser power. In addition it is interesting that the spectral shape follows the laser power reversibly and repeatedly. Particle size and laser power dependences of extinction spectra are summarized in Fig. 10. In the case of 100 mW of laser power, 30 nm-sized particles show no spectrum, whereas a clear extinction band appears in 60 nm-sized particles. This indicates that 30 nm-sized particles are hardly trapped under this laser power because of its small polarizability (particle volume). When the laser power increases, the second extinction band in longer wavelength region appears distinctly in 30-nm sized particles, though it is relatively weak in 60 nm-sized particles. This represents that smaller particles can come close to each other in shorter distance than larger particles. Namely, repulsive interparticle potential of 30 nm-sized particles would be smaller than that of 60 nm-sized
286
H. Yoshikawa, C. Hosokawa and H. Masuhara
particles, which is of course adaptable only for the colloidal gold used in the present study. Diameter Laser power 0.1 W
30 nm
40 nm
60 nm
3.. 500
550
600
650 700 750
Wavelength /nm
500 550 600 650 700 750 Wavelength/nm
500 550 600 650 700 750 Wavelength /nm
0.2 W 500 550 600 650 700 750 Wavelength /nm
500 550 600 650 700 750 Wavelength /nm
SOD 550 BOD BSD 700 760 Wavelength /nm
500 550 600 650 700 750 Wavelength /nm
SOD S5D BOO BSD 700 7SD
SOD S5D BDD BSD 700 760
0.3 W
Wavelength /nm
Wavelength /nm
Fig. 10. Extinction spectra of trapped 30,40, and 60 nm-sized gold-nanoparticles measured at different laser powers.
The general aspect of the extinction spectral shape and its laser power dependence are discussed here. Such double-peak extinction spectra were due to the aggregation of gold-nanoparticles [10] and the second extinction band shown in longer wavelength region is ascribed to the longitudinal SPR excitation for non-spherical or aggregated nanoparticles according to the generalized Mie theory [11]. Not only permanent aggregates, but also just assembled nanoparticles (closely locating each other) show the second band because of collective electromagnetic oscillations between neighboring nanoparticles. Peak height, wavelength, and band width of the first and second bands are strongly dependent on disposition of nanoparticles and interparticle distances [9]. Figure 11 shows the situation in which gold-nanoparticles dressed in electrostatic potential barriers are confined in an optical mierocage produced by the focused laser beam. The depth of the potential energy Uo is calculated as ~ 6.2kT(kis Bolzmann's constant and Tis temperature, kT- 4.12 x 10'21 J) at 100 mW of laser power and proportional to the laser power (i.e. 9.3 kT at 150 mW, 12.4 kTat 200 mW, and so on).
Optical trapping and assembling of nanoparticles
50 0
550 WO 650 TOO "Yavelenslh/nm
750
550 GOO 650 71)11 Wavelength / ntn
750
5U1J
287
551) MM 650 700 7511
Fig. 11. Top: Schematic image of reversible conformational changes of the trapped gold-nanoparticle assembly. Middle: Profiles of potential produced by the focused laser beam. Bottom: Extinction spectra
In the figure, (a) nanoparticles prefer to stay in the focal spot because the potential depth is larger than the energy of Brownian motion, but they are separated from each other by electric double layers, so that an extinction spectrum ascribed to isolated gold-nanoparticles is obtained, (b) As laser power becomes higher, nanoparticles start to approach and make small assemblies like doublets or triplets. Assemblies take a kind of one-dimensional conformation because repulsive interactions between nanoparticles prevent to make compact morphologies, presenting double peak spectra, (c) Higher laser power produces more compact assemblies consisting of various numbers of nanoparticles and morphologies, resulting in broader spectra with large components in longer wavelength region. It is noteworthy that the potential depth around trapped nanoparticles is comparable to and can be controlled with the order of the interparticle repulsive interaction: kT. If the potential depth of the optical trapping were much larger than the interparticle interactions, nanoparticles would be packed tightly into the focal spot without any space to change their locations. The soft confinement is a key point to realize the reversible control of the colloidal assemblies.
288
H. Yoshikawa, C. Hosokawa and H. Masuhara
6. CONCLUSION In conclusion, properties and anticipated effects of optical assembly of nanoparticles are summarized as follows. 1) Assembly structure can be controlled by laser power. Laser power can be tuned easily, precisely, and quickly by using various optical elements. 2) Nanoparticles possessing higher polarizability will be assembled preferably, so that obtained assembly may have different structures or properties from an assembly formed by conventional methods. 3) The potential energy of the optical trapping is comparable small to the repulsive interaction between colloidal nanoparticles. This allows trapped nanoparticles to exchange their position each other during optical trapping and to make self-assembling structures in a microcage of a focused laser beam. We have already developed fixation techniques of trapped nanoparticles on a glass substrate [12-14] and succeeded in deposition and control of polymer assemblies by using a focused laser beam [15-17]. By combining with these techniques, highly controlled colloidal microassembly could be deposited on a substrate, which becomes a useful method to produce microdevices working on the basis of surface plasmon of metallic nanoparticles. We believe that the optical trapping and assembling of nanoparticles is the promising technique which contribute the future nanophotonics and plasmonics. REFERENCES [I] [2] [3] [4] [5] [6] [7] [8] [9] [10] II1] [12] [13] [14] [15] [16] [17]
K. Svobada and S. M. Block, Opt. Lett., 19 (1994) 930. C. Hosokawa, H. Yoshikawa, and H. Masuhara, Phys. Rev. E, in press (2005). U. Gasser, E. R. Weeks, A. Schofield, P. N. Pusey, and D. A. Weitz, Science, 292 (2001) 258. C. Hosokawa, H. Yoshikawa, and H. Masuhara, Phys. Rev. E, 70 (2004) 061410. E. L. Florin, J. K. H. Horber, and E. H. K. Stelzer, Appl. Phys. Lett., 69 (1996) 446. N.G. van Kampen, Stochastic Processes in Physics and Chemistry, North-holland, Amsterdam, 1992. Chap. 8. M. P. Allen and D. J. Tildesley, Computer Simulation of Liquids, Clarendon Press, Oxford, 1987. Chap. 4. H. Yoshikawa, T. Matsui, and H. Masuhara, Phys. Rev. E, 70 (2004) 061406. U. Kreibig and M. Volkner, Optical Properties of Metal Clusters, Springer-Verlag, Berlin, 1994. J. Turkevich, G. Garton, and P. C. Stevenson, J. Colloid. Sci., 9 (1954) 26. U. Kreibig, Z. Phys. D, 3 (1986) 239. S. Ito, H. Yoshikawa, and H. Masuhara, Appl. Phys. Lett.,78 (2001) 2566. S. Ito, H. Yoshikawa, and H. Masuhara, Appl. Phys. Lett., 80 (2002) 482. S. Ito, H. Yoshikawa, and H. Masuhara, Jpn. J. Appl. Phys., 43 (2004) 885. S. Masuo, H. Yoshikawa, H. Masuhara, T. Sato, D-L. Jiang, and T. Aida, J. Phys. Chem. B, 106 (2002) 905. S. Masuo, H. Yoshikawa, H-G. Nothofer, A. C. Grimsdale, U. Scherf, K. Mullen, and H. Masuhara, J. Phys. Chem. B, 109 (2005) 6917. J. Hotta, K. Sasaki, and H. Masuhara, J. Am. Chem. Soc, 118 (1996) 11968.
Nanophotonics, Volume 2 Handai Nanophotonics, S. Kawata and H. Masuhara (Editors) © 2006 Elsevier B.V. All rights reserved. reserved.
289 289
Chapter 16
Femtosecond laser fabrication of three-dimensional metallic micro-nanostructures H.-B. Sun,3'" K. Kaneko," X.-M. Duan" and S. Kawata^ "Department of Applied Physics, Osaka University, Osaka 565-0876, Japan b
State Key Laboratory on Integrated Optoelectronics, College of Electronic Science and Engineering, Jilin University, China
Nanophotonics Lab, RIKEN, 2-1 Hirosawa, Wako, Saitama 332-0012, Japan 1. INTRODUCTION The unique optical performance of thin metal films and metal nanostructures has attracted a lot of research efforts due to the recent rebirth of the concept of surface plasmon resonance due to the pioneering work by Pendry et al [1-5]. Experimental study has been carried out on surface plasmon wave modulation of structured silver, gold, aluminium and copper thin layers for transmission and emission enhancement, and on near-field detection with metallic tips via enhanced local electromagnetic field [6-8]. Most structures studied by now are of zero, one or two dimensions. Our concern is three-dimensional (3D) metal nanostructures and the possible use of their surface plasmon wave features. As the first step towards this direction, we are now focusing on the development of technologies for fabrication of complicated-shaped 3D micro-nanostructures with metal component. 3D micro-nanostructuring is accomplished by means of laser micronanofabrieation [9-12]. The basic idea is tightly focusing the femtosecond laser pulses into a material, of which local properties is to be modified by pinpoint laser exposure. Scanning the laser focus according to a pre-programmed CAD pattern, the design is faithfully converted to a matter structure. These structures could be written either into transparent solids or into photopolymerizable liquid resins. In either case, the metal component is capable being introduced as a precursory material. In addition, structures with continuous metal layer are creatable by coating a thin metal-layer on laser-written polymer structures. In the following part of this chapter, we will introduce how metallic micronanostructures with arbitrary geometry are fabricated.
290
H.-B. Sun et al. H.-B. Sunetal.
2. STRUCTURES CONSISTING OF METAL FORMED BY IONS PHOTOREDUCTION
NANOPARTICLES
A straightforward approach to induce metal components into photopolymer to be used for laser fabrication is direct doping of nanoparticles. This method has been widely utilized in polymer industry for achieving, for example, appropriate mechanical strength and electric conductivity. It is, however, challenging to disperse the metal or metal oxide nanoparticles into a photopolymer matrix with sufficient homogeneity.
Fig. 1. Optical microscopic images of the nanopartiele-doped photopolymerizable resin: (a) TiCh nanoparticles, and (b) Z1O2 nanoparticles.
Figure 1 shows the optical microscopic images of the mixture of 5wt% 10nm-diameter TiO2 and 5wt% 50-nm-diameter Z1O2 particles dispersed into urethane acrylate resin with ethanol as the solvent. It is clearly seen that the mixtures presents milky white colour, meaning that the original nanoparticles are aggregated into various sizes so that light are universally scattered. For laser micro-nanofabrication, the best accuracy that has been achieved is less than 100nm [13], and a typical diamond-lattice PhC has a rod diameter of ~ 450 nm [14], For a uniform incorporation of these nanoparticles in the polymerized structures, their dispersion in liquid should be homogeneous at dimensions much smaller than the above scales, for example, 10% of the feature size of the diamond structure, 50 nm. Apparently such a goal is difficult to achieve with the current doping scheme. In practical fabrication, we found another problem, i.e., phase separation. When pinpoint photopolymerization occurs, the doped nanoparticles can't be included in the polymer phase; they are isolated from the polymerized structures. Such problems are not so serious when the doped polymer is used at macroscale, for example, photopainting and pasting. In order to solve the above problems, an alternative technical route has been proposed and tried by us. Instead of directly doping metal oxide nanoparticles, we employed metal ions that are better soluble to organic solvents
Femtosecond Femtosecond laser laser fabrication of of three-dimensional three-dimensional metallic metallic micro-nanostructures micro-nanostructures
291 291
[15]. Metal ions could thus be dispersed into photopolymers with a molecular uniformity. After incorporated into photopolymerized structures, the ions could be further converted to metal or metal oxide with an appropriate post-fabrication process. We doped gold ion into polyvinyl alcohol (PVA). The sample was prepared by separately dissolving 330-mg Chlorauric (HAuCL*) powder and 2-g PVA (Mw, 500) into 10-ml deionized water. And then the two aqueous solutions were mixed and agitated for 24 hours with a magnetic stirrer. Finally a near-2-(xmthick uniform film was spin-coated on a cover glass and dried. In order to study the material response to fs laser irradiation, we utilize two-beam interference method [Fig. 2] to expose the sample film. The laser system that was employed was a Ti: Sapphire laser of 780-nm wavelength, 80-fs pulse width at a repetition rate of 82 MHz. The laser beam power was attenuated by a variable reflective neutral density filter and split with a prism into two equal-power beams, one of which is optically delayed so that the two beams can coherently interact with the sample film. Fig. 3 (a) shows the scanning electron microscopic (SEM) image of a two-beam interfered pattern, which was recorded on a round area of 60-p.m diameter with a total power of 0.924 W for 31 seconds. (a) (a)
CCD Monitor
Laser
M
Beam Beam Splitter Splitter
M
M OL
M M
M
Sample
M
M
Fig. 2 Two-beam interference system used to expose the gold-ion-doped polymer film, from which grating-like structures are expected.
Absorption spectra were measured to distinguish the composition variation of the irradiated area. PVA film is colourless [curve A in Fig. 3 (b)] and Au(ffl)doped PVA film is transparent with a faint yellow colour caused by an absorption peak at 324 nm, as shown by the curve B in Fig. 3 (b). The 324-nm peak is attributed to the absorption of AuCl; complex ions, a finger feature that shows the existence of Au(III) ions [16]. After irradiation, this peak disappeared, and instead a broad band centred at 585 nm emerged [curve C and the inset of Fig. 3 (b)]. It has been reported11 that colloidal gold nanoparticles had the
292
H.-B. Sun et al. H.-B. Sunetal.
similar absorption band, which was attributed to surface plasmon absorption, the collective oscillation of free conduction electrons induced by the detecting light field in a metal particle [17]. A free-mean path theory [18] indicates that the bandwidth and peak position of surface plasmon absorption strongly depends on the particle size. When the gold particle diameter, 2R, is small (i.e., 2R< 25 nm), only dipole absorption contributes to the extinction cross section. For larger sizes, multipole modes like quadruple and octopole absorption and scattering may also dominate in the extinction cross-section, leading to a red-shift of the absorption band with increasing particle size. It has been observed [19] that the absorptions of 22-nm and 99-nm gold particles were centred at 521 nm and 575 nm, respectively. The inter-correlated disappearance of 324-nm peak and immergence of 585-nm peak indicates the photochemical conversion of Au(III) ions to atom status, which may be in forms of nanoparticles. The FWHM (full width at half maximum) of 185 nm and the absorption peak position that was maximized by a linear extrapolation to around 115 nm imply a relatively broad distribution of particle sizes.
(a)
(b)
200 300 400 500 600 700 800
Wavelength (nm) Fig. 3. (a) SEM image of interference-created pattern on the surface of ion-doped polymer thin film, (b) Absorption spectra of the irradiated samples.
To get direct proof of creation of nanoparticles and study their precise size distribution, an atomic force microscope (AFM) was utilized to image the twophoton laser patterned grating-like structures. From the sequentially reduced scanning areas, it is resolved that the grating lines in Fig. 3 (a) consist of small particles. The smallest discernible beads, judging from Fig. 4 (a) (low-left), have diameters around 15 nm. The distribution of particle size basically follows the light intensity distribution: larger particles appear at the bright interference fringes while at the dark region, both particle size and number density are small
Femtosecond Femtosecond laser laser fabrication of of three-dimensional three-dimensional metallic metallic micro-nanostructures micro-nanostructures
293 293
[Fig. 4 (b)], agreeing with the prediction of absorption spectrum [Fig. 3]. From a statistical analysis of the images, the average lateral particle diameter is around 65 nm, smaller than that predicted by the absorption spectra. Even smaller is their average height, 11 nm. Since there is sufficient space, ~ 1 (im, between grating lines, the much smaller vertical particle size is not an artificial feature due to the limitation of AFM tip geometry, but a real feature of the samples, (a) (b)
Fig. 4. AFM images of areas 20 X 20 mm (up-left), 10X10 mm (up-right), 5X5 mm" (lowright), and 1X1 mm2 (low-left)
(d)
2
4
6
8 Mm 0.0 1.0
Fig. 5. (a) Laser interferential intensity profile; (b) and (c) profiles of AFM image of nonablated and ablated particle arrays; (d) AFM image of pure PVA film irradiated with a regeneratively amplified laser pulse; and (e) AFM profile of an ablated single line. In (b), (c) and (e) the vertical axes denote height in nanometer and horizontal axes, distance in micrometer.
294
H.-B. Sun et al.
Exposing an unadulterated PVA film with the same interferometric patterns as that used for Fig. 3 (a) causes no visible response either by increasing the power of oscillator laser to the maximum (1.5 W) or by increasing the exposure time by several orders. A regeneratively amplified single pulse irradiation induced strong ablation [Fig. 5(d)] of the pure polymer surface, however no particle-like debris was found. Therefore grain formation in Figs 3 and 4 should be attributed to the doped Au ions. From the cross-sectional profile of grating lines [Fig. 5 (b)], it is found that there is no background bulge corresponding to the bright fringes [Fig. 5(a)], but particles protrude from a flat surface. These facts indicate that the gratings resulted from a gentle in-situ photochemical reaction. A possible mechanism is, when exposed to interferentially enhanced femtosecond laser pulses, electrons of AuCl ~4 complex ions located at bright fringe region gain energy by means of two-photon absorption and are dissociated, leading to the formation of Au atoms at some random sites where they are favoured by local energy fluctuation. Enhancing the irradiation may lead to new nucleation centres or growth of existing nanoparticles. The two processes compete with each other but the latter is favoured, as it requires less energy than nucleation. The average 11-nm grain height implies that the actual particle size may be larger than that is observed from AFM image, and most of their volume hides beneath the sample surface, where they may contact or even combined as seen in the surface case [white circles in Fig. 4 (a), low-left]. A simple calculation shows a mean particle diameter of 106 nm if a spherical particle shape is assumed, agreeing with the spectrum deduction (115 nm). The patterns in Fig. 4, not discernible under a microscopic CCD monitor, undergo a dramatic increase of contrast within less than 1 second when increasing the exposure time beyond a critical value, t . Figure 6 (a) shows the AFM image of the sample irradiated after this one additional second (32 s, 0.924 W). Compared with Fig. 4 (a), the grating lines become sharper and the average height increases by 5 times, to around 50 nm. The rapid change of morphology is reasonably attributed to laser ablation. However, different from Fig. 2 (e) and the general scenario of surface ablation, where material at the bright fringes [the white line in Fig. 2 (e)] is removed, protrusions out of background are formed at the light intensity maximums. This may be due to, in the course of gold nanoparticle formation, the removal of the polymer, which generally possesses much lower decomposition temperatures, TD, (~ 200°C for PVA) than metals. This assumption was confirmed by a single-line exposure, from which a particle-gathered strip sided by two channels were found [Fig. 5 (e)]. Therefore, the above ablation process is different from those induced by direct photon absorption by materials to be ablated, but by a novel process, which is assisted and launched through an energy transfer from absorption centres to the target material, here the polymer matrix.
Femtosecond laser fabrication fabrication of three-dimensional metallic micro-nanostructures
(b) (b)
8 r
Ln[t(s)]
SHtfj 2223223555
7 7
6 I
6
5
2 µm
h4 4
Ln [ (s)]
(a) a)
295 295
3 1.0 0.4 0.6 0.8 0.8 1.0 Power (W) Fig. 6. (a) AFM image of selectively ablated surface, and (b) ablation induction time versus laser power.
Under a certain metal ion concentration, the critical exposure time, % decreases exponentially with the increase of laser power, e.g., x = 1670 s (P = 0.193 W), 170 s (0.624 W), and 32 s (P = 0.924 W). We summarized the dependence with a formula: r = T0/exp(/>/P0), where the constant %- 2.73xl0 4 s is the induction time at exposure power of P(P 0.134 W. Po and ^should reflect the fact that the surface plasmon absorption band is red-shifted with the particle size increases so that even single-photon absorption becomes to play a significant role.
Laser
o
Color filter
CCD Resin
Galvano mirror set
PBS
Lamp
Substrate (on PZT)
Reflection mirror
Fig. 7. Optical setup for laser micro-nanofabrication. The femtosecond laser used here is the same as that for multi-beam interference.
296
H.-B. Sun et al. H.-B. Sunetal.
Since the metal nanoparticles could be produced by femtosecond laser irradiation. Even complicated patterns are drawn inside the doped polymer film. Fig. 7 shows the system used for single focused beam writing. Three digits " 1 " , "2" and " 3 " were sequentially written in different depths, with layer space of 4 \im [Fig. 8(a)]. Fig. 8 (b) is the top view of the written patterns, and Fig. 9 is the scattering readout by laser scanning microscope.
Fig. 8 Patterns of gold nanoparticles induced by single focused femtosecond laser beam scanning
Fig. 9. Laser scanning microscope images of the digits written at different depths as shown in Fig. 8. The scattered light is used as readout signal.
micro-nano structures Femtosecond laser fabrication fabrication of three-dimensional metallic micro-nanostructures
297
An important type of structures we used for testing the laser micronanofabrication technology is so-called photonic crystal (PhC) structure, which is 3D artificial structure with periodically varying refractive indexes.
Fig. 10. Laser scanning microscopic images of logpile PhC structures that are written inside gold-ion doped polymer film. The lines are mainly consisted of gold nanoparticles.
Fig. 11. The logpile PhC structures viewed from cross-sections of different orientations.
Figures 10 and 11 are the logpile PhC lattice, the layer-by-layer images [Fig. 10] and cross-sections along different directions [Fig. 11]. The feature lines, according to the previous study are composed of metal nanoparticles.
298
H.-B. Sun et al.
3. THREE-DIMENSIONAL MICRO-NANOSTRUCTURES COMPOSED OF NANOPARTICLES OF METAL DIOXIDES TiO2 is an important type of photonic materials, of which the refractive index of the bulk is larger than 2.4. Introduction to TiO2 to polymer matrix would greatly enhance the photonic bandgap effect due to the increased the index contrast. By a strategy similar to the case of gold ion, we chose Titanium (IV) (Ti4+) ions. Titanium (IV) ethoxide was selected as the source of Ti4* ions, and was mixed with methacrylic acid to obtain complexes of titanium (IV) acrylate. 2,2diethoxy-acetophenoe also was added into the solution as photoinitiator. Then, the complex solution was introduced into the photopolymerizable resin and fully mixed. The molecular structures and chemical reactions, which occurred in fabrication and ensuing processes, are shown in Fig. 12.
- * Ti(OCH 2 CH 3 ) m (OOCCH=CH 2 ) n OCH 2 CH 3 O TKOCH 2 CH a ) m (OOCCH=CH 2 ),,
(C)
Fig. 12. Chemical reactions associated with the Ti4* ion (a) doping, (b) polymerization, (c) exposure to air, and (d) after heating.
Shown in Fig. 13 (a) is the logpile PhC structure fabricated by this technical means. Its design is quite similar to the pure polymer structures as reported earlier, however, it includes titanium component. Fig. 13 (b) is the magnified image after the structure was exposed to air for more than 10 hours. In this course, the titanium ions at the surface layer were converted to a titanium hydroxide due to the humidity of air. The conversion continues under 250°C heating for 3 hours, until all ions become oxides. The increased surface roughness as seen in Fig. 13(c) is largely due to the formation of the Titanium oxide nanoparticles. Also the polymer rod was significantly thinned, which is one of the important reasons to cause the shift of the transmission valley in the corresponding PhC structures. A simple calculation indicates mat the refractive index change of the material is not less than 5%.
micro-nano structures Femtosecond laser fabrication fabrication of three-dimensional metallic micro-nanostructures
299
(a)
(c)
Fig. 13. Logpile structure PhCs prcxluced by two-photon photopolymerization of the titanium ion doped photopolymer. (a) the as-fabricated structure, (b) 10 hour after exposure to air, and (c) additionally heated for 2 hours at 250°C.
4. METAL NANOSHELLS So far, we introduced how to induce the metallic components to photopolymers to create metal-organic hybrid materials. The technologies are, however, difficult to be used for preparing continuous metal structures because crystal nucleuses of metals appear and grow into isolated particles in random sites of the irradiated volume. These particles are less possibly combined into a bulk due to the limited doping concentration and due to steric resistance to particle movement. Noticing the fact that growth rates are generally much larger than nucleation rate and the former depends on the number of nucleation centres, F. Stellacci et al. solve this issue by introducing nanoparticles seeds into the composite to be irradiated [20]. The metal seeds are equivalent to high concentration of ions. Metal ions would be consumed mostly by growth on the existing nanoparticles instead of producing new dispersed nucleuses. In the current research, we utilize electroless plating of metals [21-23]. A common feature of these approaches is that no electrode and no high temperature (< 100 °C) are needed. The template of our micro-nanodevices is polymer, of which the glass transition temperature is generally lower than 200 °C, for the particular case of SCR 500 we are using, 150 °C, and the substrate glass is nonconductive. This makes electroless plating an ideal technology to coat the polymer skeleton. The technical route of coating is as follows. The polymer objects are dipped into Sn2+ ions solution that functions as the initial reductant. The Sn2+ has excellent surface affinity to the polymer surface due to electrostatic interaction so that after they are uniformly adhered to the sample surface. The surface-
300
H.-B. Sun et al. H.-B. Sunetal.
modified sample is then immerged into the metal ion solution, which is prepared by dissolving metallic salts to water, in this research, AgNO3. At the beginning of the reaction, Ag+ is reduced by Sn2+ to Ag metal and coated to the sample surface. The Sn2+ itself is oxidized to Sn4+. In order that the metal deposition reaction proceeds after the exhaustion of the surface reductant Sn2+, another type of reductant of potassium sodium tartrate tetrahydrate (C4H4KNaO64H2O) were mixed beforehand to the metal salt solution. In order to avoid reduction reaction in solution, ammonia solution is also added to form the complex of [Ag (NH3)2]+, which was then reduced to metal silver through Ag autocatalytical process. It is worthy to mention that a surface to be coated is generally chemically etched to intentionally induce roughness for a better attachment of the coated the layer. In our fabrication, the pinpoint-depicted surface roughness of less than 10 nm works well for this purpose.
(c)
Silver*
Fig. 15. Illustration of the electroless plating technology, which is used to coat the micronanostructures fabricated by two-photon photopolymerization. (a) hi SnCk solution a polymer structure is coated by tin ions (Sn + ). (b) In a plating solution of metal ions (Ag+) are reduced to metal particles by tin ions (Sn2+). (c) Silver seeds on the early-coated surface grow up with the assistance of the potassium sodium tartrate reductant in metal ions solution.
Experimentally, 1-g AgNO3 was dissolved into 50-ml pure water, producing a metal salt solution. 2.0-ml and 2.5-ml ammonia solutions were separately added to a set of solution as prepared above for low- and highconcentration plating, respectively [Fig. 16]. With the gradual addition of the ammonia solution, the metal ion solution roils, presenting white-brown colour, and ultimately a transparent solution is obtained. 5-g potassium sodium tartrate is dissolved to 50-ml pure water in the other beaker, which, after fully dissolved, is mixed with [Ag(NH3)]+ complex solution as obtained above. After attaching of Sn + ions to a polymer structure by dipping it to the solution of 0.49-g tin chloride in 100-ml pure water, the polymer structure is placed in the mixture of the plating solution.
Femtosecond laser fabrication fabrication of three-dimensional metallic micro-nanostructures
301
Fig. 16. SEM image of silver-coated polymer micro-coils. The deposition was conducted at room temperature, 23 °C. The amount of the ammonia used for forming [Ag(NHj)2]+ in the plating solution are (a)2.0 ml and (b)2.5 ml, respectively. Note the scatter silver nanoparticles in regions other than the coil surface in (a), and the larger surface roughness in (b). According to these results, an idea amount of ammonia is roughly determined to 2.25 ml. The scale bar is 1 um.
In order to test the effectiveness of the above approach, a coil structure of 350-nm diameter is fabricated. The SEM images of the plated structure are shown in Fig. 16. In Fig. 16(a), where ammonia concentration is low, silver particles appear on the substrate, implying the insufficient stabilization of the silver ions. In contrast, no such undesired particles were deposited when the ammonia concentration is high [Fig. 16(b)]. However, the higher surface roughness, or in another word the generated larger-sized particles should also be ascribed to a relatively high-speed reduction reaction due to the high ammonia concentration. A suitable complex concentration should be characterized as the production of good surface roughness while the number of scattered silver nanoparticles is minimized. The ammonia amount was thus determined to 2.25ml. The coated thickness dependence on the dipping time in the current experimental condition is investigated. A linear relation in the first 10 minutes is obtained, where the coating rate is around 4.2 nm/minutes. A nanometer-order accuracy of the thickness control is thus obtained because the growth of metal film thickness is slow compared with plating duration. Compared with the above coil structures, the diamond-lattice PhCs are more complicated. Whether the obtained plating condition supports the complicated periodic structures is still a problem that needs investigation. The answer is given by Fig. 17, which shows the SEM image of 8 X 8 X 2 period diamond structures with different coated thickness.
302
H.-B. Sun et al. H.-B. Sunetal.
".---..•
, "^
(d)
(f)
Fig. 17. Metal-coated diamond-lattice PhC structures with layer thicknesses of (a) 14 nm, (b) 20 nm, (c) 23 nm, (d) 40 nm, (e) 55 nm and (f) 66 nm.
Reflection experiments were conducted on the metal plated diamond-lattice PhCs since almost no transmission signal could be detectable for these structures. 60-, 60
50 66.5nm
Reflectance [%]
40
55nm Polymer structure
30
20
14nm 41nm
20nm 23.5nm
10
0 2000 2000
27.5nm
3000 3000
4000 Wavenumber [cm [cm"-1]]
5000 5000
6000 6000
Fig. 18. Reflection spectra for the metal-coated diamond-lattice PhCs with different layer thickness.
micro-nano structures Femtosecond laser fabrication fabrication of three-dimensional metallic micro-nanostructures
303 303
The reflection peak of a polymer-air lattice, which has been ascribed to photonic band gap effect from the simultaneous transmission and reflection spectra and from the scaling effect, occurs at the wave number of 3700 cm"1 with the peak intensity 12 % and the peak width 0.3um (Fig, 18). The peak intensity decreases with the increase of the thickness of deposited metal film from 14nm to 20nm, and disappears when the thickness reaches 23,5 nm. The reduction of the reflected power is reasonably ascribed to material absorption by metal. In this case photonic band structure remains similar to the pure polymer one and the introduction of metal on the surface layer could be treated as a small perturb in the band gap calculation. However, with coated layer thickness increased, the photonic band structure undergoes a significant change, which finally behaves more and more like a pure metal lattice. Fig. 18 shows that the original 3700 cm"1 band gap appearing at polymer structure disappears and a broad band gap opens in the range from 2000 cm"1 to 4000 cm"1. We believe surface plasmon wave absorption coupling to the structure should be responsible to the appearance of the large reflection peak, on which theoretical work is ongoing for a reliable assignment of the features. 5. CONCLUSION Different from other chapter, we didn't discuss much on the characteristics of the metal nanoobjects, but instead we focus on how to technically produce complex-shaped 3D metal and metal-oxide structures. Two typical technologies are metal ion doping followed with photoreduction of ions in photopolymer structures created by laser micro-nanofabrication, and electroless coating of laser fabricated 3D polymer structures. Each of the technologies will find use in fields like nanophotonics and biophotonics. An immediate task of our future research is investigation of surface plasmon property of the fabricated structures and on this basis, design of novel nano plasmonic devices. REFERENCES [1] J. B. Pendry, Phys. Rev. Lett., 85 (2000) 3966. [2] D. R. Smith, W. J. Padilla, D. C. Vier, S. C. Nemat-Nasser, and S. Shultz, Phys. Rev. Lett., 84 (2000) 4184. [3] C. G. Parazzoli, R. B. Greegor, K. Li, B. E. C. Koltenbah, and M. Tanielian, Phys. Rev. Lett, 90 (2003) 107401. [4] A. A. Houck, J. B. Brock, and I. L. Chuang, Phys. Rev. Lett., 90 (2003) 137401. [5] A. Alu and N. Engheta, IEEE T. Microw. Theory, 52 (2004) 199. [6] Y. Inouye and S. Kawata, Opt. Lett., 19 (1994) 159. [7] T. Ichknura, N. Hayazawa, M. Hashimoto, Y. Inouye, and S. Kawata, Phys. Rev. Lett., 92 (2004)220801. [8] J. Feng, T. Okamoto, and S. Kawata, Opt. Lett. 30 (2005) 2302. [9] S. Maruo, O. Nakamura, and S. Kawata, Opt. Lett. 22 (1997) 132.
304
H.-B. Sun et al.
[10] H. B. Sun, S. Matsuo, and H. Misawa, Appl. Phys. Lett., 74 (1999) 786. [11] S. Kawata, H.-B. Sun, T. Tanaka, and K, Takada, Nature, 412 (2001) 697. [12] H.-B. Sun and S. Kawata, Adv. Polymer Sci., 170 (2004) 169. [13] K. Takada, H.-B. Sun, and S. Kawata, Appl. Phys. Lett., 86 (2005) 071122. [14] K. Kaneko, H.-B. Sun, X. M. Duan and S. Kawata, Appl. Phys. Lett., 83 (2003) 2091. [15] X. M. Duan, H. B. Sun, K. Kaneko, and S. Kawata, Thin Solid Film, 453 (2004) 518. [16] P. Xu and H. Yanagi, Chem. Mater., 11 (1999) 2626. [17] G. C. Papavassiliou, Prog. Solid State Chem., 12 (1980) 185. [18] U. Kreibig and C. v. Fragstein, Z. Phys., 224 (1969) 307. [19] S. Link and M. A. El-Sayed, J. Phys. Chem. B, 103 (1999) 4212. [20] F. Stellacci, C. A. Bauer, T. M.-Friedrichsen, W. Wenseleers, V. Alain, S. M. Kuebler, S. J. K. Pond, Y. Zhang, S. R. Marder, and J. W. Perry, Adv. Mater., 14 (2002) 194. [21] P. Jiang, J. Cizeron, J. F. Berton, and V. L. Colvin, J. Am. Chem. Sac., 121 (1999) 7957. [22] Z. Chen, P. Zhan, Z. Wang, J. Zhang, W. Zhang, N. Ming, C. T. Chan, and P. Sheng, Adv. Mater., 16 (2004) 417. [23] G. O. Mallory, J. B. Hajdu, Electroless Plating: Fundamentals and Applications, American Electroplaters and Surface Finishers Society, Orlando, 1990.
Handai Nanophotonics, Volume 2 S. Kawata and H. Masuhara (Editors) © 2006 Elsevier B.V. All rights reserved.
305 305
Chapter 17
Nanophotolithography based on surface plasmon interference T. Ishiharaa'b and X. Luoa "Frontier Research System, RIKEN, Wako 351-0198, Japan b
Department of Physics, Tohpku University, Sendai 980-8578, Japan
1. INTRODUCTION Photolithography has been a practical technique for microstracture fabrication. The ability of parallel process is the key issue for the large-scale production. Because the minimum feature size is limited to the half of wavelength, the conventional photolithography is forced to employ shorter wavelengths as smaller features are required. When the wavelength reaches deep ultra violet, further shortening becomes extremely difficult because of the imaging optics as well as materials. One of the alternative ways is to utilize near field distribution combined with visible or UV light sources. In 1999, Alkaisi et al. demonstrated that the near field at narrow slits on an opaque conformable mask can generate sub-wavelength pattern onto a resist [1]. Line width of 50 nm and grating with 140 nm period were achieved for a broad band light of which dominant component was 436nm. Because the fine features were generated at the edge of the mask, the pattern suffered from relatively large line edge roughness. Blaikie and McNab numerically analysed a photolithography scheme in which a near-field and an interference lithography techniques are combined [2]. When the excitation is close to the lowest cut-off of a Cr 270-nm-period grating structure embedded in material with refractive index of 1.6, the incident light (450nm) is able to generate a high-contrast standing wave pattern in the near field region with period of 135 nm. Goodberlet et al. also used conformable Cr mask to make sub-50nm lines, holes and posts using 220 nm light from an arc-lamp [3]. They demonstrated that isolated patterns are affected by interference effect. In 2004, Luo and Ishihara demonstrated that sub-wavelength periodic structure can be fabricated on a resist layer with metallic photonic crystal slabs [4]. In order to enhance surface plasmon characteristic, we used Ag film instead of Cr
306
T. Ishihara and X. Luo
used by the predecessor. Stronger resonance is expected in this case, as was first discovered by Ebbesen et al. in 1998 [5] and later confirmed by many others. This method was named as Surface Plasmon Resonant Interference Nanolithography Technique (SPRINT). Further we have reported the sub-wavelength pattern transfer in thicker Ag film with spacer layer in [6] and increase of spatial resolution for thin film mask [7]. In this paper, we will discuss the advantage of SPRINT and the outlook of this technique. Some comments are also given on our former publications. 2. SIMULATION In order to design masks for SPRINT demonstration, we first estimated the first resonant period Ax for Hg g-line (436 nm, 2.843 eV) from surface plasmon dispersion for a semi-infinite metal. k x
^(o e,e2{m) .m c £, + ^(0))
=
2x A,
(1)
where £\ and £2 are dielectric constant of substrate and Ag, respectively. The frequency dependent dielectric constant was taken from [8]. The period was chosen to be 300 nm so that the first resonance is excited by 436nm light incident to the normal to the sample. Next we calculated transmission spectra for several widths of slits, keeping the film thickness to be 60nm. We found it has resonant transmission when the slit width was 60 nm. At the transmission resonance condition, the far field is enhanced. Since the (quasi-) eigenmodes consist of a series of spatial frequency components, near field components are also enhanced at the transmission resonance as is seen in the simulation. We performed FDTD calculation for the structure to see the near field distribution. Periodic boundary condition was applied for x direction, while perfectly matched layers were assumed at the domain boundaries in the z-direction. Incident light is propagating downwards (+z) with polarization vector in x-direction. Figure l(a) shows the electric field distribution. The 60 nm thick Ag mask is located below the quartz substrate. The field intensity is plotted as a pseudo-color in linear scale. Note that the high field region is not only at the openings of the mask, but also at the flat part of the lower interface of the mask. In fact at the opening of the grating slits, the field is not uniform with a singularity at the corners due to the accumulation of the induced charges. The associated field has short decay length. The fine features especially clear in the mask layer seem to be an artifact originated from the finite domain size in the simulation. In order to make a quantitative discussion, the field intensity at a certain distance from the lower interface is plotted in Fig.l (b). At 10 nm from the lower interface, the electric field is highest at the edge, while by 25 nm, the field intensity at the corner is damped to the level of other interference fringes. Thus
Nanophotolithography based on surface plasmon plasmon interference
307
appropriate thickness of an index-matched spacer layer allows providing a mask that gives more-or-less periodic gratings. Although the intensity decreases as the distance, good contrast is kept even at 40nm. Light intensity distribution
Arb. unit
I
3.5 3.0 2.5 2.0 1.5
SO
1.0 (tun)
-180 -120
-120
120 Location (ran)
-60
0
1 SO
I
0.5
60
Position (rim) Fig.l. (a) Electric field distribution in the 60 nm-thick Ag mask. P-polarized light is propagating from the top. (b) Electric field profile below the Ag mask. The field is measured at 10,15,20,25,30,35 and 40 run from the mask-resist interface.
Figure 2 shows field distribution calculated by FDTD of two resonant structures for 436 nm with period of A2 and A4.
A2
A4
The slit was fixed to be narrow, one tenth of the metallic strip width.
308
T. Ishihara and X. Luo Electric Field Distribution
Electric Field TTHft
iv
D = 41 nrn
= 82om 990 nm
Fig.2. Electric field distribution for structures resonant to 436 nm light, (a) the second order resonance (450 nm). (b) forth order resonance (900 nm).
In the first case, the period of 450 nm gives 2nd resonance, which is seen from the four lobes in the lower side of the mask. Similarly, the plasmon is resonantly excited when a structure has the period of 900 nm. Eight lobes are seen in this case, corresponding to the fourth resonance. Note that there is no significant field at the opening of the slit, which is generally the case when the opening width is not resonant to the incident light. Irrespective of the order of the resonance, the feature size of the interference lobes are similar, which is given by the surface plasmon wavevector at the frequency of exciting light. As for the example presented in Fig.l, the field pattern is different because of the additional resonance to the slit opening. Thus slit width is one of the geometrical parameters which can be used for fine tuning In Ref. [6], polarization-resolved field distribution was discussed. It cannot be justified, however, to claim that the changing polarization is a useful tool for fabricating complex patterns. Although field distributions for Ex and Ez component of the field in the structure are quite different, we do not have simple means to excite them separately. 3. EXPERIMENT The mask for SPRINT was prepared as following. The features on the mask have a periodicity of 300 nm and the opening is 1/5 of the periodicity as was assumed in the simulation. The substrate of the mask is a 2 mm-thick unconformable quartz plate, on which electron beam resist (Zeon ZEP 520) and thin conducting layer were spin-coated. The latter is to prevent charging up during the electron beam patterning. After developing the pattern, a 60 nm-thick Ag film was evaporated in vacuum and a grating was obtained by a lift-off process.
Nanophotolithography based on surface plasmon plasmon interference
309
Excitation light
Photoresist
Fig. 3. (a) Schematic setup of SPRINT, and (b) SEM image of resist pattern
Figure 3 (a) shows schematic structure of our setup for photolithography. Onto the mask, 20 nm of photo resist (Tokyo Oka TSMR) was spin coated as a spacer layer which also protects Ag from oxidization. The photo-exposure was carried out in a conventional mask aligner (Mikasa MA-10). The light source was an unpolarized broadband light from Hg discharge lamp, with the effective dominant component being 436 nm. A 12 s exposure followed by a 60 s development in the developer (NMD-W 2.38%) was used to obtain the resist structure, of which SEM image of the resist pattern after development is shown in Fig. 3(b). Grating structure with period of 300 nm was recorded on the resist. The line width was 50 nm which corresponds to 1/6 of the excitation light source. The structures seen in the SEM image corresponds well with the FDTD simulation in Fig.l (a). 4. DISCUSSIONS Here we discuss on the resolution limit of SPRINT. As we have seen above, when the excitation light excites a standing wave in the periodically arranged grooves in the mask, well defined light intensity pattern is located under the mask. In the grating structure, there are basically 2m lobes where m is the order of the
310
T. Ishihara and X. Luo
resonance. Besides, the corners at the grating corners may contribute to additional field maxima. When the photon energy approach to the surface plasmon energy defined by e2 (m) = -e1
(3)
the wavevector diverges as can be seen from Eq. (1), which results in large effective index of refraction. In reality, any metal has imaginary part in its dielectric constant, and the dispersion does not diverge. Yet good metals such as Ag have a chance to provide much larger wavevector which can give higher spatial resolution. Besides, if the metal film is thin enough, surface plasmons at two interfaces interact with each other to form symmetric and anti-symmetric modes [9]. The symmetry is referred in terms of magnetic field H which is perpendicular to the wavevector. The .//-symmetric mode propagates longer because it so widely spreads in the dielectric that it does not see the imaginary dielectric constant of the metal. The H-anti-symmetric mode is concentrated in the interfaces, and has larger wavevector which is given by
tanh(a2d) = k = kr-iki where rfis the metal thickness, £j= e3 and e% (m) are dielectric constant for the clad and metal layers, respectively. The dielectric constant of Ag was taken from ref. [8]. Dispersion relation for the H-antisymmetric mode in dielectric/Ag/dielectric structure is drawn in Fig.4 for several Ag layer thicknesses. (Some of the dispersion relations for finite thicknesses shown in Ref. [7] were incorrect near the surface plasmon energy due to the insufficient numerical accuracy.) Generally the ^-vector reaches its maximum at the surface plasmon energy. Furthermore as can be seen from the figure, the wavevector increases rapidly as the thickness decreases. For g-line (436 nm, shown in blue line in Fig.4), a 10 nm-thick Ag film gives Re (k) =100 1/um, which corresponds to a spatial resolution of A/2 = %/Re(k) = 30 nm. If the photon energy was 3.3 eV, it might be possible to achieve 10 nm according to the scheme. The large imaginary part in this region, on the other hand, may deteriorate the interference. Detailed investigation is required to demonstrate such a high resolution. As can be seen in the figure, we cannot expect significant enhancement of wavevector for thickness of 60 nm at g-line frequency. The reason why small features were achieved in Fig.l may be ascribed to the slits in the film. Note that the dispersion relation shown in Fig. 4 is valid only for a flat Ag film. Just as the dispersion relation is dependent on thickness, structures in nano-scale can also modify the dispersion relation significantly due to the charge distribution at the surfaces, which is responsible for fine features appearing in the simulation.
Nanophotolithography based on surface plasmon plasmon interference
311 311
g-line 1=1.7 / Ag (Johnson-Christy) / n=1.7 H- antisymmetric mode
100
200
300
Re Fig.4. Dispersion relation for the short range surface plasmon mode in Ag thin films. The numbers on the curve refer to the thickness of the film
As we discussed in Ref. [4], it is useful to separate light collection and near-field generation functions, for achieving arbitrary patterns. The first interface of the metal film should be periodic for collecting the propagating light in a free space, while only near-field distribution is what we concern for photolithography. 5. OUTLOOK The essential origin of the high-spatial resolution in SPRINT is ascribed to the large wavevector of quasi-eigenstates in the structure. SPRINT is by definition utilizing interference for generating patterns. Therefore it is naturally applied to periodic structures. As for fabricating arbitrary structures, highly damped modes in the short range surface plasmons in the thin film may serve for higher spatial resolutions. Since g-line is not the best choice for exploiting the modes in Ag, extensive simulation as well as experiments is indispensable for future development. Recently Luo et al reported demonstration of SPRINT for imperforated Ag structures [10]. They also proposed to use the technique for parallel nano data recording. It is definitely one of the important directions of surface-plasmon-based nano-optics. 6. SUMMARY Grooves with feature size of 50 nm have been fabricated by means of Surface Plasmon Resonant Interference Nanolithography Technique (SPRINT) applied
312
T. Ishihara and X. Luo
for Ag grating excited by Hg-line (436 nm). The successful nanofabrication is ascribed to resonant excitation of surface plasmon in the metallic nanostructure. The Finite Difference Time Domain (FDTD) simulation was able to reproduce the experimental observation. When the metal is thin enough to form coupled surface plasmon polaritons, the short range mode may give even larger wavevectors which may be utilized for even higher spatial resolution. ACKNOWLEDGEMENT We wish to express our gratitude to N. A. Gippius for helping to calculate dispersion relation of metallic thin films. REFERENCES [1] M. M. Alkaisi, R. J. Blaikie, S. J. McNab, R. Cheung, and D. R. S. Gumming, Appl. Phys. Lett., 75 (1999)3560. [2] R. J. Blaikie and S. J. McNab, Appl. Opt, 39 (2000) 20. [3] J. G. Goodberlet and H. Kavak, Appl. Phys. Let, 81(2002) 1315. [4] X. Luo and T. Ishihara, Appl. Phys. Lett., 84 (2004) 4780. [5] T. W. Ebbesen, H. L. Lezec, H. F. Ghaemi, T. Tbio, and P. A. Wolff, Nature, 391 (1998) 667. [6] X. Luo and T. Ishihara, Jpn. J. Appl. Phys., 43 (2004) 4017. [7] X. Luo and T. Ishihara, Opt. Express, 12 (2004) 3055. [8] P. B. Johnson and R. W. Christy, Phys. Rev. B, 6 (1972) 4370. [9] F. Yang, J. R. Samples, and G. W. Bradberry, Phys. Rev. B, 44 (1991) 5855. [10] X. Luo, J. Shi, H. Wang, and G. Yu, Mat. Phys. Lett. B, 18 (2004) 945.
313 AuthorIndex AbeS. Asahi T. Bravo-Abad J. Duan X. M. Feng J. Fujii A. Fukui M. FutamataM. Garda-Vidal F. J. Haraguchi M. Hashimoto K. Hayashi S. Hayazawa N. H'DhiliF. Hosokawa C. Ikebata A. Inouye Y. Ishida A. Ishibara T. Itoh T. Kajikawa K. Kaneko K. Kawata S. KikkawaY.
185 219 15 289 231 155 31 101 15 31 197 151 81 231 275 197 81 153 305 197 185 289 81,231,289 197
Kobayashi T. Lopez-TejeiraF. LuoX. Martfn-Moreno L. MaruyamaY. Masuhara H. Niidome Y. Okamoto T. Okamoto T. Ozaki Y. Pileni M. P. Sarychev A. K, Shalaev V. M. Shvets G. Simonen J. Sotokawa Y. SunH.B. Takahara J, Tsuboi K. Uwada T. Watanabe H. Yamada S. Yoshikawa H.
55 15 305 15 101 219, 275 255 31 231 197 247 3 3 3 231 185 289 55 185 219 81 255 275
This Page is Intentionally Left Blank
315 Subject Index absorption loss, 237 adenine nanocrystals, 82 apertureless metallic probe tip, 81 artificial magnetism, 4 band edge, 233 beam effect, 27 blinking, 102,122 gharge transfer (CT), 198 CT transition, 198 CTband, 204 gollision frequency, 45 golloidal aggregate, 275 goncentrie metallic shell, 34 diamond lattice, 301 diffraction limit 55 dispersion relation of surface plasmon, 232 232 Drude model, 7 electrical SPR, 4 electroless plating enhancement factor, 87,101,153, 265 extinction spectrum, 285 extraordinary optical transmission (EOT), 15 Fermi surface 61 field enhancement, 160 gap mode, 142 high-pressure induced Raman spectroscopy 98 hot particle, 108 hot site, 141 incoherent coupling devices, 75 Kerr material 45 Kerr nonlinearity, 45 laser ablation, 294 laser-induced transformation, 268 laser micro-nanofabrication, 289 laser scanning microscope, 296 left-handed material (LHM), 3 localized plasmon resonance, 262 long-range surface plasmon (LRSP), 240 low-dimensional optical wave, 56 magnetic plasmon resonance (MPR), 4,7 magnetic polarizability, 7 metamaterials, 3 metal waveguide, 55
Monte Carlo, 282 nanoaggregate, 201 nanoantenna, 8,9 nanocrystal, 247 nano-optical waveguide, 56 nanorod (NR), 255 nanosphere, 255 nanosphere lithography, 105 nanowell, 171 NDgap, 68 negative dielectric (ND), 59 negative dielectric waveguide, 55 negative permeability (NP), 59 negative refraction, 3 optical bistability, 39 optical magnetism, 3 optical storage, 268 optical switching, 39 optical trapping, 275 organic light-emitting diode (OLED), 243 photolithography, 305 photonic crystal (PhC), 297,232 photonic bandgap, 298 potential depth, 287 plasma frequency, 7 plasma angular frequency, 34 plasmonic bandgap, 231 plasmonic crystal, 231 plasmonies, 55 permittivity, 7 propagation length 75 radiation loss, 237 ring-breathing mode, 87 second harmonic generation, 186 self-assembly, 275 self-assemble monolayer, 156 self-focusing, 42 short-range surface plasmon (SRSP), 240 silver-adenine complex, 89 single aperture, 15 single molecule sensitivity (SMS), 101 single molecule detection (SMD), 101 skin depth, 16 split-ring resonator (SPR), 4,12 superstructure, 247
316
Subject Index
surface electromagnetic mode, 26 surface impedance boundary condition (SIBC), 16 surface leaky mode 20 surface enhanced Raman scattering (SERS), 81,101,141, 197 SERS band, 208 surface plasmon (SP), 15, 64 surface plasmon polariton, 31,55,101 surface plasmon resonant interference nanolithography, 306 tip-enhanced near-field Raman scattering (TERS), 81 two-beam interference, 291 Wood's anomaly 22 wavenumber surface 58
