Nanoplasmonics (Vol. 2): From Fundamentals to Applications

Citation preview

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) be completed on-line via the Elsevier homepage (http:// 1865 853333, e-mail: [email protected]. [email protected]. Requests may also be 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 of this work may be reproduced, stored in in aa retrieval system or transmitted in in any form or by Except as outlined above, no part of 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 is available from the Library B r i t i of sh Congress. Library. catalogue record is 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

ofANSI/NISO Z39.48-1992 (Permanence of ofPaper). @ The paper used in this publication meets the requirements requii-ements of 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 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 0,1

7v

nn 0,0 400

A 1 \

^ 500

VI

600

700

800 800

900 900

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

"E w

0,6

1E-3

1E-6

1E-9

1E-12

760

780

800

820

840

Wavelength (nm)

0,4

E

0,2 0,0 770 770

780 780

790 790

800 800

810 810

820 820

Wavelength (nm) (nm) Wavelength 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

0,01 0,01

τ12 τ23

1E-3 1E-3

1E-4 1E-4 740

760

780

800

820

840 840

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 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 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

FF

£• -

5 4 3 2 1 0 0 =

-80

-60 -60

-40 -40

-20 -20

0 0

20 20

40 40

60 60

80 80

θ8 (deg)

Fig. for a x ig. 9. z/, Radial xvauicu component vuixipuiiviii of ui the nil/ Poynting x ujutiLiK vector v w i u i evaluated bvaiumbu in iix the uib far leu field iiviu versus v u a u o angle ai 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

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

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

30 20 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 'o| vs. h for the linear response at points A and A', results, we evaluated 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

i•

1

'

1

_ /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
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)

Laser

Controller

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 1000

1200 1200

1400 1400

w c

(b) (b) 723

c

CO

(0 (0

l V '\WVi^^^^^J^ v v w ^ ^ - l ^ ^ 600 600

800 800

1000 1000

1200 1200

1400 1400

[cm-–1] Raman shift [cm 1

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. Watanabeetetal. al. H. Watanabe

450 420 390 360 330 300 270 240 210 180 150 120 90 60 30 0

(a) (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 942 961 797 790 723 733

"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. Ad-NIO Ad-N7 Ad-N3 Ad-Nl Adenine UB3LYP Crystal" B3LYP 2.946 2.502 2.502 2.524 -112.2 0.1 0.0 0.7 1.342 1.339 1.335 1.347 1.338 1.341 N1-C2 1.334 1.335 1.338 C 2 -N 3 1.332 1.334 1.328 1.334 1.342 1.334 1.340 1.338 1.336 N 3 -C 4 1.399 1.396 1.382 1.395 1.396 1.397 Q-C5 1.405 1.413 1.410 1.410 C 5 -C 6 1.409 1.409 1.384 1.383 1.382 1.385 1.385 C 5 -N 7 1.387 1.312 1.308 1.308 1.308 1.312 N 7 -C 8 1.308 1.380 1.381 1.380 1.380 C 8 -N 9 1.371 1.367 1.371 1.349 1.349 1.346 1.337 1.353 C6-N10 N,-C 2 -N 3 128.2 128.4 127.9 128.0 129.0 128.5 111.4 112.4 110.8 111.5 111.7 111.7 C 2 -N 3 -C 4 126.9 126.6 126.9 125.9 126.7 126.7 N 3 -C 4 -C 5 C4-C5-C6 115.8 116.1 116.3 116.4 116.9 116.0 111.2 110,4 111.0 111.3 C 4 -C 5 -N 7 111.3 110.7 104.2 C 5 -N 7 -C g 105.0 104.3 104.2 103.9 104.2 112.6 113.3 113.2 113.2 113.8 113.2 N7-C8-N9 122.3 123.1 122.4 122.6 123.4 122.3 C 5 -C 6 -N, o -19.4 -0.6 Ni-Cs-Nio-H 0.0 -0.5 -1.3 3.9 0.5 3.5 3.7 M -0.3 3.0 3.4 3.2 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) (c)

4 1 |iO m u -–1 10AÅ4aamu

,. , , . II

'l f

-

\

I

i

f

|

i

(d) (d)

i

(0 (0

-

| (e) (e)

I . . I . I

DC -

I I

L.i.

600 600

(f) (f)

800 800

h , 1000 1000

(g) (g)

-

II

1200 1200

1400 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 970 968 968 939 941 942 895 899 903 846 853 852 804 800 799 722 724 721 680 679 681 664 664 663 615 615 615 572 572 574 544 542 536 529 531 535 517 520 522 514 529 535 300 301 297 275 276 277 215 220 224 163 175 188 51 149 157 85 43 51 41 40 35

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

c

CO

E CO

-

13701370

(a) (a)

1360 7 i1360

a. - -&

ν88 v a

-

_

13501350 13401340 1330, 1330 750 750, 740 * 740

ν221 1 V

730730 ( 720 720710710

Binding energy [kcal/mol]

700700

0) £

.—

o

-8

•

(b)

-4

—o--.

•

Ad-N3 Ad-N3 Ad-N7 Ad-N7

0 4

2

2.5 2.5

3 3

∞

Bond Bond distance 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 PC PC

CCD

Monitor

Raman/Optical Image age

Raman spectra

Notch filter Xe lamp

Polychromator CCD "

λ /2 plate XI2 plate Mobile Mobile mirror mirror

PC 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 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 -A— Bare ;g=20nm

j-

Tjrfl'H

10x10*

400

n M

800

l-

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. (b) D=2nm D=2nm (b)

(a) D=5nm D=5nm (a)

•« •

(c) D=0.5nm D=0.5nm (c)

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 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?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

4

absorbance

3)

0.02 0S>2 x4 2 0.01 0.01

0

400 400

500 500

600 700 600 700 wavelength // nm nm wavelength

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. Photoinduced Charge Charge Separation Separation Photoinduced Ps-l hv hν Electron

+

hν \

Ps-ll Electron l' .Transport Transport

4H+++O 4H O22 \ A 2H22 O -JMn

'P680 P680

A00 . . Transport Transport

A

\

...:.

ft

"

Excess Electron

P700 Damage

Oxygens Reactive Oxygens

Photoinhibition Photoinhibition

immobilization on gold surface microfluidics system substrates H2O

Calvin Cycle Cycle Calvin FNR 2NADPH

RBISCOetc RBISCO etc CO2

2NADP

sugar

suaar

Artificially Supplemented Supplemented Artificially Redox Enzyme Enzyme Redox substrates

useful products

SP excitation excitation SP useful products products useful

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

1/2 O2

N

O

PS-I

thylakoids

+ 2 e-

Cl

H2O

ONa O Na

DCIP λmax=600 nm

DCIP2H DCIP2H

colorless colorless

λex= = 632.8nm 632.8nm 12.5 12.5 mW mW

X= 600 nm λ= LED

θ=73

mt, DCIP [/DCIP~|

Multichannel Multichannel Spectrometer 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 30

60 90 60 90 time// min time

120 120

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-

NADP NADP

?imax =340 nm nm λ max=340

- 2 e- PS-II

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/ min

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 of plasmon be calisation of photoillumination direct photoillumination interaction between the the neighboring nanowells

metal

metal metal glass substrate

expected functions

future applications

light antenna fluorescence enhancement addressing nanovessel nanoarray nanoarray µ -TAS H-TAS fibre fibre sensor sensor light light emitting emitting device device etc. 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

400 400

500 6 00 500 600 wavelength // nm wavelength nm

700 700

800 800

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

532 532 nm nm 1µ VJ 3+ Nd -YAG Laser Nd3+ -YAG Laser

85nm 85 nm

nanowells

t

500 nm

0.0 nm

Nikon Eclipse 300

position controllable pinhole turret

Perkin Elmer SPCM module AQR-14 APD module

x60 × 60 NA1.2

0 nm

0 nm

500 nm

1000 nm nrr 1000

Near-Field Image

BP520 BP520

1000 nm

10m • mV V -10

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)

R

ε 1(Ω)

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(, 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 Wavelength (nm) (nm) Wavelength

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].

249

toself-organization of of5nm silver nanocrystals Intrinsic properties due to

' • •: " jtif

s*fc

2 nm

P* rt r

*• :•••

2 nm>y

sjfk*f

'

•

:~f •? +

'

# Si?.**: • i . v . . r

^

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

/** •
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)

Absorbance

(d)

100 nm

1.2 t• 1.2 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;