Zikang Tang, Ping Sheng Nanoscale Phenomena Basic Science To Device Applications Lecture Notes in Nanoscale Science and Technology 2007

Nanoscale Phenomena
Basic Science to Device Applications

LECTURE NOTES IN NANOSCALE SCIENCE AND TECHNOLOGY
Series Editors:
Zhiming M. Wang, Department of Physics, University of Arkansas, Fayetteville, AR, USA
Andreas Waag, Institut für Halbleitertechnik, TU Braunschweig, Braunschweig, Germany
Gregory Salamo, Department of Physics, University of Arkansas, Fayetteville, AR, USA
Naoki Kishimoto, Quantum Beam Center, National Institute for Materials Science, Tsukuba, Ibaraki,
Japan
Volumes Published in this Series:

Volume 1: Self-Assembled Quantum Dots, Wang, Z.M., 2008
Volume 2: Nanoscale Phenomena: Basic Science to Device Applications, Tang, Z., and Sheng, P., 2008
Forthcoming Titles:
Volume 3: One-Dimensional Nanostructures, Wang, Z.M., 2008
Volume 4: Epitaxial Semiconductor Nanostructures, Wang, Z.M., and Salamo, G., 2008
Volume 5: B-C-N Nanotubes and Related Nanostructures, Yap, Y.K., 2008
Volume 6: Towards Functional Nanomaterials, Wang, Z.M., 2008
Nanoscale Phenomena
Basic Science to Device Applications
Zikang Tang and Ping Sheng

Editors
Hong Kong University of Science and Technology

Clear Water Bay, Hong Kong
Zikang Tang Ping Sheng
Department of Physics and Institute Department of Physics and Institute
of Nano Science and Technology of Nano Science and Technology
Hong Kong University Hong Kong University
of Science & Technology of Science and Technology
Clear Water Bay Clear Water Bay
Kowloon, Hong Kong Kowloon, Hong Kong
ISBN-13: 978-0-387-73047-9 e-ISBN-13: 978-0-387-73048-6
Library of Congress Control Number: 2007931614
c 2008 Springer Science+Business Media, LLC

All rights reserved. This work may not be translated or copied in whole or in part without the written
permission of the publisher (Springer Science+Business Media, LLC, 233 Spring Street, New York,
NY 10013, USA), except for brief excerpts in connection with reviews or scholarly analysis. Use in
connection with any form of information storage and retrieval, electronic adaptation, computer software,
or by similar or dissimilar methodology now known or hereafter developed is forbidden.
The use in this publication of trade names, trademarks, service marks and similar terms, even if they are
not identified as such, is not to be taken as an expression of opinion as to whether or not they are subject
to proprietary rights.
Printed on acid-free paper.
9 8 7 6 5 4 3 2 1
springer.com
Preface
The Third International Workshop of the Croucher Advanced Study Institute (ASI)
on Nano Science and Technology: from Basic Science to Device Applications, was
held at Hong Kong University of Science and Technology from 8 January to 12
January 2007. The first and second workshops took place in January 1999, and
January 2002, respectively.
Collected in this volume are 20 articles, 16 from invited talks and four from
contributed presentations. The speakers are from the United States, Europe, Japan,
Korea, Chinese Mainland, Taiwan, and Hong Kong. During the workshop, the
vivid presentations captured the audience’s attention not only with the great poten-
tial of nanotechnology, but also brought out the relevant underlying science. In
Prof. Supriyo Datta’s talk, a rather unique “bottom-up” view of electrical con-
duction was presented which is particularly relevant to nanoscale devices. He
raised several thoughtful questions on transport physics such as dissipation and
entanglement that could be operative in these devices. Professor Morinobu Endo
described the advanced technology of large-scale synthesis of nanostructured car-
bons, with emphasis on novel applications that have already had an impact on
our daily lives. His presentation impressed on us that nanotechnology applica-
tions are already here, with far-reaching implications ahead. Professor Ming-Chou
Lin described his group’s successful integration of silicon-based nano devices into
biomolecular technology. With this integrated technology, the potential of silicon
nanowires as a detector for DNA hybridization is demonstrated. Professor Vivian
Yam delivered an excellent review of on her group’s recent work on functional
molecular materials. Through rational design and judicious functionalization and
assembly strategies, she has demonstrated that many transition-metal molecular
complexes possess structure-dependent light emitting characteristics with poten-
tial application as templates for the preparation of nano-sized materials. A rather
unique presentation by Prof. Ping Sheng on nanoscale hydrodynamics showed that
it is only recently that continuum hydrodynamics can quantitatively reproduce the
nanoscale flow characteristics as simulated by molecular dynamics, and in so doing
revises the traditional non-slip boundary condition. Taken together, these papers
from the Third Croucher ASI gave a snapshot of the frontier research carried out
over the past few years, and demonstrates that great progress has indeed been
achieved both in fundamental research and in applications of nanoscience and
technology.
v
vi Preface
We would like to express our gratitude to all of the authors for their excellent
contributions. Special thanks go to Ms. Helen Lai and her team for the success-
ful organization of this workshop, and to the Croucher Foundation for its financial
support.
Zikang Tang and Ping Sheng

Kowloon, Hong Kong
May, 2007
Contents
Preface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . v
Part I Nano Structured Carbons and their Applications
1 Science and Technology of Nanocarbons

Morinobu Endo . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
2 Catalytically-Grown Carbon Nanotubes and Their Current

Applications
Morinobu Endo . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
3 Transparent Conducting Films by Using Carbon Nanotubes

Hong-Zhang Geng and Young Hee Lee . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
4 Raman Spectroscopy on Double-Walled Carbon Nanotubes

Wencai Ren and Hui-Ming Cheng . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
5 Interface Design of Carbon Nano-Materials for Energy Storage

Feng Li, Hong-Li Zhang, Chang Liu and Hui-Ming Cheng . . . . . . . . . . . 41
6 Formation Mechanism of 0.4 nm Single-Walled Carbon Nanotubes

in CoAPO-5 Single Crystals
J. P. Zhai, I. L. Li, Z. M. Li, J. T. Ye and Z. K. Tang . . . . . . . . . . . . . . . . . 49
Part II Quantum Dots and Molecular Spintronics
7 Nanodevices and Maxwell’s Demon

Supriyo Datta . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
8 Manipulating Electron Spins in an InGaAs/InAlAs Two-

Dimensional Electron Gas
C. L. Yang, X. D. Cui, S. Q. Shen, H. T. He, Lu Ding, J. N. Wang,
F. C. Zhang and W. K. Ge . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
vii
viii Contents
9 Continuum Modelling of Nanoscale Hydrodynamics

Ping Sheng, Tiezheng Qian and Xiaoping Wang . . . . . . . . . . . . . . . . . . . . 99
10 Defect in Zinc Oxide Nanostructures Synthesized by a

Hydrothermal Method
A. B. Djurišić, K. H. Tam, C. K. Cheung, Y. H. Leung, C. C. Ling,
C. D. Beling, S. Fung and W. K. Chan . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117
Part III Nano Materials Design and Synthesis
11 Towards Surface Science Studies of Surfaces Formed by Molecular

Assemblies Using Scanning Tunneling Microscopy
Chen Wang and Shengbin Lei . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133
12 Electronic Transport Through Metal Nanowire Contacts

Y. H. Lin, K. J. Lin, F. R. Chen, J. J. Kai and J. J. Lin . . . . . . . . . . . . . . . . 139
13 Synthesis and Properties of Quasi-One-Dimensional Nitride

Nanostructures
Yong-Bing Tang, Dai-Ming Tang, Chang Liu, Hong-Tao Cong
and Hui-Ming Cheng . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 149
14 Electron Energy-Loss Spectroscopy for Nannomaterials

C. H. Chen and M. W. Chu . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 179
15 Fabrication of Photovoltaic Devices by Layer-by-Layer

Polyelectrolyte Deposition Method
Wai Kin Chan, Ka Yan Man, Kai Wing Cheng and Chui Wan Tse . . . . . . 185
16 Optical Properties of Arrays of Iodine Molecular Chains Formed

Inside the Channels of AlPO4 -5 Zeolite Crystals
J. T. Ye and Z. K. Tang . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 191
Part IV Molecular Electronics
17 Quantum Manipulation at Molecule Scale

J. G. Hou . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 201
18 Silicon-Based Nano Devices for Detection of DNA Molecules

M. C. Lin, C. J. Chu, C. J. Liu, Y. P. Wu, C. S. Wu, and C. D. Chen,
I. C. Cheng, L. C. Tsai, H. Y. Lin, Ya-Na Wu, Dar-Bin Shieh, H. C. Wu
and Y. T. Chen . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 209
19 From Simple Molecules to Molecular Functional Materials

and Nanoscience
Vivian Wing-Wah Yam and Keith Man-Chung Wong . . . . . . . . . . . . . . . . 217
Contents ix
20 First-Principles Method for Open Electronic Systems

Xiao Zheng and GuanHua Chen . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 235
Author Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 245
Subject Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 247

Contributors
C. D. Beling, Department of Physics, F. R. Chen, Department of Engineering

The University of Hong Kong, and System Science, National Tsing
Pokfulam Road, Hong Kong Hua University, Hsinchu, 30013,
Taiwan
W. K. Chan, Department of Chemistry,
The University of Hong Kong, Y. T. Chen, Department of Chemistry,
Pokfulam Road, Hong Kong National Taiwan University, Taipei 106,
Taiwan
Wai Kin Chan, Department of
Hui-Ming Cheng, Shenyang National
Chemistry, The University of Hong
Laboratory for Materials Science,
Kong, Pokfulam Road, Hong Kong
Institute of Metal Research, Chinese
Academy of Sciences,
Cheng Hsuan Chen, Center for Shenyang, 110016, China
Condensed Matter Sciences,
National Taiwan University, Taipei, Kai Wing Cheng, Department of
Taiwan Chemistry, The University of Hong
Kong, Pokfulam Road, Hong Kong
Chii Dong Chen, Institute of Physics,
Academia Sinica, Nankang I. C. Cheng, Department of Material
11529, Taipei, Taiwan Science, National University of Tainan,
Tainan, 70005 Taiwan
G. H. Chen, Department of Chemistry,
C. K. Cheung, Department of Physics,
The University of Hong Kong,
The University of Hong Kong,
Pokfulam Road, Hong Kong
X. D. Cui, Department of Physics, The M. W. Chu, Center for Condensed

University of Hong Kong, Matter Sciences, National
Hong Kong, China Taiwan University, Taipei, Taiwan
Chen Wang, National Center for C. J. Chu, Institute of Physics,

Nanoscience and Technology, Academia Sinica, Nankang
Beijing, 100080, China 11529, Taipei, Taiwan
xi
xii Contributors
Hong-Tao Cong, Shenyang National H. T. He, Department of Physics and

Laboratory for Materials Institute of Nano-Science
Science, Institute of Metal Research, and Technology, The Hong Kong
Chinese Academy of Sciences, University of Science and Technology,
Shenyang, 110016, China Hong Kong, China
Supriyo Datta, School of Electrical J. G. Hou, Hefei National Laboratory

for Physical Sciences at Microscale,
and Computer Engineering,
University of Science and Technology
Purdue University, West Lafayette, IN
of China, Hefei, Anhui 230026,
47907, USA
P.R. China
Lu Ding, Department of Physics and J. J. Kai, Department of Engineering

Institute of Nano-Science and System Science, National Tsing
and Technology, The Hong Kong Hua University, Hsinchu 30013, Taiwan
University of Science and Technology,
Hong Kong, China Young Hee Lee, Department of
Nanoscience and Nanotechnology,
A. B. Djuriši¢, Department of Physics, Department of Physics, and Center for
The University of Hong Kong, Nanotubes and Nanostructured
Pokfulam Road, Hong Kong Composites, Sungkyunkwan Advanced
Institute of Nanotechnology,
Sungkyunkwan University, Suwon
Morinobu Endo, Faculty of
440–746, R.O. Korea
Engineering, Shinshu University,
4-17-1 Wakasato, Nagano-shi Y. H. Leung, Department of Chemistry,
380–8553, Japan The University of Hong Kong,
S. Fung, Department of Physics, The
University of Hong Kong, Feng Li Shenyang, National
Pokfulam Road, Hong Kong Laboratory for Materials Science,
W. K. Ge., Department of Physics Academy of Sciences, Shenyang
and Institute of Nano-Science and 110016, China
Technology, The Hong Kong University
I. L. Li, School of Engineering and
of Science and Technology,
Technology, Shenzhen
Hong Kong, China
University, Shenzhen, China
Hong-Zhang Geng, Department of Z. M. Li, Department of Physics,

Nanoscience and Nanotechnology, Hong Kong University of Science and
Department of Physics, and Center for Technology, Clear Water Bay, Kowloon,
Nanotubes and Nanostructured Hong Kong, China
Composites, Sungkyunkwan Advanced
Institute of Nanotechnology, Y. H. Lin, Institute of Physics, National
Sungkyunkwan University, Suwon Chiao Tung University, Hsinchu 30010,
440–746, R.O. Korea Taiwan
Contributors xiii
J. J. Lin, Institute of Physics & S. Q. Shen, Department of Physics,

Department of Electrophysics, National The University of Hong Kong, Hong
Chiao Tung University, 1001 Ta Hsueh Kong, China
Road, Hsinchu 30010, Taiwan
Ping Sheng, Department of Physics
Ming-Chou Lin, Institute of Physics, and Institute of Nano Science &
Academia Sinica, Nankang 11529, Technology, Hong Kong University of
Taipei, Taiwan Science and Technology, Clear Water
Bay, Kowloon, Hong Kong China
K. J. Lin, Department of Engineering
Shengbin Lei, National Center for
and System Science, National Tsing
Nanoscience and Technology
Hua University, Hsinchu 30013, Taiwan
Beijing, 100080, China
H. Y. Lin Department of Molecular Dar-Bin Shieh, Institute of Oral
Science and Engineering, National Medicine and Institute of
Taipei University of Technology Molecular Medicine, National
Cheng-Kung University
C. C. Ling, Department of Physics,
The University of Hong Kong, Dai-Ming Tang, Shenyang National
Pokfulam Road, Hong Kong Laboratory for Materials Science,
C. J. Liu, Institute of Physics, Academy of Sciences
Academia Sinica, Nankang Shenyang 110016, China
11529, Taipei, Taiwan
K. H. Tam Department of Physics,
Chang Liu, Shenyang National The University of Hong Kong,
Laboratory for Materials Science, Pokfulam Road, Hong Kong
Academy of Sciences, Shenyang Z. K. Tang, Department of Physics,
110016, China Hong Kong University of Science and
Technology, Clear Water Bay, Kowloon,
Hong Kong, China
Ka Yan Man, Department of Chem-
istry, The University of Hong Kong,
Y.-B. Tang, Shenyang National
T. Qian, Department of Mathematics, Academy of Sciences, Shenyang
The Hong Kong University of 110016, China
Science and Technology, Clear Water
Bay, Kowloon, Hong Kong, China L. C. Tsai, Department of Material
Science, National University
Wencai Ren, Shenyang National of Tainan, Tainan 70005, Taiwan
Institute of Metal Research, Chinese Chui Wan Tse, Department of
Academy of Sciences, Shenyang Chemistry, The University of Hong
110016, China Kong, Pokfulam Road, Hong Kong
xiv Contributors
J. N. Wang, Department of Physics Vivian Wing-Wah Yam, Department

and Institute of Nano-Science and of Chemistry, The University of
Technology, The Hong Kong University Hong Kong, Pokfulam Road,
of Science and Technology, Hong Kong
Hong Kong, China
C. L. Yang, Department of Physics and
X. Wang, Department of Mathematics, Institute of Nano-Science
The Hong Kong University of and Technology, The Hong Kong
Science and Technology, Clear Water University of Science and Technology,
Bay, Kowloon, Hong Kong, China Hong Kong, China
Keith Man-Chung Wong, Centre for J. T. Ye, Department of Physics,

Carbon-Rich Molecular and Hong Kong University of Science and
Nano-Scale Metal-Based Materials Technology, Clear Water Bay, Kowloon,
Research, and Department of Hong Kong, China
Chemistry, The University of Hong
Kong, Pokfulam Road, Hong Kong J. P. Zhai, School of Engineering
P. R. China and Technology, Shenzhen University,
Shenzhen and Department of Physics,
Hong Kong University of Science and
Y. P. Wu, Institute of Physics,
Technology, Clear Water Bay, Kowloon,
Academia Sinica, Nankang 11529,
Hong Kong, China
Taipei, Taiwan
Hong-Li Zhang, Shenyang National
C. S. Wu, Institute of Physics, Laboratory for Materials Science,
Academia Sinica, Nankang 11529, Institute of Metal Research Chinese
Taipei, Taiwan Academy of Sciences,
Shenyang 110016, China
Ya-Na Wu, Institute of Oral Medicine
and Institute of Molecular F. C. Zhang, Department of Physics,
Medicine, National Cheng-Kung The University of Hong Kong, Hong
University Kong, China
H. C. Wu, Department of Chemistry, X. Zheng, Department of Chemistry,

National Taiwan University, The University of Hong Kong,
Taipei 106, Taiwan Pokfulam Road, Hong Kong, China
Chapter 1
Science and Technology of Nanocarbons
Morinobu Endo∗
1.1 Introduction
Nanocarbons can be defined as carbon materials built from sp2 bond building block
in a nanoscale, and include various forms of carbons in the range from fullerene,
carbon nanotube to nano-porous materials. Among them, one-dimensional carbon
nanotubes have attracted a wide range of scientists due to their unique morphology
and nano-sized scale, furthermore their versatile expected applications, in which the
challenge is that how to utilize their intrinsic properties of carbon nanotubes, such
as mechanical, thermal and electrical properties. It is no doubt that the recent explo-
sive carbon research is due to the discovery of nano-scale carbon materials. After
the identification of fullerenes and carbon nanotubes as new allotropes of carbon
element in the end of the 20th century, chemists, physicists, scientists and engineers
started to work intensively due to their extraordinary mechanical and electronic
properties. In general, carbon is a unique element that has three types of bonding
nature: sp3 , sp2 and sp bonds, which result in diverse allotropes such as diamond,
graphite, carbyne and fullerene. So, carbon differs chemically from Si, Ge, and Sn
(the other group IV elements, which all have sp3 bonding). Carbon has plenty of
potential since it is the lightest atom in group IV of the periodic table. Now it is
becoming the main element instead of silicon for electronic applications in this era,
as Professor H. W. Kroto predicted, “21st century is the carbon age”. The 19th
century can be remembered as an iron age while the 20th century is recognized
to be the century founded on silicon technology. Now much attention is paid on
carbon “C” as one of the promising candidates and leading element in the 21st
century. To date, carbons have contributed to welfares of human being as a “silent
force behind the science”. This is greatly due to the diverse structural forms and
functions of carbons. Inherent advantage of carbons materials, originating from the
extraordinary ability of the chemical element carbon to combine itself and other
chemical elements in different ways, is blooming in the 21st century. Tailoring of
carbon structure, that is, controlling the physicochemical properties of carbon mate-
rials on the nanometer scale, will be core technology for obtaining novel carbons
∗ Faculty of Engineering, Shinshu University, 4-17-1 Wakasato, Nagano-shi 380–8553, Japan
Z. Tang, P. Sheng (eds), Nano Science and Technology. 3

C Springer 2008
4 M. Endo
with new and extraordinary functions. The function and the added value of novel
carbons based on nanotechnology will provide the industry of our society with
a great business opportunity. Day by day, the request for revitalizing the global
economy is growing. The only way to answer for this request is the technologi-
cal innovation. According to model of Kondratieff the economist, new science will
create a new industry and technology, and as a result, it will bring the wave of big
economic activity. The generalization of science, technology and products in the
20th century, which led to the recession of world market, is said to be one of the
reasons to the present economy slump. Now it is the time for new carbons based on
carbon science to give rise to the creation of new functional materials, and finally
technological innovation. As literally one of the major newspaper reported, “Carbon
is hot now”. Many scientists are being “hot” on carbon materials. The author and
co-workers have been studying for many years how to control the nano-structure
in order to enhance the function and performance, and develop new functions of
carbon materials. Our ultimate goal is to be a “carbon tailor” that creates novel
carbon structure by making use of sp3 , sp 2 and sp or hexagonal carbon structure
to meet our need just like an experienced tailor making a made-to-order outfit. In
this account, new carbons and their physical properties and novel functions from the
viewpoint of nano-structure and its control will be presented with an emphasis on
my own works.
1.2 Carbon Nanotubes
Nano-sized carbon nanotubes with hollow core were observed when hydrocarbons
were decomposed in the existence of nano-sized catalyst such as iron at higher tem-
perature (Oberlin et al., 1976). Interestingly, vapor-grown carbon fibers (VGCFs)
shows very unique morphology such as an annular ring texture, and also contain
single walled carbon nanotube or multi-walled carbon nanotube in the core part of
the fibers (Fig. 1.1). Through a precise control of the synthetic conditions such as
size of metal particles, it is possible to tailor the diameter, crystallinity and also
the orientation of the cone angle with regard to the tube axis (Endo et al., 2001a;
Terrones et al., 2001). As compared with those of arc discharge or laser vaporization,
Fig. 1.1 FESEM images of vapour grown carbon fibers containing (a) SWNT and (b) MWNT
1 Science and Technology of Nanocarbons 5
synthesis of carbon nanotube via a catalytic chemical vapor deposition (CVD)

method has shown to be more controllable and profitable for the large-scale produc-
tion. As a result, the diversity of microstructure and morphology of one-dimensional
carbons are expected to be useful in various fields such as electronic devices, com-
posites, biomaterials, batteries etc.
1.3 Anode Material for Lihitum Ion Battery
Since commercialization of Li-ion batteries (LIB) by Sony in 1992, they have been
used in a wide range of mobile electronic devices, and now became leading product
of secondary batteries (five hundred million battery cells per year). It is well known
that carbon or graphite materials have been used as a main component of anode
electrode in LIB. Actually, the advance of carbon science has been supporting the
tremendous improvement of capacity of LIB for past 10 years (10% per year). These
days, the new generation cellular phones having diverse functions with high data
transfer rate require further enhancement of the battery performance. When consid-
ering the diverse and still growing LIB, e.g. from new type of cellular phone, high
performance notebook, PC (personal computer), to power supply of 42V-vehicle and
hybrid vehicle, it is critical to develop a higher performance LIB with advanced car-
bon materials as anode. In this sense, the outstanding mechanical properties and the
high surface to volume ratio (due to their small diameter) enable carbon nanotubes
to be incorporated as an additive in the commercial lithium-ion battery systems
(Endo et al., 2001b). Astonishingly, only by adding small amount of high purity
carbon nanotubes to the synthetic graphite, the cyclic efficiency was sustained at
almost 100% up to 50 cycles. At higher concentrations, the nanotubes interconnect
graphite powder particles together to form a continuous conductive network.
1.4 Electrode Material for Electric Double Layer Capacitor
Electric double layer capacitor (EDLC) has been expected as the electric energy
storage device of the next generation, such as hybrid electric vehicle (HEV), fuel-
cell vehicle, and power back-up system. Since the new interpretation of the pore
structure is required in order to understand the mechanism of EDLC, we have devel-
oped an effective method for analyzing the nanopore with the help of transmission
electron microscope (TEM) observations and image processing. For capacitance on
EDLC, the theory for the conventional parallel plate condenser has been applied.
The structure of the pore has been assumed to be the ideal structure such as slit
and tube etc. However, actual pores that are quite different from these ideal mod-
els should be determined according to the TEM observations and image analysis
process. The result of novel method we have developed is consistent with that of
conventional gas adsorption method (Endo et al., 2001c,d,e). Furthermore, it is very
helpful to understand the pore structure because the novel method analyzes the pore
6 M. Endo
Fig. 1.2 Tree of nanocarbons with basic properties and applications
structure from different viewpoint. The novel method will be used as a powerful
tool to develop a new carbon electrode for EDLC with high specific capacitance.
Consequently, novel method shows some clues to control the nano-structure of car-
bon, and to get a final product with high performance. This enables us to effectively
form an activated-carbon electrode with ideal pore size, which has the right room
for the solvated ions in an aqueous and aprotic solvent system. All parameter in the
process of preparing the capacitor (i.e. the choice of precursor, pyrolysis, activation
etc.) should be considered for the end use. The creation of the new industry in rela-
tion with carbon based EDLC will be realized by the cooperation of carbon-related
science and technology.
1.5 Conclusion and Prospect
It is estimated that the world market of nano materials is a twenty seven thousand
billion yen market. This sum doubles when we incorporate the correlated indus-
tries and as a result, exceeds present automobile industry of this country. Novel
carbons with new function and technology, constructed by tailor-made technology
1 Science and Technology of Nanocarbons 7
in nanoscale, will function as a starting point of industrial innovation. The technical

development until the 20th century was mainly raised from the market needs. In the
21st century when there is no clear market needs, needs provoking-type technical
innovation based on material technology is awaited. Applications fields of new car-
bons such as fuel cell, new type battery, composite with novel function, gas storage,
and purification of water/air is considered to be the fundamental technology of the
century. Conversion from conventional gasoline based vehicle to hybrid vehicle and
fuel cell based vehicle, supported by new carbon technology, will cut the gas con-
sumption to less than a half than usual. It has been expected that HEV will continue
to increase its share around 10–20% of the domestic vehicles sales in 2007–2008,
and at the same time, nano carbon industry is also expected to bloom. Widespread
impact on nanocarbon-related technology on automobile industry and the housing
field and promotion of the structural reform of the industry and economy is main
target. Japanese consumers are renowned to be “green consumers” in the world,
and this fact has much to do with the expansion of environmentally friend vehicle.
Also, nano electron mechanical system will be developed by the contribution of
nanocarbons as a sensor and actuator. New carbon as a key material of the 21st
century will grow to a great industry and contribute to environmental, energy and IT
both directly and indirectly. Intellectual production of nano carbon, that is, scientific
results has been accumulated to a considerable amount, and it is time to the creation
of the value (Fig. 1.2). When considering the recent research trend of carbons, we
believe that this is not anymore a dream, and we are on the starting line.
References
Endo, M., Kim, Y.A., Fukai, Y., Hayashi, T., Terrones, M., Terrones, H., Dresselhaus, M.S., 2001a,
Comparison study of semi-crystalline and highly crystalline multiwalled carbon nanotubes.
Applied Physics Letter, 79, pp. 1531–1533.
Endo, M., Kim, Y.A., Hayashi, T., Nishimura, K., Matushita, T., Miyashita, K., Dresselhaus, M.S.,
2001b, Vapor-grown carbon fibers (VGCFs): Basic properties and their battery applications.
Carbon, 39, pp. 1287–1297.
Endo, M., Maeda, T., Takeda, T., Kim, Y.J., Koshiba, K., Hara, H., Dresselhaus, M.S., 2001c,
Capacitance and pore-size distribution in aqueous and nonaqueous electrolytes using various
activated carbon electrodes. Journal of Electrochemical Society, 148, pp. A910–A914.
Endo, M., Kim, Y.J., Takeda, T., Hayashi, T., Koshiba, K., Hara, H., Dresselhaus, M.S., 2001d,
Poly (vinylidene chloride)-based carbon as an electrode material for high power capacitors
with an aqueous electrolyte. Journal of Electrochemical Society, 148, pp. A1135–A1140.
Endo, M., Kim, Y.J., Maeda, T., Koshiba, K., Katayama, K., 2001e, Morphological effect on the
electrochemical behavior of electric double-layer capacitors. Journal of Material Research, 16,
pp. 3402–3410.
Oberlin, A., Endo, M., Koyama, T., 1976, Filamentous growth of carbon through benzene decom-
position. Journal of Crystal Growth, 32, pp. 335–349.
Terrones, H., Hayashi, T., Munoz-Navia, M., Terrones, M., Kim, Y.A., Grobert, N.,
Kamalakaran, R., Dorantes-Davila, J., Escudero, R., Dresselhaus, M.S., Endo, M., 2001,
Graphitic cones in palladium catalysed carbon nanofibers. Chemical Physics Letter, 343,
pp. 241–250.
Chapter 2
Catalytically-Grown Carbon Nanotubes
and Their Current Applications
Morinobu Endo∗
2.1 Introduction
One-dimensional carbon nanotubes have attracted a particular interest due to their

unique morphology, nano-sized scale, novel physico-chemical properties and fur-
thermore, their versatile applications (Dresselhaus et al., 1995; Iijima 1991; Oberlin
et al., 1976; Saito et al., 1998). Carbon nanotubes could be visualized as rolled
sheets of graphene with man-sized dimension (sp2 carbon honeycomb lattice). They
could be either single-walled with diameters as small as about 0.4 nm, or multi-
walled consisting of nested tubes with outer diameters ranging from 5 to 100 nm.
The constituent cylinders within multi-walled carbon nanotubes (MWNTs) may
possess different chiral structures. In particular, the electronic properties of single-
walled carbon nanotubes (SWNTs) vary from semiconductors to metals, depending
upon the chiral angle (the way the hexagons are positioned with respect to the tube
axis) (Dresselhaus et al., 1995; Saito et al., 1998).
In this chapter, we will describe large-scale synthetic methods for producing
high-purity carbon nanotubes and their current usages in various fields from elec-
trochemical system, multifunctional filler to medical devices. In addition, we will
report the selective synthesis on highly pure and crystalline double walled carbon
nanotubes (DWNTs) by the right combination of the catalytic chemical vapor depo-
sition (CVD) method and the subsequent purification steps.
2.2 Large Scale Production of Carbon Nanotubes
In order to use carbon nanotubes in novel devices, it is necessary to produce these

materials with a high crystallinity in large-scale at economic costs. In this context,
the catalytic CVD method is considered to be the most optimum to produce bulk
amounts of carbon nanotubes, particularly with the use of a floating catalyst method
(Oberlin et al., 1976; Endo, 1988). This technique is more controllable and cost
∗ Faculty of Engineering, Shinshu University, 4-17-1 Wakasato, Nagano-shi 380-8553, Japan

C Springer 2008
10 M. Endo
efficient when compared to arc-discharge and laser ablation methods. Regarding the
bulk production of nanotubes for industrial applications, it is important to mention
that in the early 1990s, Hyperion Catalysis International, Inc. (Cambridge, MA)
started the large-scale production of MWNTs. Other companies such as Applied
Sciences, Inc and Showa Denko have already a large-scale production capacity for
MWNTs, which exhibits relatively wide distribution of diameters ranging from 50
to 120 nm (avg. = 80 nm). The most interesting point is that all companies selected
a catalytic CVD method, and furthermore three companies except Hyperion adopted
the floating reactant technique for the large-scale production of MWNTs. The devel-
opment of the floating reactant techniques allows a three-dimensional dispersion of
the hydrocarbon together with the catalytic particles derived from the pyrolysis of
organometallic compounds (Endo, 1988; Tibbetts et al., 1994; Kim et al., 2005).
Therefore, it is clear that a great deal of progress has been achieved related to the
bulk synthesis of carbon nanotubes. It is noteworthy that metallic compound in as-
grown nanotubes, which is essential for nanotube growth but invoke some toxic
properties, has to be removed as low as possible. In this sense, thermal treatment
has been utilized as the powerful tool for reducing metal content below 100 ppm
(Kim et al., 2005).
2.3 Selective Fabrication of Double Walled Carbon Nanotubes
A recently highlighted topic in carbon nanotube science and technology has been the
synthesis and characterization of DWNTs. DWNTs with a co-axial structure, that
occupies a position between MWNTs and SWNTs, have attracted much attention
because they are expected to exhibit interesting structural and electronic proper-
ties. Recently we reported the fabrication of highly purified DWNTs (Endo et al.,
2005a) through the right combination of a catalytic CVD method with an opti-
mized two-step purification process. Interestingly, they have relatively homoge-
neous and small-sized inner (ca. 0.9 nm) and outer diameters (ca. 1.5 nm), they are
thus expected to exhibit the 1D character of a quantum wire. By very carefully
pouring a stable suspension of nanotubes into a PTFE filter-attached funnel, very
thin (ca. 30 μm), round (diameter = 3.4 cm), light (weight = ca. 15 mg) and black
bucky paper was produced (Fig. 2.1a). The resultant bucky paper is the result of
the physical entanglement of DWNT bundles (10–30 nm) (Fig. 2.1b). From the
cross-sectional HRTEM image (Fig. 2.1d), we observed that our tubes consist of
two relatively round, small and homogeneous-sized (below 2 nm in the outer shell)
two concentric individual tubules. Furthermore, these co-axial tubes are packed in
a hexagonal array. This DWNT buckypaper is highly flexible and is mechanically
strong enough to fold an origami. The high structural integrity in DWNT-derived
bucky papers is thought to be derived mainly from their long lengths (up to μm
scale), because the longer the tubes, the greater the mechanical robustness of the
intermingled bundle.
2 Catalytically-Grown Carbon Nanotubes and Their Current Applications 11
Fig. 2.1 (a) Photographs of round and thin DWNT-derived bucky paper (inset, thickness is
ca. 30 μm). Low resolution transmission electron microscope (b) and field emission electron micro-
scope (c) images indicate a bundled structure. (d) Cross-sectional high resolution transmission
electron microscope image of a bundle of DWNT (inset is its schematic model, two concentric
shells were regularly packed in a hexagonal array)
2.4 Novel Microcatheter from Carbon Nanotube-Filled Nylon
Among a wide range of potential applications from nano-composite, energy storage

and energy conversion system, sensor, field emission displays and radiation source,
nano-device, to probes, carbon nanotubes hold promise for medical application
field. In this sense, the biocompatibility of carbon nanotubes has been actively
investigated over the past few years. Recently, we fabricated carbon nanotube-filled
nanocomposite-derived microcatheter via a conventional extruding system. Highly
small-sized and black novel microcatheter exhibited highly reduced thrombus
formation in vivo and low blood coagulation in vitro (Fig. 2.2) (Endo et al., 2001,
2005a,b). In addition, handleability of this novel microcatheter was enhanced
due to increased elastic modulus (ca. 1.5 times) intrinsically caused by carbon
nanotube-reinforcing effect. Novel microcatheter from nanotube-filled nanocom-
posite derived microcatheter exhibited good biological properties (e.g., reduced
thrombus formation and low inflammatory reaction) that make them applica-
ble in various fields from hemodialysis, cardiopulmonary bypass, intra-vascular
catheters, left ventricular support to vascular dilating devices. Furthermore, our
stable (almost inert) biological results based on the systematic T-cells study of
high purity carbon nanotube clarified the potential toxic problem, and resultantly,
will contribute to the active research on application of carbon nanotubes including
medical fields.
12 M. Endo
Fig. 2.2 (a) Photographs of transparent nylon-based and black nanotube-filled nanocomposite-
derived microcatheter, blood coagulation of pure (b) and (c) nanotube-filled micro catheters, throm-
bus formation of (d) pure and (e) nanotube-filled micro catheters
2.5 Conclusion
This account has mainly described the current status and prospect regarding the
large-scale synthesis of carbon nanotubes, with emphasis on their applications. The
unique electronic properties of nanotubes make them also good candidates in the
electronic-industry so that silicon-based technologies could be replaced by nano-
carbons. At present, researchers are working on the following areas in order to
apply carbon nanotubes in emerging technologies: (i) production of defect-free and
high purity carbon nanotubes; (ii) establishment of useful techniques for quantify-
ing the number of defects (types, where and amount etc.) in the nanotube structure;
(iii) development of effective purification techniques (below ppm level) for the metal
particles within carbon nanotube, especially for biological and electronic applica-
tions, and (iv) achieving homogeneous carbon nanotubes dispersions in polymer
composites. The possibilities of applications using carbon nanotubes range from
electronics, field emission display to energy storage devices and functional fillers
in composites. Nowadays the carbon nanotubes-derived products have smeared into
our life step by step, and before long, they will function as essential components for
technological innovations.
2 Catalytically-Grown Carbon Nanotubes and Their Current Applications 13
References
Dresselhaus, M.S., Dresselhaus, G., Eklund, P.C., 1995, Science of Fullerene and Carbon Nan-
otube (Academic Press: Tokyo).
Endo, M., 1988, Grow carbon fibers in the vapour phase. Chem. Tech. 18, pp. 568–576.
Endo, M., Saito, R., Dresselhaus, M.S., Dresselhaus, G., 1997, From carbon fibers to carbon nan-
otube. In Carbon Nanotubes: Preparation and Properties, edited by Ebbesen, T.W. (New York:
CRC Press), pp. 35–110.
Endo, M., Kim, Y.A., Hayashi, T., Nishimura, K., Matushita, T., Miyashita, K., Dresselhaus, M.S.,
2001, Vapor-grown carbon fibers (VGCFs): Basic properties and their battery applications.
Carbon, 39, pp. 1287–1297.
Endo, M., Muramatsu, H., Hayashi, T., Kim, Y.A., Terrones, M., Dresselhaus, M.S., 2005a, ‘Buck-
ypaper’ from coaxial nanotubes. Nature, 433, p. 476.
Endo, M., Koyama, S., Matsuda, Y., Hayashi, T., Kim, Y.A., 2005b, Thrombogenicity and
blood coagulation of a microcatheter prepared from carbon nanotube-nylon-based composite.
Nanoletters, 5, pp. 101–106.
Iijima, S., 1991, Helical microtubules of graphitic carbon. Nature, 354, pp. 56–58.
Kim, Y.A., Hayashi, T., Endo, M., Kaburagi, Y., Tsukada, T., Shan, J., Osato, K., Tsuruoka, S.,
2005, Synthesis and structural characterization of thin multi-walled carbon nanotubes with a
partially facetted cross section by a floating reactant method. Carbon, 43, pp. 2243–2250.
Oberlin, A., Endo, M., Koyama, T., 1976, Filamentous growth of carbon through benzene decom-
position. Journal of Crystal Growth, 32, pp. 335–349.
Saito, R., Dresselhaus, G., Dresselhaus, M.S., 1998, Physical Properties of Carbon Nanotubes
(Singapore: Imperial College Press).
Sun, X., Kiang, C.H., Endo, M., Takeuchi, K., Furuta, T., Dresselhaus, M.S., 1996, Stacking char-
acteristics of graphene shells in carbon nanotubes. Physical Review B, 54, pp. R12629–R12632.
Tibbetts, G.G., Bernnardo, C.A., Gorkiewicz, D.W., Alig, R.L., 1994, Role of sulfur in the produc-
tion of carbon fibers in the vapor phase. Carbon, 32, pp. 569–576.
Chapter 3
Transparent Conducting Films by Using
Carbon Nanotubes
Hong-Zhang Geng and Young Hee Lee∗
3.1 Introduction
The interests on the flexible transparent conducting films (TCFs) have been growing
recently mainly because of strong needs for electrodes of flexible or wearable dis-
plays in the future (Cairns and Crawford, 2005). The current technology adopts
indium-tin oxide (ITO) for TCFs that meets the requirement of low resistance and
high transmittance. Yet, bending of the ITO film generates cracks in the film, result-
ing in poor flexibility reported by Saran et al. (2004). This is a serious drawback
in flexible display applications. Carbon nanotube (CNT), which was discovered by
Iijima (1991), is a new functional material that can be treated as graphitic sheets
with a hexagonal lattice being wrapped up into a cylinder (Baughman et al., 2002).
It has a high aspect ratio of typically 1,000 or greater with a diameter of a few
nanometers. It is generally known that CNTs have high elastic modulus of 1–2 TPa
as well as high electrical conductivity which is 1,000 times larger than Cu-wire
(Bernholc et al., 1998; Salvetat et al., 1999; Mamalis et al., 2004). Wu et al. (2004)
found that the CNTs were capable of forming naturally robust random network in
the film and provide low sheet resistance and high transmittance with a minimal
amount of CNTs. CNT films have been known to exhibit an excellent bending
characteristics over the conventional ITO film. These structural and physical prop-
erties of CNTs are superb features for flexible TCFs, providing high performance
of transparence and conductivity. Although CNTs have been suggested for flexi-
ble TCFs with outstanding film performance, the performance strongly relies on
the film preparation conditions (Kaempgen et al., 2003; Sreekumar et al., 2003;
Kim et al., 2003; Saran et al., 2004; Wu et al., 2004; Ferrer-Anglada et al., 2004;
Hu et al., 2004; Zhang et al., 2005; Kaempgen et al., 2005; Moon et al., 2005;
Unalan et al., 2006). The understanding for the film fabrication methods, material
dependence, and criteria to determine the sheet resistance and transmittance is still
in its infant stage.
∗ Department of Nanoscience and Nanotechnology, Department of Physics, and Center for Nan-
otubes and Nanostructured Composites, Sungkyunkwan Advanced Institute of Nanotechnology,

Sungkyunkwan University, Suwon 440–746, R.O. Korea
e-mail: [email protected]

C Springer 2008
16 H.-Z. Geng, Y. H. Lee
The fabrication of TCFs includes three steps: (i) the CNT treatment, (ii) the CNT-
solution preparation, and (iii) the film preparation. The choice of CNTs, the degree
of purity and defects of CNTs, and further post-treatment of CNTs are presumably
important factors in determining the film performance. The CNTs are easily bundled
with a diameter of typically a few tens nanometers due to their strong van der Waals
interactions (∼ 1000 eV) originating from the micrometer long CNTs (Thess et al.,
1996; Pan et al., 1998; Sun et al., 2005). Thus, the macrodispersion (small-size bun-
dles) or the nanodispersion (dispersion into individual nanotubes) is strongly desired
in order to minimize the absorbance (O’Connell et al., 2001; Richard et al., 2003;
Lee et al., 2005; Kim et al., 2005). The choice of solvents and the related dispersants
are crucial factors in achieving the best dispersion conditions (O’Connell et al, 2002;
Vaisman et al, 2006). Once the CNT solution is ready, the film preparation is another
important step to consider, since this determines the networking of CNTs.
The purpose of this work is to determine which type of nanotubes is the best
for flexible TCFs. Several single-walled CNTs (SWCNTs) prepared by different
methods such as arc discharge (Arc), catalytic chemical vapor deposition (CVD),
high pressure carbon monoxide (HiPCO), and laser ablation (Laser) were intro-
duced to search for the material parameters. These CNTs were dispersed in water
with sodium dodecyl sulfate (SDS) with sonication. This CNT solution was then
sprayed onto the Poly(ethylene terephthalate) (PET) substrate by a spray coater
to form the CNT film. The quality of CNTs, the degree of dispersion, film mor-
phology and the performance of TCFs were characterized by scanning electron
microscopy, transmission electron microscopy, thermogravimetric analysis, Raman
spectroscopy, optical spectra, and four-point probe measurements. We found that
the material parameters such as the diameter and crystallinity of CNTs, bundle size,
networking of CNTs of the film, the degree of dispersion of CNTs in solvent are
crucial parameters to improve the film performance. We proposed a material quality
factor to evaluate the quality of SWCNTs and further discussed the film performance
that relied on the absorbance and on the degree of entanglement in the CNT network.
Our analysis demonstrates that the Arc SWCNTs are the best candidate to give rise
to the lowest sheet resistance and highest transmittance among other species. The
CNT film performance shows a good candidate for replacing the conventional ITO
transparent conducting film in various application fields.
3.2 Experimental
Different kinds of SWCNT materials were used in this study. The CVD SWCNTs
were purchased from SouthWest NanoTechnologies, Inc., HiPCO SWCNTs and
Laser SWCNTs were from Carbon Nanotechnology, Inc. Arc SWCNTs were from
Iljin Nanotech Co. Ltd. The CNTs were characterized by the field-emission scan-
ning electron microscopy (FE-SEM-JSM7000F), transmission electron microscopy
(TEM-JEM2100F), Raman spectroscopy (Renishaw RM1000-Invia), and thermo-
gravimetric analysis (TGA-Seiko Exstar 6000 (TMA6100), SEICO INST. JAPAN).
3 Transparent Conducting Films by Using Carbon Nanotubes 17
Deionized water was used to disperse SWCNTs with sonication and centrifuge
process. SWCNTs with a concentration of 0.3 mg/ml and SDS (Sigma–Aldrich) of
3 mg/ml were dissolved in water and sonicated in a bath type sonicator (Power sonic
505) at 400 W for 10 hours. The SWCNT solution was centrifuged at 10,000 g for
10 minutes by centrifuger (Hanil Science Industrial, MEGA 17R). The upper 50%
of the supernatant solution was carefully decanted for characterization. Absorbance
of the SWCNT supernatant was recorded by UV-VIS-NIR spectrophotometer
(Cary-5000). The SWCNT supernatant was directly sprayed with air brush pistol
(GUNPIECE GP-1, FUSO SEIKI Co. LTD) onto the PET substrate. A special care
for the choice of CNT concentration was taken in order to keep the nozzle from
being clogged.
During the spray process, the PET substrate was kept on the sample holder
at 100 ◦C to accelerate evaporation of the fine droplets on the surface. When the
spray process was terminated, the TCF film was immersed into deionized water for
10 minutes to remove the surfactant and dried in a dry oven at 80 ◦ C for 30 minutes.
This process was repeated twice to further remove surfactants and enhance adhe-
sion between CNTs and PET film. The transmittance was recorded by UV-VIS-NIR
spectrophotometer in the visible range (400–800 nm). Measurements of the sheet
resistance were carried out by four-point probe method (Keithley 2000 multimeter)
at room temperature.
3.3 Results and Discussion
3.3.1 Analysis on SWCNT Materials
The general properties of SWCNT powders were characterized by different meth-

ods. Figure 3.1 presents the FE-SEM and TEM morphologies of different SWCNT
powders used in our study. The average diameters of individual CVD, HiPCO,
Laser, and Arc SWCNTs were 0.8, 1.0, 1.2, and 1.4 nm, respectively, as deter-
mined from the TEM images. The CVD SWCNTs had the smallest average bundle
size, as estimated from the SEM images, whereas the Laser sample had the largest
average bundle size among all the samples. Some carbonaceous particles on the
CNT bundles were present in the CVD SWCNTs. The Arc SWCNTs had relatively
well-defined crystallinity without amorphous carbons on the tube walls, although
the bundle size of the Laser sample was larger than that of the Arc sample from the
FE-SEM images. The presence of carbonaceous particles on the nanotube walls is an
important factor for the application of TCFs, since they determine the contact resis-
tance among the CNT network (Dai et al., 1996; Jost et al., 1999; Hu et al., 2006).
The purity of each sample can be obtained from the TGA. Figure 3.2 presents the
TGA and differential TGA of the SWCNTs that were used in this work. The SWC-
NTs started to burn off in air with increasing temperature. The remaining material at
900 ◦C was metal oxides. All the samples contained some amounts of catalysts. The
burning temperature obtained from the peak of differential TGA varied with several
Fig. 3.1 FE-SEM (scale bar: 200 nm) and TEM (inset, scale bar: 20 nm) images of CNTs: (a) CVD
SWCNTs, (b) HiPCO SWCNTs, (c) Laser SWCNTs, and (d) Arc SWCNTs
Fig. 3.2 TGA and differential TGA of the CNTs: (a) CVD SWCNTs, (b) HiPCO SWCNTs,
(c) Laser SWCNTs, and (d) Arc SWCNTs
parameters, for instance, the content of metals, the bundle diameter of CNTs, the
number of walls, and the crystallinity of CNTs (Arepalli et al., 2004). It is interesting
to note the strong correlation between the sample purity (or catalyst amount) and the
burning temperature for all the SWCNTs in Fig. 3.2a–d. The unusual endothermic
burning behavior was observed at the sample typically with high metal content,
which can be seen in the CVD SWCNTs in Fig. 3.2a. The Laser SWCNTs in
Fig. 3.2c revealed two peaks in the differential TGA. The first peak could be iden-
tified as the burning temperature of amorphous carbons (Shi et al., 2000; Sadana
et al., 2005). The second peak could be regarded as the SWCNT peak but the burn-
ing temperature was low mainly due to the presence of large amount of catalysts
(∼ 10 wt%). The burning temperature can be also related to the number of defects
on the CNT walls.
The SWCNT powders were also characterized by Raman spectroscopy.
Figure 3.3 shows the Raman shift of the samples at excitation energies of 514 nm
(2.41 eV) and 633 nm (1.96 eV). The metallicity of each sample can be determined
from the radial breathing modes (RBMs). For a given laser excitation energy, the
SWCNTs whose allowed transition energies between van Hove singularities are
in agreement with the excitation energy, will be excited to produce the Raman
signal. This is known as the resonant Raman spectroscopy. Since the positions of
van Hove singularities are dependent specifically on the diameter and chirality,
Fig. 3.3 Raman spectra of

the SWCNT materials at
excitation energy of
(a) 514 nm and (b) 633 nm.
Each region of van Hove
singularities is provided in
the box
the metallicity of the excited SWCNTs can be determined thoroughly (Rao et al.,
1997; Jorio et al., 2001; Kuzmany et al., 2001). At 514 nm, the Laser and the Arc
samples revealed only the semiconducting nanotubes. On the other hand, the CVD
and the HiPCO samples contained both metallic and semiconducting nanotubes.
The metallic component of SWCNTs was further evidenced by the presence of
long tail at lower energy side in the G-band, i.e., the Fano line representing the
metallic contribution (Brown et al., 2001a). At 633 nm, the Laser and the Arc
samples picked up mostly metallic SWCNTs, whereas the CVD sample retained
mostly semiconducting SWCNTs and the HiPCO sample contained both metallic
and semiconducting SWCNTs. In addition to RBMs and G-band, the D-band was
also present in all samples. D-band near 1320 cm−1 originates in general from the
presence of defects, although its exact origin was not certain (Brown et al., 2001b).
The intensity of D-band indicates the amount of defects on the nanotube walls.
Another intriguing parameter is the intensity of the G -band. The G -band is the
second-order Raman signal, which is the first overtone of the D-band. Geng et al.
(2007) found that G -band is governed by the metallicity of the CNTs, i.e., the
intensity of G -band is directly proportional to the metallic composition of nan-
otubes. At 514 nm, the Arc and Laser samples revealed relatively small intensities,
because they contained mostly semiconducting nanotubes, as shown in Fig. 3.3a.
On the other hand, the CVD and HiPCO samples contained a certain portion of
metallic composition, therefore yielding the large intensities of the G -band. At
633 nm, similar trend of the changes in the intensity of G -band was observed. The
intensity of G -band increased with increasing metallic composition in each sample.
3.3.2 Material Quality Factor for SWCNTs
We obtained several parameters of the SWCNTs used in our study. Figure 3.4
presents the summary of the material parameters of the SWCNT samples. The aver-
age diameters of each sample were determined from the TEM. Arc SWCNTs had
the largest average diameter among other types, as shown in Fig. 3.4a. Figure 3.4b
shows the burning temperature changing with the purity of SWCNTs. More impu-
rities such as transition metals and carbonaceous particles indicate degradation of
conductivity. Figure 3.4c presents the intensity ratio of D-band to G-band and G -
band to G-band at excitation energies of 514 and 633 nm, respectively. The presence
of defects is represented in terms of the relative intensity ratio of D-band to G-band.
This leads to lower the electrical conductivity of metallic CNTs. The abundance of
metallic components is represented by the intensity ratio of G -band to G-band.
More metallic nanotubes results in higher conductivity of the materials.
The band gap of semiconducting nanotubes is inversely proportional to the diam-
eter, E g = 2 ac-c γo /D (eV), where ac-c is 0.142 nm and γo is an empirical tight-
binding parameter taken as 2.9 eV (Dresselhaus et al., 2005). The conductivity of
SWCNTs can generally be simply expressed as
σ = neμn + peμ p (3.1)

Fig. 3.4 Summary of the

material parameters for
SWCNTs: (a) diameters,
(b) burning temperature and
purity, (c) The ratio of the
intensity of D-band to G-band
and G -band to G-band of the
SWCNTs at excitation
energies of 514 and 633 nm
where n and p are the n-type (electrons) and p-type (holes) carrier concentrations,
respectively, and μn and μ p are the respective electron and hole mobility. The
mobility is dominated by a succession of random scattering from collisions with
lattice atoms, impurity atoms, and other scattering centers. The intrinsic carrier
concentration decreases exponentially with band gap, n i = n o ex p(−E g /2k B T ),
where k B and T are the Boltzmann constant and temperature of the system, respec-
tively. The p-type and n-type nanotube carrier concentrations are, p = n i ex p[(E i −
E f )/k B T ], and n = n i ex p[(E f − E i )/k B T ], where the intrinsic Fermi level E i
is frequently used as a reference level when the extrinsic semiconductors are dis-
cussed with a Fermi level of E f . The conductivity is proportional to the carrier
concentration, σ ∼ ex p(−E g /2k B T )ex p[(E i −E f )/k B T ] for semiconducting nan-
otubes. For intrinsic semiconducting nanotubes, E i = E f . On the other hand, in
metallic nanotubes, π and π∗ overlap at the Fermi level, i.e., metallic nanotubes are
always metallic independent of the diameters. Nevertheless, Delaney et al. (1998)
had suggested that the nanotubes are usually bundled and a pseudogap of ∼ 0.1 eV
is open due to the tube-tube interaction. This pseudogap is small compared to the
direct bandgap of semiconductors with diameters of 1 ∼ 1.4 nm corresponding to
bandgaps of 0.7 ∼ 1.0 eV (An and Lee, 2006). The pseudogap in metallic SWCNT
bundles is inversely proportional to the tube diameter via E pg ≈ 0.105/D (eV)
after fitting to the observed values (Ouyang et al., 2001). Thus, the conductivity of
metallic nanotubes reveals the similar diameter dependence to semiconducting ones,
σ ∼ ex p(−E pg /2k B T ).
As we described before, the intensity of the G -band is a measure of the abun-
dance of metallic nanotubes. Therefore we propose this concept in quantifying
the metallicity of nanotubes. It is also noted in our samples that the intensity of
the G -band is strongly correlated to the intensity of metallic peaks in RBMs.
Despite the abundance of metallicity, the presence of defects on the nanotube walls
that may act as scattering centers degrades the conductivity of the SWCNT net-
work. The intensity of the D-band indicates the amount of defects on the nanotube
walls. Therefore, an appropriate parameter to express conductivity of nanotubes for
SWCNT films is the intensity ratio, G -band/D-band. High abundance of metallicity
and few defects on the nanotube walls will be desired for high conductivity of the
SWCNT films.
From these analyses, it may be useful to extract a practical parameter as a figure
of merit to evaluate the metallicity of SWCNTs. The purity affects the conductivity.
The diameter contributes to the conductivity via bandgap described in the previous
paragraph. More defects reduce the mean free path of carriers and decrease the
mobility of carriers in nanotubes. The intensity ratio of G -band/D-band may repre-
sent the mobility of the carriers. The conductivity is proportional to the metallicity of
nanotubes and inversely proportional to the number of scattering centers or defects.
Here, we define an effective material quality factor Q m that governs the conductivity
of SWCNTs
E i −E f
−E g
−E pg

Q m = P ×(e 2k B T × e kB T
× I S + e 2k B T × IM ) (3.2)
where E g = 0.82/D (eV) and E pg = 0.105/D (eV) from the previous paragraph,
D (nm) is the average diameter of individual SWCNTs and P is the purity of the
sample. The intensity ratio of G /D was averaged over the excitation energies of
Raman spectroscopy for each metallic and semiconducting nanotubes. Here I S (I M )
is defined as
A S (A M )
I S (I M ) = IG /D × (3.3)
AM + AS
where A S (A M ) is the area intensity of semiconducting (metallic) peaks of RBMs

from Raman shift in Fig. 3.3. The first exponential term in Eq. (3.2) represents
the carrier concentration and the second exponential term represents the mobility
that changes with doping effect from semiconducting nanotubes. The second term
represents the contribution from metallic nanotubes. Thus this formula resembles
the conductivity of Eq. (3.1). Only two wavelengths were used in our study but it
Fig. 3.5 The material quality

factor of SWCNTs defined by
Eq. (3.2) in the text
seemed satisfactory to explain the TCF performance in terms of Q m . The abundance

of metallic nanotubes is implicitly expressed in the intensity ratio of G /D. Since
the exponent of metallic nanotubes is much greater than that of semiconducting
nanotubes with typical diameters of 1 ∼ 2.0 nm, the first term can be negligible
in Eq. (3.2) in intrinsic nanotubes when E i ≈ E f . The semiconducting nanotubes
generally are sensitive to the environment invoking a doping effect. In such cases,
the Fermi level can be shifted and the first term may not be negligible. Although the
pseudogap of metallic nanotubes can be affected by a doping effect (Kwon et al.,
1998; Zhu et al., 2005), we didn’t consider it in Eq. (3.2), since it is relatively small
in comparison to that of semiconducting nanotubes. We calculated the material qual-
ity factor Q m in case of E i ≈ E f . Figure 3.5 shows the material quality factor Q m
of the four SWCNTs in our study. From this point of view, the Arc sample was
presumably the best sample that provided the highest conductivity than other types
of SWCNTs considered in this work, as can be compared from Fig. 3.5.
3.3.3 Film Characteristics of SWCNTs
Once the prepared SWCNT powder was dispersed in deionized water with SDS, as
described in experiment, this solution was sprayed onto the PET film to form thin
SWCNT TCFs. Figure 3.6 presents the characteristic curves for sheet resistance-
transmittance. For a direct comparison, at 80% of transmittance, the Arc TCFs
showed the lowest sheet resistance of about 160 /sq, whereas the CVD TCFs
showed the highest sheet resistance. From the material point of view, as discussed
in the previous paragraph, this general trend of the film performance seemed to be
quite understandable.
We calculated the sheet conductance of TCFs at transmittance of 70 and 80%
from Fig. 3.6 and plot as a function of material quality factor. Figure 3.7 shows the
sheet conductance of TCFs at a transmittance of 70 and 80% as a function of the
material quality factor Q m . We clearly observed a direct correlation between the
sheet conductance and the material quality factor. The sheet conductance reveals a
linear relationship with the material quality factor. Although this empirical formula
Fig. 3.6 Characteristic

curves of sheet
resistance-transmittance for
various SWCNT TCFs. Each
curve contains several data
points from TCFs with
different number of spray
times. Transmittance at
550 nm was selected for
comparison. SWCNTs were
dispersed in deionized water
with SDS
is not rigorous, it can provide at least a means of estimating material quality that
governs the conductivity of the SWCNT TCFs. For instance, large diameter, higher
purity, less defects (lower intensity of D-band), and more metallic nanotubes (higher
intensity of G -band) will give better conductivity of the SWCNT TCF. From this
point of view, the Arc TCF is the best sample providing the highest conductivity in
comparison to TCFs made by other types of SWCNTs considered in this work, as
can be seen from Fig. 3.6. However, it may be conjectured that different optimiza-
tion conditions for dispersion and film preparation may change the TCF properties.
In spite of such a possibility, the argument for the material quality dependence
described above still holds true. The trend of the change in sheet conductance is
similar at different transmittance regions, although the slopes are different. This
suggests that our definition of the effective material parameter is quite understand-
able from a material point of view to describe the TCF performance.
The film morphologies of the SWCNT network of TCFs are presented in Fig. 3.8.
All other processing parameters except types of SWCNTs were kept constant
through the TCFs preparation process. It is noted that the film morphologies of
the CVD and the HiPCO TCFs were significantly different from those of the Laser
and the Arc samples. The circular spots were visible in the CVD and the HiPCO
samples, as can be seen in Fig. 3.8a–b. These are the remaining SDS surfactants
in the film in spite of the repeated washing process. This resulted presumably from
Fig. 3.7 The sheet

conductance changes as a
function of the material
quality factor at different
transmittance of 70and 80%
Fig. 3.8 FE-SEM images of the as-prepared TCFs using different types of SWCNTs (scale bar:
400 nm): (a) CVD SWCNTs, (b) HiPCO SWCNTs, (c) Laser SWCNTs, and (d) Arc SWCNTs
the stronger binding of surfactants with defects on the SWCNT walls that were
hardly removable in the current process compared to the Laser and the Arc samples.
Another important feature is the sparseness and networking capability of the CNT
bundles. The Laser and the Arc films revealed low CNT density but better network
formation by an entanglement of the SWCNT bundles. On the other hand, the CNT
density in the CVD and the HiPCO films was high but with relatively less degree
of entanglement. As a consequence, the transmittance in the CVD and the HiPCO
films was significantly dropped at high density of the CVD and HiPCO films. From
the SWCNT density and networking capability of the CNT films, the Arc sample
again seemed to be the best material for TCFs among these samples we studied in
this work.
3.4 Conclusion
In summary, the flexible TCFs with various types of SWCNTs have been inves-
tigated by using spray method. The characteristics of TCFs were analyzed by both
the material quality factors and the film quality factors. We extracted all the material
parameters of SWCNT powders such as diameter, purity and burning temperature,
D-band and G -band in Raman spectra. We found that G -band represented well the
metallicity of SWCNTs. We defined the material quality factor that dominated the
conductivity of SWCNTs by using the diameter, purity, and the average of inten-
sity ratio of G -band to D-band. The sparseness of the SWCNT network and the
degree of entanglement in the film were also important film parameters to determine
the conductivity of the film. By taking account of all these associated effects, we
were able to provide an explanation of the SWCNT film performance. A further
improvement in the film performance could be possible by choosing highly con-
ductive CNTs evaluated by material quality factor and by enhancing the degree
of dispersion of CNTs in various solvents. In addition, with enhancing the film

properties, it will make the CNT-based TCFs possible for flexible display appli-
cations.
Acknowledgments We acknowledge the financial support by the kosef through cnnc at skku and
in part financially supported by the moe, mocie and molab through the fostering project of the
laboratory of excellence.
References
An, K. H. and Lee, Y. H., 2006, Electronic-structure engineering of carbon nanotubes. Nano, 1,
pp. 115–138.
Arepalli, S., Nikolaev, P., Gorelik, O., Hadjiev, V. G., Holmes, W., Files, B. and Yowell, L., 2004,
Protocol for the characterization of single-wall carbon nanotube material quality. Carbon, 42,
pp. 1783–1791.
Baughman, R. H., Zakhidov, A. A. and de Heer, W. A., 2002, Carbon Nanotubes—the Route
Toward applications. Science, 297, pp. 787–792.
Bernholc, J., Brabec, C., Nardelli, M. B., Maiti, A., Roland, C. and Yakobson, B. I., 1998, Theory
of growth and mechanical properties of nanotubes. Applied Physics A, 67, pp. 39–46.
Brown, S. D. M., Jorio, A., Corio, P., Dresselhaus, M. S., Dresselhaus, G., Saito, R. and Kneipp, K.,
2001a, Origin of the Breit-Wigner-Fano lineshape of the tangential G-band feature of metallic
carbon nanotubes. Physical Review B, 63, pp. 155414.
Brown, S. D. M., Jorio, A., Dresselhaus, M. S. and Dresselhaus, G., 2001b, Observation of
the D-band feature in the Raman spectra of carbon nanotubes. Physical Review B, 64,
pp. 073403.
Cairns, D. R. and Crawford, G. P., 2005, Electromechanical properties of transparent conducting
substrates for flexible electronic displays. Proceedings of the IEEE, 93, pp. 1451–1458.
Dai, H., Wong, E. W. and Lieber, C. M., 1996, Probing electrical transport in nanomaterials: con-
ductivity of individual carbon nanotubes. Science, 272, pp. 523–526.
Delaney, P., Choi, H. J., Ihm, J., Louie, S. G. and Cohen, M. L., 1998, Broken symmetry and
pseudogap in ropes of carbon nanotubes. Nature, 391, pp. 466–468.
Dresselhaus, M. S., Dresslhaus, G., Saito, R. and Jorio, A., 2005, Raman spectroscopy of carbon
nanotubes. Physics Reports, 409, pp. 47–79.
Ferrer-Anglada, N., Kaempgen, M., Skákalová, V., Dettlaf-Weglikowska, U. and Roth, S., 2004,
Synthesis and characterization of carbon nanotube-conducting polymer thin films. Diamond
and Related Materials, 13, pp. 256–260.
Geng, H.-Z., Kim, K. K., So, K. P., Lee, Y. S., Chang Y. and Lee, Y. H., 2007, Effect of acid treat-
ment on carbon nanotube-based flexible transparent conducting films. Journal of the American
Chemical Society, 129, pp. 7758–7759.
Hu, L., Hecht, D. S. and Grüner, G., 2004, Percolation in Transparent and conducting carbon
nanotube Networks. Nano Letters, 4, pp. 2513–2517.
Hu, Y. F., Liang, X. L., Chen, Q., Peng, L.-M., and Hu, Z. D., 2006, Electrical characteristics of
amorphous carbon nanotube and effects of contacts. Applied Physics Letters, 88, pp. 063113.
Jorio, A., Saito, R., Hafner, J. H., Lieber, C. M., Hunter, M., McClure, T., Dresselhaus, G. and
Dresselhaus1, M. S., 2001, Structural (n, m) determination of isolated single-wall carbon nan-
otubes by resonant Raman scattering. Physical Review Letters, 86, pp. 1118–1121.
Jost, O., Gorbunov, A. A., Pompe, W., Pichler, T., Friedlein, R., Knupfer, M., Reibold, M.,
Bauer, H.-D., Dunsch, L., Goldem, M. S. and Fink, J., 1999, Diameter grouping in bulk sam-
ples of single-walled carbon nanotubes from optical absorption spectroscopy. Applied Physics
Letters, 75, pp. 2217–2219.
Kaempgen, M., Dettlaff, U. and Roth, S., 2003, Characterization of carbon nanotubes by optical
spectra. Synthetic Metals, 135–136, pp. 755–756.
Kaempgen, M., Duesberg, G. S. and Roth, S., 2005, Transparent carbon nanotube coatings. Applied
Surface Science, 252, pp. 425–429.
Kim, K. K., Bae, D. J., Yang, C.-M., An, K. H., Lee, J. Y. and Lee, Y. H., 2005, Nanodispersion of
single-walled carbon nanotubes using dichloroethane. Journal of Nanoscience and Nanotech-
nology, 5, pp. 1055–1059.
Kim, Y., Minami, N., Zhu, W., Kazaoui, S., Azumi, R. and Matsumoto, M., 2003, Langmuir-
Blodgett films of single-wall carbon nanotubes: layer-by-layer deposition and in-plane orienta-
tion of tubes. Japanese Journal of Applied Physics, 42, pp. 7629–7634.
Kuzmany, H., Plank, W., Hulman, M., Kramberger, Ch., Grüneis, A., Pichler, Th., Peterlik, H.,
Kataura, H. and Achiba, Y., 2001, Determination of SWCNT diameters from the Raman
response of the radial breathing mode. The European Physical Journal B, 22, pp. 307–320.
Kwon, Y.-K., Satio, S. and Tománek, D., 1998, Effect of intertube coupling on the electronic struc-
ture of carbon nanotube ropes. Physical Review B, 58, pp. R13314–R13317.
Lee, J. Y., Kim, J. S., An, K. H., Lee, K., Kim, D. Y., Bae, D. J. and Lee, Y. H., 2005,
Electrophoretic and dynamic light scattering in evaluating dispersion and size distribu-
tion of single-walled carbon nanotubes. Journal of Nanoscience and Nanotechnology, 5,
pp. 1045–1049.
Mamalis, A. G., Vogtländer, L. O. G. and Markopoulos, A., 2004, Nanotechnology and nanostruc-
tured materials: trends in carbon nanotubes. Precision Engineering, 28, pp. 16–30.
Moon, J. S., Park, J. H., Lee, T. Y., Kim, Y. W., Yoo, J. B., Park, C. Y., Kim, J. M. and Jin,
K. W., 2005, Transparent conductive film based on carbon nanotubes and PEDOT composites.
Diamond and Related Materials, 14, pp. 1882–1887.
O’Connell, M. J., Bachilo, S. M., Huffman, C. B., Moore, V. C., Strano, M. S., Haroz, E. H.,
Rialon, K. L., Boul, P. J., Noon, W. H., Kittrell, C., Ma, J., Hauge, R. H., Weisman, R. B. and
Smalley, R. E., 2002, Band Gap Fluorescence from individual single-walled carbon nanotubes.
Science, 297, pp. 593–596.
O’Connell, M. J., Boul, P., Ericson, L. M., Huffman, C., Wang, Y., Haroz, E., Kuper, C., Tour, J.,
Ausman, K. D. and Smalley, R. E., 2001, Reversible water-solubilization of single-walled car-
bon nanotubes by polymer wrapping. Chemical Physics Letters, 342, pp. 265–271.
Ouyang, M., Huang, J.-L., Cheung, C. L. and Lieber, C. M., 2001, Energy gaps in metallic single-
walled carbon nanotubes. Science, 292, pp. 702–707.
Pan, Z. W., Xie, S. S., Chang, B. H., Wang, C. Y., Lu, L., Liu, W., Zhou, W. Y., Li, W. Z. and
Qian, L. X., 1998, Very long carbon nanotubes. Nature, 394, pp. 631–632.
Rao, A. M., Richter, E., Bandow, S., Chase, B., Eklund, P. C., Williams, K. A., Fang, S.,
Subbaswamy, K. R., Menon, M., These, A., Smalley, R. E., Dresselhaus, G. and Dressel-
haus M. S., 1997, Diameter-selective Raman scattering from vibrational modes in carbon nan-
otubes. Science, 275, pp. 187–191.
Richard, C., Balavoine, F., Schultz, P., Ebbesen, T. W. and Mioskowski, C., 2003, Supramolecular
self-assembly of lipid derivatives on carbon nanotubes. Science, 300, pp. 775–778.
Sadana, A. K., Liang, F., Brinson, B., Arepalli, S., Farhat, S., Hange, R. H., Smalley, R. E. and
Billups, W. E., 2005, Functionalization and extraction of large fullerenes and carbon-coated
metal formed during the synthesis of single wall carbon nanotubes by laser oven, direct current
arc, and high-pressure carbon monoxide production methods. The Journal of Physical Chem-
istry B, 109, pp. 4416–4418.
Salvetat, J.-P., Briggs, G. A. D., Bonard, J. M., Basca, R. R., Kulik, A. J., Stöckli, T., Burn-
ham, N. A. and Forró, L., 1999, Elastic and shear moduli of single-walled carbon nanotube
ropes. Physical Review Letters, 82, pp. 944–947.
Saran, N., Parikh, K., Suh, D.-S., Muñoz, E., Kolla, H. and Manohar, S. K., 2004, Fabrication
and characterization of their films of single-walled carbon nanotube bundles on flexible plastic
substrates. Journal of the American Chemical Society, 126, pp. 4462–4463.
Shi, Z., Lian, Y., Liao, F. H., Zhou, X., Gu, Z., Zhang, Y., Iijima, S., Li, H., Yue, K. T. and
Zhang, S.-L., 2000, Large scale synthesis of single-wall carbon nanotubes by arc-discharge
method. Journal of Physics and Chemistry of Solids, 61, pp. 1031–1036.
Sreekumar, T.V., Liu, T. and Kumar, S., 2003, Single-wall Carbon nanotube films. Chemistry of
Materials, 15, pp. 175–178.
Sun, C.-H., Yin, L.-C., Li, F., Lu, G.-Q. and Cheng, H.-M., 2005, Van der Waals interactions
between two parallel infinitely long single-walled nanotubes. Chemical Physics Letters, 403,
pp. 343–346.
Thess, A., Lee, R., Nikolaev, P., Dai, H., Petit, P., Robert, J., Xu, C., Lee, Y. H., Kim, S. G.,
Rinzler, A. G., Colbert, D. T., Scuseria, G. E., Tománek, D., Fischer, J. E. and Smalley, R. E.,
1996, Crystalline ropes of metallic carbon nanotubes. Science, 273, pp. 483–487.
Unalan, H. E., Fanchini, G., Kanwal, A., Pasquier, A. D. and Chhowalla, M., 2006, Design criteria
for transparent single-wall carbon nanotube thin-film transistors. Nano Letters, 6, pp. 677–682.
Vaisman, L., Marom, G. and Wagner, H. D., 2006, Dispersions of surface-modified carbon nan-
otubes in water-soluble and water-insoluble polymers. Advanced Functional Materials, 16,
pp. 357–363.
Wu, Z. C., Chen, Z., Du, X., Logan, J. M., Sippel, J., Nikolou, M., Kamaras, K., Reynolds, J. R.,
Tanner, D. B., Hebard, A. F. and Rinzler, A. G., 2004, Transparent, conductive carbon nanotube
films. Science, 305, pp. 1273–1276.
Zhang, M., Fang, S., Zakhidov, A. A., Lee, S. B., Aliev, A. E., Williams, C. D., Atkinson, K. R. and
Baughman, R. H., 2005, Strong, transparent, multifunctional, carbon nanotube sheets. Science,
309, pp. 1215–1219.
Zhu, H., Zhao, G.-L., Masarapu, C., Young, D. P. and Wei, B., 2005, Super-small energy gaps of
single-walled carbon nanotube strands. Applied Physics Letters, 86, pp. 203107.
Chapter 4
Raman Spectroscopy on Double-Walled
Carbon Nanotubes
Wencai Ren and Hui-Ming Cheng∗
4.1 Introduction
Double-walled carbon nanotube (DWNT) is a special MWNT only consisting of two

coaxial SWNTs. It is an ideal model to study the effect of interlayer interaction on
the phonons and electronic structures of carbon nanotubes (CNTs), and have some
unique properties and potential applications as nanodevices, due to its unique double
wall structure. For example, it is expected to have some potential applications as
molecular conductive wire or molecular capacitor in a memory device, depending
on the electronic properties of the two constituent tubes (Saito et al., 1993). Recent
studies also indicated that DWNTs are better field effect transistor (FET) channels
than SWNTs (Shimada et al., 2004). Therefore, it is very important to realize the
controllable synthesis of DWNTs and to identify the diameters and the electronic
properties of the two constituent tubes for their promising applications as nano-scale
electronic devices.
It is well established that every possible nanotube has a distinct electronic and
vibrational spectrum, so that there is a one-to-one relation between the nanotube
and the singularities in the 1D joint electronic density of states (JDOS) (Dresselhaus
et al., 2002). Moreover, the electronic states are highly sensitive to the diameter of
nanotubes. In the resonant Raman effect, a large enhancement in the Raman signal
occurs only when the incident or scattered photon is in resonance with a singular-
ity in the 1D JDOS of nanotubes (Dresselhaus et al., 2002). Therefore, resonance
Raman spectroscopy can be used as a sensitive probe of the geometrical and elec-
tronic structure of CNTs through coupling between electrons and phonons in this
one-dimensional system.
In this review, we will present our recent progress on the controllable synthe-
sis and detailed structural characterization of DWNTs. A floating catalyst chem-
ical vapor deposition (CVD) method was proposed for the synthesis of DWNTs
∗ Shenyang National Laboratory for Materials Science, Institute of Metal Research, Chinese
Academy of Sciences, Shenyang 110016, China


C Springer 2008
30 W. Ren, H.-M. Cheng
with high purity, good alignment and narrow diameter distribution with a relatively
large scale. Moreover, we performed systematical Raman studies on the as-prepared
DWNTs to probe their detailed structure characteristics and several unique Raman
features were found.
4.2 Experimental
The synthesis equipment was described in detail in our previous paper (Cheng et al.,
1998a). In general, methane was used as the carbon source, hydrogen as the carrier
gas, ferrocene as the catalyst precursor, and thiophene was carried into the reactor
by methane to enhance the growth of CNTs. For the DWNT growth, the reaction
tube was first heated to reach 1373 K in a hydrogen atmosphere. Then, methane and
hydrogen, accompanying with the catalyst precursor and thiophene, were introduced
into the reaction tube for 10 min, followed by cooling the CVD system in hydrogen
to room temperature. Depending on the experimental conditions, web-like DWNTs
(Ren et al., 2002) or macroscopically long DWNT strands (Ren and Cheng, 2005)
as long as 10 cm can be selectively formed during the reaction process. It is worthy
of note that this method can be used to synthesize DWNTs in a large scale due to the
continuous introduction of the carbon source, carrier gas, catalysts and thiophene.
The morphology and microstructure of the product were observed using scan-
ning electron microscopy (SEM), high-resolution transmission electron microscopy
(HRTEM JOEL 2010 200kV) and Raman spectroscopy (Jobin Yvon LabRam
HR800, excited by 632.8 nm laser). It should be pointed out that all Raman spec-
tra were taken in backscattering configuration, with the incident and scattered light
propagating perpendicular to the rope axis (Ren et al., 2002, 2005, 2006). For the
polarized Raman studies on the G band of DWNT ropes, the spectra were recorded
with parallel VV and crossed VH polarizations of the incoming and scattered light.
For the related discussions on the selection rules, we define Z as the DWNT axis
direction and Y as the photon propagation direction.
4.3.1 Synthesis of DWNTs
Figure 4.1 and Fig. 4.2 show, respectively, SEM, TEM and HRTEM images of the
as-prepared web-like DWNTs (Ren et al., 2002) and DWNT ropes (Ren and Cheng,
2005). It is apparent that the products are very pure with less amorphous carbon
and catalyst particles attached, which is a common characteristic for the nanotubes
produced by the floating catalyst CVD method. In the DWNT ropes, DWNTs grew
along one direction, which was parallel to the direction of the flow gas in the reac-
tion system. Therefore, we infer that the orientation force for the growth of the
4 Raman Spectroscopy on Double-Walled Carbon Nanotubes 31
Fig. 4.1 SEM, TEM and HRTEM images of the web-like DWNTs (Ren et al., 2002)
long DWNT ropes is from the gas flow. Another point we need point out is that
the diameter distribution of DWNTs prepared by our method is much narrower
than those prepared by hydrogen arc discharge method (Hutchison et al., 2001). For
the web-like product, the inner and outer diameters of DWNTs are in the range of
1.0–2.0 nm and 1.6–2.8 nm, respectively. For the DWNT ropes, the inner and outer
diameter of DWNTs are dominantly in the range of 1.0–1.3 nm and 1.7–2.0 nm,
respectively, which is approach to those prepared by C60 coalescence in SWNTs at
high temperature (Bandow et al., 2001). The successful synthesis of these nanotubes
opens up the possibility for their further property studies and applications.
4.3.2 Radial Breathing Mode of DWNTs
It is well established that the radial breathing mode (RBM) is a particularly valu-
able tool for detecting SWNTs and measuring the diameter of SWNTs (Dresselhaus
et al., 2002). Therefore, probing and analyzing the RBM feature of DWNTs is a key
for their further structural characterization. We consider that the effects of coupling
between adjacent graphene layers of DWNTs on the vibrational and electronic states
are not of sufficient strength to significantly affect the physical properties of each
constituent due to the weak van der Waals interaction, as similar to the intra-bundle
and inter-tube coupling in a SWNT bundle (Dresselhaus and Eklund, 2000). There-
fore, it is not expected that many new Raman modes actually be observed, and
the spectral intensity should still be dominated by the selection rules for the two
constituent SWNTs.
The Raman spectra taken from the web-like sample are shown in Fig. 4.3. It
is worth noting that different from that observed for SWNTs, the RBM band of
Fig. 4.2 Optical photo, and SEM, HRTEM images of DWNT ropes (Ren and Cheng, 2005)
Fig. 4.3 RBM feature of the web-like DWNTs (Ren et al., 2002)
DWNTs mainly consists of two components at low frequency region forming pair
peaks. The two RBM bands centered at 112 and 197 cm−1 . Another characteristic
of the RBM band of DWNTs is that each RBM band has some Raman shifts with
a small frequency interval, which is thought to be caused by the two-graphene
Table 4.1 Raman peak positions and the calculated diameters of DWNTs (Ren et al., 2002)
Outer diameter Inner diameter
ωr (dt ) cm−1 (nm) ωr (dt ) cm−1 (nm)
92.3 (2.75) 128.3 (1.98)
130.9 (1.94)
103.9 (2.44) 146.2 (1.74)
155.2 (1.64)
112.9 (2.25) 164.1 (1.55)
170.5 (1.49)
175.6 (1.45)
119.3 (2.13) 190.9 (1.33)
197.2 (1.29)
128.3 (1.98) 197.2 (1.29)
207.4 (1.22)
212.5 (1.20)
130.9 (1.94) 207.4 (1.22)
212.5 (1.20)
139.8 (1.82) 246.6 (1.03)
146.2 (1.74) 275.6 (0.92)
284.4 (0.89)
layers in the DWNT and their chirality. Table 4.1 summarizes the peak positions
and their calculated tube diameters using the formula: ωr (dt ) = 254(cm−1 nm)/dt
(nm) (Cheng et al., 1998b), where ωr is the Raman shift, and dt is the diameter
of the tubes. It is found that the diameters corresponding to the lower frequency
band obtained by the calculation are almost equal to the outer diameters obtained
by the measurements based on HRTEM images, and the higher frequency band
correspond to the inner diameter. Moreover, HRTEM observation shows that few
SWNTs with diameter about 2.3 and 1.3 nm were found. Therefore, we can conclude
that the two mean peaks are the RBM of the two coaxial SWNTs constituting a
DWNT. Moreover, from Table 4.1, we can see that an outer diameter incorporates
with several possible candidates of inner tubes. This result is consistent with our
HRTEM observations. That is, the interlayer spacing of DWNTs is not a constant,
ranging from 0.37 to 0.41 nm.
Figure 4.4 shows two RBM mappings obtained along the directions parallel and
perpendicular to the DWNT rope axis, respectively. Similar to those of web-like
DWNTs, they mainly consist of two components at low frequency region form-
ing pair peaks. The outer diameters and inner diameters calculated using relation
dt = 254/ωr are roughly in the range of 1.7–2.0 nm and 1.0–1.3 nm, respectively.
The good agreement of there values with those from HRTEM investigations further
prove that these low-frequency Raman peaks are the intrinsic diameter-dependent
RBM feature of DWNTs.
4.3.3 Polarized Raman Studies on the G Band of DWNTs
The symmetry analysis of the G band is helpful for understanding of the interlayer
interaction, the assignments of the mode symmetry, the consequent determinations
of diameters, and the electronic properties of nanotubes (Dresselhaus et al., 2002).
Fig. 4.4 RBM mapping recorded in a step of 0.5 μm along the directions parallel (left panel) and
perpendicular (right panel) to the rope axis, respectively. The ordinate unit is μm and the vertical
color bar represents the intensity of Raman spectra (Ren and Cheng, 2005)
Fig. 4.5 Lineshape analysis for the four different polarized spectra of the DWNT bundles. The
frequencies (widths) of the modes in the figure are displayed in cm−1 (Ren et al., 2005)
As stated above, for DWNTs, it is assumed that the interlayer interactions are similar
to the intertube interactions in a SWNT bundle, so that the free nanotube symme-
tries can also be used to analyze the G-band feature of DWNTs (Dresselhaus et al.,
2002; Jorio et al., 2000). Figure 4.5 shows the lineshape analysis results on the
G band of DWNTs in different polarized geometries. It is noteworthy that all the
modes exhibit different polarization-dependent relative intensities (see Fig. 4.5 and
Table 4.2), suggesting that they are all intrinsic to the DWNTs.
The analysis on the RBM band of aligned DWNT ropes indicates that the inner
and outer tubes excited by 632.8 nm laser are metallic and semiconducting, respec-
tively. This result provides an useful information on the symmetry assignment of
the G band of DWNTs. Based on symmetry analysis on SWNTs, the low frequency
mode 1531 cm−1 was firstly assigned to the A (A1g ) mode of the metallic inner
Table 4.2 Relative intensities of the G-band modes belonging to different irreducible representa-
tions (IR) in the different polarized Raman spectra of DWNTs. For the modes at 1567, 1591 cm−1
and 1531, 1582 cm−1 , we used their measured intensities in the (ZZ) polarized spectra as a refer-
ence, while for the modes at 1556 and 1603 cm−1 , we used the (ZZ) intensity of the 1591 cm−1
mode as a reference (Ren et al., 2005)
ωG (cm−1 ) IR (ZZ) (XX) (ZX) (XZ)
1556 E2 0.12 0.07 0.02 0.02
1567 A1 + E 1 1.00 0.40 0.14 0.16
1591 A1 + E 1 1.00 0.27 0.15 0.14
1603 E2 0.00 0.07 0.00 0.00
1531 A1 1.00 0.32 0.13 0.12
1582 A1 1.00 0.26 0.10 0.11
+
tubes because of the asymmetric BWF line shape. Generally, the ωG is 1590 cm−1
for S-SWNT bundles, and 1582 cm−1 for M-SWNT bundles, which is independent
of dt (Jorio et al., 2000; Brown et al., 2001). Therefore, the mode 1582 cm−1 with a
Lorentzian line shape was assigned to high frequency A (A1g ) mode of the metallic
inner tubes. Moreover, the intensity ratios I(Z Z ):I(X X ) for the above two modes
are consistent with the theoretical value of 1.00:0.25. The residual modes with
Lorentzian line shape were all assigned to originate from the semiconducting outer
tubes. Our experimental results show that the highest frequency 1603 cm−1 tangen-
tial mode only appears with a relatively low intensity in the (XX) parallel-polarized
spectra (see Fig. 4.5 and Table 4.2). This result is in agreement with group the-
ory predictions for the E 2 (E 2g ) mode. The 1556 cm−1 mode was also assigned to
E 2 (E 2g ) symmetry of the semiconducting outer tubes, because of its lower intensity
in all the polarization configurations and the approximately equal frequency shifts
with 1603 cm−1 from the central frequency of 1580 cm−1 for graphite. The peaks
at 1567 and 1591 cm−1 exhibit a comparable intensity ratio I(Z Z ) :I(X X ) to the theo-
retical value of 1.00:0.25 for A symmetry modes (Saito et al., 1998). Moreover, the
cross-polarized spectra also exhibit the above two features. Therefore, we assigned
them to two A (A1g ) + E 1 (E 1g ) modes of the semiconducting outer tubes.
To further determine the effect of interlayer interaction on the G band, we per-
formed a comparison of mode symmetry and frequency for SWNTs and DWNTs
with a similar diameter and the same electronic property. Previous research results
indicated that, for S-SWNT bundles with an average diameter of 1.85 nm, four
peaks 1549, 1567, 1590, and 1607 cm−1 were found and respectively assigned to
the E 2 (E 2g ), two A (A1g ) + E 1 (E 1g ), and high frequency E 2 (E 2g ) mode (Jorio
et al., 2000). Our assignment results for the semiconducting outer tubes are con-
sistent with the above results except for small frequency shifts due to the different
average diameters. Moreover, it is worth noting that the parameters of the BWF
lineshape for metallic inner tubes, such as the frequency and linewidth, are in good
agreement with those for the M-SWNTs with similar diameters (Brown et al., 2001).
All these results indicate that the two constituent tubes independently retain their
original mode symmetries and vibrations, consistent with the theoretical predictions
(Damnjanovic et al., 2003) and above assumption.
4.3.4 D Band and G Band of DWNTs
For a SWNT, it was well established that its structure can be uniquely determined
by its (n, m) indices (Dresselhaus et al., 2002). However, a DWNT cannot be deter-
mined only by two different (n, m) combinations, but also by the atomic arrangement
of its two constituent tubes, i.e., their relative circumferential rotation angle and their
axial translation is an important determining parameter (Saito et al., 1993). More-
over, theoretical calculations for energy bands of DWNTs have concluded that their
electronic structures are sensitive to their structural symmetry, that is, the atomic
correlation between adjacent nanotubes (Saito et al., 1993, and Sanvito et al., 2000).
Fig. 4.6 Raman spectra of the RBM, D band and G band measured with a laser energy of 1.96
eV of (a) sample containing more than 90% DWNTs with outer and inner diameter of 1.85 ± 0.15
and 1.15±0.15 nm, respectively, (b) sample containing about 30% DWNTs and 70% SWNTs with
the similar outer diameter of 1.85 nm, and (c) SWNTs with average diameter of 1.85 nm. The inset
shows a fit to the D band of DWNTs with a larger diameter of 2.26 nm. A Lorentzian line shape
was used in the fitting process. The frequencies/widths of the observed modes are also displayed
(Ren et al., 2006)
Therefore, it is quite important to identify the atomic correlation between the two
constituent tubes of a DWNT and to determine its electronic structure for electronic
applications. Based on the physical origin of the D-band and G -band features,
i.e., the double resonance mechanism, their Raman profiles are very sensitive to
the electronic structure of the nanotubes. Therefore, we will try to probe the atomic
correlation of the two constituent tubes of DWNTs by analyzing their D-band and
G -band features.
Figure 4.6 shows the D-band and G -band Raman features of different samples
measured with the 1.96 eV laser excitation, in which Fig. 4.6a is from the DWNT
rope with a narrow diameter distribution, Fig. 4.6b from the sample containing
about 30% DWNTs, and Fig. 4.6c from the SWNTs with an average diameter of
1.85 nm. It is clear that the D band and G band of the samples containing DWNTs
are composed of four well-resolved peaks and those of the SWNTs are a single peak.
The inset of Fig. 4.6 shows the D band of DWNTs with a larger outer diameter of
2.26 nm. Note that the four-peak Raman features of the D band and G band could
be observed in both DWNT samples with different mean diameters, and their fre-
quencies were dependent on the mean nanotube diameter. Moreover, the four peaks
can be reproducibly found in the D band at the different positions of the sample,
as shown in Fig. 4.7, with the same position and linewidth. Therefore, we conclude
that the four components arise from the intrinsic structural features of DWNTs.
It is important to note that the four components in the D band of DWNTs
have comparable intensity and exhibit very small linewidth of about 9–13 cm−1 ,
comparable to the intrinsic linewidth of a single SWNT (Jorio et al., 2002). This
result indicates that (1) incident and scattered phonons are very close to the vHSs
Fig. 4.7 The reproducible

D-band feature at different
positions of an aligned
DWNT rope (Ren et al.,
2006)
(Jorio et al., 2002) (i.e., satisfying the triple resonance condition), and (2) the den-
sity of states is very singular, leading to a long phonon lifetime. Moreover, it was
worthy of noting that only one main peak was found in the D-band feature for
an upshift of about 0.2 eV in the E ii energy transition resulting from the intertube
interactions in SWNT bundles (Souza Filho et al., 2001). Therefore, we deduce that
the occurrence of additional vHSs in the JDOS plays a key role in the unique feature
of the D band and G band of DWNTs.
Theoretically, the incommensurate DWNTs have only 2π rotational symmetry
along the chiral vectors which are both perpendicular to the nanotube axis (Saito
et al., 1993). Therefore, the electronic property of an incommensurate DWNT, can
be considered to be the sum of the electronic structures of two independent nan-
otubes, expect for some small modification due to a weak interlayer interaction. As
a consequence, two main peaks could be expected in the D band of the incom-
mensurate DWNTs, having approximate frequencies with those of independent
SWNTs with similar diameters. According to the relation between ω D and dt for
Elaser = 1.96 eV, ω D = 1331−16.5/dt (Dresselhaus et al., 2002), the ω D of the outer
tube with a mean diameter of about 1.85 nm can be predicted to be ∼ 1322 cm−1 ,
which is in a good agreement with the observed high frequency component of the
D band (1322 cm−1 ). For an average interlayer spacing of 0.37 nm, the ω D of the
inner tube should be ∼ 1316 cm−1 (corresponding to 1.11 nm), which approaches
the observed 1313 cm−1 . Therefore, the inner two peaks can be assigned to origi-
nate from the incommensurate DWNTs. The obtained D-band frequency agreement
between DWNTs and SWNTs with the similar diameters provides a strong evidence
for the weak interlayer interaction between the two constituents of the incommen-
surate DWNTs.
Commensurate DWNTs have a better atomic correlation and a stronger van der
Waals interaction between the two constituents than those of incommensurate ones.
Two important changes were predicted theoretically for the electronic structures
of commensurate DWNTs: occurrence of additional vHSs due to the separation
of original degenerated energy bands, and very singular density of states at many
anticrossings of energy bands due to the interlayer interaction (Saito et al., 1993).
The occurrence of additional vHSs makes it possible to give rise to some additional
vHSs satisfying the triple resonance conditions. Moreover, the very singular den-
sity of states makes related D-band frequencies distinguishable due to the resulting
very narrow linewidth. Consequently, some new peaks were expected to appear in
the D band of commensurate DWNTs, having different frequencies with those of
incommensurate DWNTs. Therefore, the two observed outer peaks in the D band
can be deduced to originate from the occurrence of additional vHSs related to a
stronger interlayer interaction for the commensurate DWNTs in the sample. The
above analysis can be further proved by the recent studies on the D band and G
band of graphene, where only one peak is found for the monolayer graphene, and
four peaks are observed for the bi-layer graphene with AB stacking due to the modi-
fication of electronic structures by strong interlayer interaction (Ferrari et al., 2006).
4.4 Conclusion
A brief review about the synthesis and Raman spectroscopy of DWNTs is here pre-
sented, emphasizing in particular how such spectra can be observed for DWNTs
and the use of these spectra for the further structural characterization of DWNTs. A
floating CVD method was proposed for the synthesis of DWNTs with high purity,
good alignment and a very narrow diameter distribution. Double-peak feature was
found in the RBM of DWNTs, which is considered to be associated with the outer
and inner diameter, respectively. From the polarized Raman studies on the G band
of DWNTs, it was found that the two constituent tubes of a DWNT independently
remain their symmetries and vibrations. Four well-resolved peaks with very narrow
linewidths were found in the D-band and G -band features of double-walled car-
bon nanotubes (DWNTs). Combined with theoretical predictions, we constitute the
first Raman evidence for atomic correlation and the resulting electronic structure
modification of the two constituent tubes in DWNTs. However, we must point out
that the obtained results about Raman spectroscopy of DWNTs are based on bun-
dles and further investigations on individual nanotubes are required to justify these
deductions.
Acknowledgments This work was supported by National Natural Science Foundation of China
(No. 90606008), Ministry of Science and Technology (No. KJCX2-YW-M01), and Chinese
Academy of Sciences 2006CB932701.
References
Bandow, S., Takizawa, M., Hirahara, K., Yudasaka, M., and Iijima, S., 2001, Raman scattering
study of double-wall carbon nanotubes derived from the chains of fullerenes in single-wall
carbon nanotubes. Chemical Physics Letters, 337, pp. 48–54.
Brown, S.D.M., Jorio, A., Corio, P., Dresselhaus, M.S., Dresselhaus, G., Saito, R., and Kneipp, K.,
2001, Origin of the Breit-Wigner-Fano lineshape of the tangential G-band feature of metallic
carbon nanotubes. Physical Review B, 63, p. 155414.
Cheng, H.M., Li, F., Su G., Pan, H.Y., He, L.L., Sun, X., and Dresselhaus, M.S., 1998a, Large-
scale and low-cost synthesis of single-walled carbon nanotubes by the catalytic pyrolysis of
hydrocarbons. Applied Physics Letters, 72, pp. 3282–3284.
Cheng, H.M., Li, F., Sun, X., Brown, S.D.M., Pimenta, M.A., Marucci, A., Dresselhaus, G., and
Dresselhaus, M.S., 1998b, Bulk morphology and diameter distribution of single-walled carbon
nanotubes synthesized by catalytic decomposition of hydrocarbons. Chemical Physics Letters,
289, pp. 602–610.
Damnjanovic, M., Dobardzic, E., Milosevic, I., Vukovic, T., and Nikolic, B., 2003, Lattice dynam-
ics and symmetry of double wall carbon nanotubes. New Journal of Physics, 5, pp. 148.
Dresselhaus, M.S. and Eklund, P.C., 2000, Phonons in carbon nanotubes. Advances in Physics, 49,
pp. 705–814.
Dresselhaus, M.S., Dresselhaus, G., Jorio, A., Souza Filho, A.G., and Saito, R., 2002, Raman
spectroscopy on isolated single wall carbon nanotubes. Carbon, 40, pp. 2043–2061.
Ferrari, A.C., Meyer, J.C., Scardaci, V., Casiraghi, C., Lazzeri, M., Mauri, F., Piscanec, S.,
Jiang, D., Novoselov, K.S., Roth, S., and Geim, A.K., 2006, Raman spectrum of graphene and
graphene layers. Physical Riview Letters, 97, pp. 187401.
Hutchison, J.L., Kiselev, N.A., Krinichnaya, E.P., Krestinin, A.V., Loutfy, R.O., Morawsky, A.P.,
Muradyan, V.E., Obraztsova, E.D., Sloan, J., Terekhov, S.V., and Zakharov, D.N., 2001,
Double-walled carbon nanotubes fabricated by a hydrogen arc discharge method. Carbon, 39,
pp. 761–770.
Jorio, A., Dresselhaus, G., Dresselhaus, M.S., Souza, M., Dantas, M.S.S., Pimenta, M.A.,
Rao, A.M., Saito, R., Liu, C., and Cheng, H.M., 2000, Polarized Raman study of single-wall
semiconducting carbon nanotubes. Physical, Review Letters, 85, pp. 2617–2620.
Jorio, A., Fantini, C., Dantas, M.S.S., Pimenta, M.A., Souza Filho, A.G., Samsonidze, Ge, G.,
Brar, V.W., Dresselhaus, G., Dresselhaus, M.S., Swan, A.K., Ünlü, M.S., Goldberg, B.B., and
Saito, R., 2002, Linewidth of the Raman features of individual single-wall carbon nanotubes.
Physical Review B, 66, p. 115411.
Ren, W.C. and Cheng, H.M., 2005, Aligned double-walled carbon nanotube long ropes with a
narrow diameter distribution. Journal of Physical Chemistry B, 109, pp. 7169–7173.
Ren, W.C., Li, F., and Cheng, H.M., 2005, Polarized Raman analysis of aligned double-walled
carbon nanotubes. Physical Review B, 71, p. 115428.
Ren, W.C., Li, F., Chen, J., Bai, S., and Cheng, H.M., 2002, Morphology, diameter distribution and
Raman scattering measurements of double-walled carbon nanotubes synthesized by catalytic
decomposition of methane. Chemical Physics Letters, 359, pp. 196–202.
Ren, W.C., Li, F., Tan, P.H., and Cheng, H.M., 2006, Raman evidence for atomic correlation
between the two constituent tubes in double-walled carbon nanotubes. Physical Review B, 73,
p. 115430.
Saito, R., Dresselhaus, G., and Dresselhaus, M.S., 1993, Electronic-structure of double-layer
graphene tubules. Journal of Applied Physics, 73, pp. 494–500.
Saito, R., Takeya, T., Kimura, T., Dresselhaus, G., and Dresselhaus, M.S., 1998, Raman intensity
of single-wall carbon nanotubes. Physical Review B, 57, pp. 4145–4153.
Sanvito, S., Kwon, Y.K., Tomanek, D., and Lambert, C.J., 2000, Fractional quantum conductance
in carbon nanotubes. Physical Review Letters, 84, pp. 1974–1977.
Shimada, T., Sugai, T., Ohno, Y., Kishimoto, S., Mizutani, T., Yoshida, H., Okazaki, T., and
Shinohara, H., 2004, Double-wall carbon nanotube field-effect transistors: Ambipolar transport
characteristics. Applied Physics Letters, 84, pp. 2412–1414.
Souza Filho, A.G., Jorio, A., Dresselhaus, G., Dresselhaus, M.S., Saito, R., Swan, A.K.,
Ünlü, M.S., Goldberg, B.B., Hafner, H., Lieber, M., and Pimenta, A., 2001, Effect of quantized
electronic states on the dispersive Raman features in individual single-wall carbon nanotube,
Physical Review B, 65, p. 035404.
Chapter 5
Interface Design of Carbon Nano-Materials
for Energy Storage
Feng Li∗ , Hong-Li Zhang, Chang Liu and Hui-Ming Cheng
5.1 Introduction
Lithium ion batteries (LIB) have been a forerunner and market leader since late
1990s in the field of rechargeable battery industry especially for portable electronic
devices, and are also very attractive in electric vehicles and hybrid electric vehicles.
In order to achieve high-performance LIB, developing novel electrode materials,
on one hand, is pivotal; on the other hand, a deeper understanding of the related
electrochemical phenomena along with electrode reactions is definitely necessary
(Tarascon and Armand, 2001).
It is well known that the interface between electrode materials and electrolytes
plays an important role in affecting the first columbic efficiency, cyclability, rate
capability, and safety of LIB (Aurbach, 2003). Especially for carbonaceous anode
materials widely used in current LIB, there is a kind of solid electrolyte interphase
(SEI) film formed on the surface due to electrolytes decomposition (Balbuena and
Wang, 2004). The SEI film is insulating for electrons but conductive for Li ions,
and can significantly influence the cyclability of carbon anode. Here, we introduce
the formation and evolution of SEI film on the surface of natural graphite (NG)
spheres, which show obvious advantages over artificially graphitized meso-phase
carbon micro-beads owing to low cost and intrinsically high crystallinity (Vetter
et al., 2005). Based on the interface results, we propose core-shell design to modify
the electrode materials for LIB by the chemical vapour deposition process. We can
obtain the full coating amorphous carbon core-shell structure without the catalyst
and nanotube/nanofiber coating nano/micro urchin-like structure with catalyst by
chemical vapor deposition (Zhang et al., 2005, 2006a,b).
5.2 Experimental
An electrochemical cell was assembled in Ar-filled glove box (Mbraun, Unilab, H2 O

and O2<1ppm) with a working electrode composed of anode materials (85 wt%),


C Springer 2008
42 F. Li et al.
carbon black (5 wt%) and poly(vinylidene fluoride) (PVDF) binder (10 wt%), a
lithium foil counter electrode and a porous separator (Celgard, 2400). The elec-
trolyte was 1 M LiPF6 in a mixture of EC/DMC (1:1 volume ratio) (battery grade).
The cells were then galvanostatically charged and discharged for certain cycles in
the voltage range of 0.001–2.5 V versus Li/Li+ at a current density of 0.2 mA/cm2 .
The detailed description on the characterization of cycled electrodes by focused
ion beam (FIB) workstation and the preparation of the core-shell composite anode
materials can be found in (Zhang et al., 2005, 2006a,b).
5.3.1 SEI Film Observations
Figure 5.1a shows the SEM image of a typical original natural graphite sphere, we
can see that there are many cracks and flakes on the surface. Figure 5.1b–d present
the secondary electron FIB images and corresponding elemental line scan analysis
(ELSA) curves of a NG sphere that was from a working electrode discharged to
0 V from OCV. In Fig. 5.1b, a rough film can be obviously discerned on the surface
of the NG sphere, which shields the original surface cracks as shown in Fig. 5.1a.
Moreover, a small split (Fig. 5.1c) on the film is also observed. The split is possibly
due to the increase of internal pressure from gas production causing the expansion
of the graphite sphere. The ELSA curves in Fig. 5.1d indicate the relative content
variation of fluorine (F), oxygen (O), carbon (C) and phosphorus (P) along the black
line in Fig. 5.1c. According to the variation of peaks and valleys in the ELSA curves,
it can be concluded that the surface species of the freshly formed SEI film are dis-
tributed non-uniformly, which is due to different reactivity toward electrolytes on
different local parts of NG spheres.
With the aid of FIB workstation, we successfully cut the NG sphere as shown
in Fig. 5.1b, and obtain a clear cross-section image as displayed in Fig. 5.2a. The
thickness of SEI film on the surface is directly measured, and an element analysis
is performed inside the sphere. The results of ELSA curves demonstrate that there
are peaks of F and O around the internal cracks, which reveals that electrolytes can
penetrate into the inside of NG spheres through the surface cracks and decompose
there to form a kind of internal SEI film.
Figure 5.3 presents the FIB image and ELSA curves of a NG sphere experienced
24 discharge/charge cycles. It can be seen that the SEI film splits upon repeated
cycles due to the increase of internal pressure from gas (e.g. C2 H2 ) production inside
the sphere and of volume change stress caused by intercalation/deintercalation of
Li+ . The ELSA curves (Fig. 5.3b) show alternate peaks and valleys around the
cracks, and the average content of each element keeps nearly horizontal in the whole
scanning range. The result indicates that the SEI film becomes relatively uniform in
composition after reaching its stable state.
5 Interface Design of Carbon Nano-Materials for Energy Storage 43
Fig. 5.1 (a) SEM image of a typical NG sphere; (b and c) FIB secondary electron images showing
the surface of a NG sphere which have experienced the first discharge process; (d) ELSA curves
along the black line in (c) (Zhang et al., 2005)
Fig. 5.2 (a) A secondary electron FIB image showing the cross section of a NG sphere experienced
the first discharge cycle; (b) ELSA curves along the white line in (a) (Zhang et al., 2005)
44 F. Li et al.
Fig. 5.3 (a) A secondary electron FIB image showing the changes of surface SEI film of a NG
sphere experienced 24 cycles; (b) ELSA curves along the white line in (a) (Zhang et al., 2005)
5.3.2 Electrode Materials Design
In order to avoid the occurrence of internal SEI film and the formation of a thick
SEI film, the core-shell structure can be used as controlling the SEI film. There-
fore, we propose core-shell design to modify the electrode materials for LIB by the
chemical vapour deposition process (Fig. 5.4). The advantages of this design are
(i) forming a continuous conductive network in the bulk of electrodes to decrease
electrode polarization; (ii) improving the adsorption and immersion of electrolytes
on the surface of electroactive materials to facilitate the electrode reaction kinet-
ics; (iii) as a buffer among electroactive materials due to their superior resiliency;
(iv) being a kind of electroactive material itself without lowering the whole capacity
of anode materials notably.
We obtained pyrolytic carbon coated NG core-shell structure and nan-
otube/nanofiber coating nano/micro urchin-like structure with catalyst by chemical
vapor deposition.
Fig. 5.4 The evolution of electrode materials of LIB

5.3.3 Characterizations of Electrode Materials
Figure 5.5a and b presents more direct evidence about the existence of pyrolytic
carbon layer by the cross-section images of a coated NG sphere, which was obtained
using a FIB workstation that has the ability to accurately cut a specimen (in the scale
of micrometers or nanometers) at any user-defined position and clearly display cor-
responding secondary electron images. An obvious layer in white contrast can be
identified at the periphery of the NG sphere, indicating that the pyrolytic carbon
layer has a thickness of 250 nm. According to the thickness and some other basic
data such as the average diameter of NG spheres, the densities of NG spheres and
pyrolytic carbon, one can estimate that the approximate weight percent of deposited
carbon was 13%, which was very close to the result (15%) obtained by a ther-
mogravimetric analysis (TG) method. Meanwhile, many cracks inside the graphite
sphere are also clearly shown in Fig. 5.5a, which are unavoidable during the sphere-
making process from common flaky graphite. NG spheres were coated in one step by
a uniform layer of pyrolytic carbon with a thickness of 250 nm. The coated sample
as anode material of a lithium-ion battery displayed excellent cyclability, withthe
capacity above 320 mAh/g at the 15th cycle in comparison with less than 200 mAh/g
for the original NG spheres. Meanwhile, the first CE was also improved to 88% from
original 80%. The remarkably improved electrochemical properties for the coated
NG spheres can be mainly attributed to the following factors: the lowered specific
surface area, the inhibition of internal SEI film formed around the internal cracks,
and the formation of a thin and compact SEI film on the outer surface of the coated
NG spheres.
In order to avoid the occurrence of internal SEI film and the formation of a thick
SEI film, we propose to grow CNFs on the surface of NG spheres in situ by catalytic
chemical vapor deposition. In this way, a nano/micro hybrid urchin-like composite
(NG/CNF) is achieved as shown in Fig. 5.6a, where NG spheres act as inner cores
and CNF as surrounding spines. From the Raman spectra in Fig. 5.6b, it can be noted
that the grown-CNF is of non-graphitic. Therefore, the urchin-like composite is also
a combination of graphitic (core) and non-graphitic (spines) structure. Figure 5.6c
Fig. 5.5 Secondary electron FIB images showing the cross section of a coated NG sphere. The
inset in panel (a) is the magnified image of the rectangle; panel (b) is the magnified image of the
circle area in panel (a). (Zhang et al., 2006a).
46 F. Li et al.
Fig. 5.6 (a) A SEM image showing the urchin-like NG/CNF in situ composite; (b) Raman spectra
of the pristine NG and the NG/CNF in situ composite; (c and d) Comparison of cyclability and
rate capability among the pristine NG, the NG with CNF as direct fillers, and the NG/CNG in situ
composite (Zhang et al., 2006b)
displays the comparison of cyclability among the pristine NG spheres, the NG with
CNF as direct fillers, and the NG/CNF in situ composite. It is clearly seen that a
great improvement in cycling performance is achieved for the in situ composite, the
capacity of which is still kept at above 300 mAh/g up to 45 cycles, about 150 mAh/g
higher than that of the pristine NG. On the contrary, the capacity of the pristine NG
and the NG with CNF as fillers drops continuously in the whole cycles. Figure 5.4b
displays rate capability at different C-rates ranging from 0.1 to 1.2 C. Again, the
NG/CNF in situ composite shows a better behavior than the pristine NG and the NG
with CNF as fillers.
5.4 Conclusion
The interfaces with electrolytes are key factors for carbon nano-materials in their
applications of energy storage. In this report, the formation and evolution of the SEI
film on the surface of natural graphite spheres in the electrolyte of 1 M LiPF6 in
ethylene carbonate and diethyl carbonate were investigated by using FIB technique.
A concept of “internal SEI film” is first proposed based on the characterization of
the cross section of the natural graphite spheres with the aid of FIB. Secondly, based
on the results from interface investigation, urchin-like nano/micro hybrid materials
are designed and achieved to modify conventional micrometer-sized electroactive
materials for LIB. Moreover, a pyrolytic carbon coated NG core-shell structure and
a nanotube/nanofiber coated NG nano/micro urchin-like structure were prepared.
These composite anode materials display significantly improved cyclability and
rate capability in comparison with the pristine NG. We can obtain the full coating
amorphous carbon core-shell structure without the catalyst and nanotube/nanofiber
coating nano/micro urchin-like structure with catalyst by chemical vapor deposition.
Electrochemical measurements indicate that the cyclability and rate capability of the
composite as anode material for LIB are significantly improved. Furthermore, the
design also demonstrates its effectiveness in other kinds of anode and cathode mate-
rials such as transition metal oxides. These results may be inspirable for exploring
high performance electrode materials.
Acknowledgments This work was supported by National Natural Science Foundation of China
(No. 50632040, 50328204).
References
Aurbach, D., 2003, Electrode–solution interactions in Li-ion batteries: a short summary and new
insights. Journal of Power Sources, 119–121, pp. 497–503.
Balbuena, P.B., Wang, Y.X. (eds) 2004, Lithium-Ion Batteries Solid-Electrolyte Interphase (Impe-
rial College Press).
Tarascon, J.-M. and Armand, M., 2001, Issues and challenges facing rechargeable lithium batteries.
Nature, 414, p. 359
Vetter, J., Novák, P., Wagner, M.R., Veit, C., Möller, K.-C., Besenhard, J.O., Winter, M., Wohlfahrt-
Mehrens, M., Vogler, C. and Hammouche, A., 2005, Ageing mechanisms in lithium-ion batter-
ies. Journal of. Power Sources, 147, p. 269
Zhang, H. L., Li, F., Liu, C., Tan, J., Cheng, H. M., 2005, New insight into the solid electrolyte
interphase with use of a focused ion beam. Journal of Physical Chemistry B, 109, pp.
22205–22211.
Zhang, H.L., Liu, S.H., Li, F., Bai, S., Liu, C., Tan, J., Cheng, H.M., 2006a, Electrochemical
performance of pyrolytic carbon-coated natural graphite spheres. Carbon, 44, pp. 2212–2218.
Zhang, H.L., Zhang, Y., Zhang, X.G., Li, F., Liu, C., Tan, J., Cheng, H.M., 2006b, Urchin-like
nano/micro hybrid anode materials for lithium ion battery, Carbon, 44, pp. 2778–2784.
Chapter 6
Formation Mechanism of 0.4 nm Single-Walled
Carbon Nanotubes in CoAPO-5 Single Crystals
J. P. Zhai, I. L. Li, Z. M. Li, J. T. Ye and Z. K. Tang∗
6.1 Introduction
Carbon nanotubes (Iijima, 1991) have been extensively studied in the past decade
due to their promising applications, which largely derives from their particular
structural and electronic properties. Especially, single-walled carbon nanotubes
(SWNTs) can behave as metal or semiconductor depending on the tube diame-
ter and chirality. Therefore, it is very important to control the purity in diameter,
chirality and alignment of SWNTs. In our previous work, we reported fabrication
of mono-dispersed SWNTs of diameter as small as 0.4 nm inside the channels of
AlPO4 -5 (structure code AFI) zeolite crystals (Tang et al., 1998; Wang et al., 2000).
They constitute the best example of near one-dimensional (1D) quantum wires, and
show extremely interesting physical properties (Tang et al., 2001; Li et al., 2001;
Mei et al., 2005). Although the 0.4 nm SWNTs have been systematically studied,
some of their basic physical and chemical properties are not yet well understood.
For instance, the understanding of the formation mechanism is still not precisely
known.
The 0.4 nm SNTs are synthesized by pyrolyzing tripropylamine (TPA) hydro-
carbon molecules which are incorporated inside the channels during hydrothermal
growth of the AlPO4 -5 crystal (Tang et al., 1998). However, the mechanisms by
which the TPA molecules transform to the SWNTs inside the channels are poorly
understood. This problem arises from the fact that the channel wall is relatively
inertial and has very weak local dipole moment because of the regular alteration
of tetrahedral (AlO4 )− and (PO4 )+ in AlPO4 -5 framework. Thus, the adsorption
force of the channel wall to guest molecules is relatively weak. As a result, a signif-
icant amount of TPA organic molecules can escape from the channels before they
are thermally decomposed. The negatively charged framework and Brønsted acid
sites are generated, when Al3+ is replaced by divalent cation Co2+ in the AlPO4 -5
frameworks, which can play an important catalytic role and increase the adsorp-
tion force by improving the adsorption potential on the channel walls, the resulting
∗ Department of Physics, Hong Kong University of Science and Technology, Clear Water Bay,
Kowloon, Hong Kong, China


C Springer 2008
50 J. P. Zhai et al.
SWNTs have higher quality and density (Li et al., 2004b; Zhai et al., 2006a,b).
These works led us to explore the carbonization mechanism of organic species in
CoAPO-5 crystal channels in more detail and optimize SWNTs synthesis process
further.
In this paper, we report the occurred forms and detailed decomposition process of
organic molecules TPA confined in the channels of CoAPO-5 crystals. By coupling
micro-Raman spectrometry and mass spectrometry, it has been possible to show
that tripropylammonium cation decomposition occurs via abstraction of propylene
and stepwise formation of dipropyl- and n-propylammonium cations. Micro-Raman
measurements at various temperatures show that the 0.4 nm nanotubes are formed
at about 673 K. Compared with that of AlPO4 -5 single crystal, the carbonization
temperature of organic species inside the channels of CoAPO-5 crystals is notably
decreased.
6.2 Experimental
An AFI crystal is a type of microporous aluminophosphate crystallites. Its frame-

work consists of alternating tetrahedral (AlO4 )− and (PO4 )+ , which form parallel
open channels packed in the hexagonal structure. The inner diameter of the 12-ring
channel is 0.73 nm, and the distance between two neighboring parallel channels is
1.37 nm. During the hydrothermal growth of the AFI crystals, TPA molecules were
introduced into the channels as growth template. The TPA molecules are oriented
in a head-tail manner along the [001] direction of the channels (Qiu et al., 1989). In
this work, CoAPO-5 single crystals were synthesized by the hydrothermal method.
Cobalt acetate, aluminum tri-isopropoxide [(iPrO)3 Al] and phosphoric acid (H3 PO4
85 wt. %) were used as cobalt, aluminum and phosphorus sources, respectively.
The aluminum tri-isopropoxide and cobalt acetate were first hydrolyzed in water,
and then H3 PO4 , and TPA template was dropwisely added into the aluminum tri-
isopropoxide solution under vigorous stirring. The gel formed from the final reaction
mixture was sealed in a Teflon-lined stainless autoclave and heated at 448 K for 24
hours. The solid products were filtered, washed with distilled water, and dried at
353 K in atmosphere.
The composition of the CoAPO-5 crystal was analyzed by X-ray fluorescence
(XRF) using element analyzer JSX-3201Z. The volatiles evolved from CoAPO-5
crystals during pyrolysis process were measured by an ABB mass spectrometer.
In order to eliminate the influence of physisorbed water inside the channels, the
as-synthesized sample is firstly treated under a vacuum of 10−3 mbar at 373 K for
3 hours. After the water was desorbed, these crystals were treated at temperatures
between 373 and 873 K, with a heating rate of 3 K/min. Raman spectra of the sam-
ples were measured at various temperatures by Jobin Yvon-T64000 micro-Raman
spectroscopy using 514.5 nm line of an Ar ion laser excitation. The equipped CCD
detector was cooled using liquid nitrogen. The incident laser light was polarized
parallel to the tube axis.
6 Formation Mechanism of 0.4 nm Single-Walled Carbon Nanotubes 51
The as-grown CoAPO-5 crystals have typical size of ∼ 50 μm in diameter and

∼ 400 μm in length. As confirmed by X-ray diffraction and FTIR, the CoAPO-5
crystal has the same crystal structure as that of AlPO4 -5 crystals. The composition
of crystal is tested by X-ray fluorescence (XRF). Figure 6.1 shows the composition
of the CoAPO-5 crystal plotted as a function of the cobalt molar ratio in the starting
gel. The vertical axes are the stoichiometric ratio of Al, P (left) and Co (right) in
the CoAPO-5 crystals. The analytical data indicated an increase in the cobalt con-
tent in the as-grown CoAPO-5 crystals with increasing the cobalt molar ratio x =
[Co]/([Co] + [Al]) in the starting gel, while the aluminium content in the crystals is
correspondingly decreased. In contrast, the phosphorus content in the crystals kept a
constant. These results strongly suggested that Al3+ was substituted by Co2+ , which
lead to the generation of negatively charged framework and Brønsted acid sites.
The vibration spectra provide useful information about the organic template
molecules inside the channels of the as-grown CoAPO-5 crystals. Micro-Raman
spectra of TPA molecules inside the channels of the CoAPO-5 crystal in the C−H
bond-stretching region (2700 ∼ 3150 cm−1 ) measured at ambient condition are
shown in Fig. 6.2. The spectra of liquid TPA and the aqueous solution of tripropy-
lammonium fluoride are also shown in the figure as a reference. From this figure,
we can observe that the C−H bond-stretching modes of the TPA inside the channels
of the CoAPO-5 single crystal at 2890, 2948, and 2975 cm−1 obviously shifted
Fig. 6.1 The molar ratios of cobalt, aluminium, and phosphorus measured by the X-ray fluores-
cence element analyzer
Fig. 6.2 Raman spectra of:

(a) CoAPO-5 single crystal
with TPA inside the channels,
(b) tripropylammonium
fluoride, aqueous solution,
and (c) liquid TPA
to a higher frequency in comparison with that of liquid-phase TPA. The corre-

sponding shift in the CH3 anti-symmetrical deformation mode at 1455 cm−1 (not
shown) has also been observed. The Raman-active vibration frequencies of the TPA
molecules in the CoAPO-5 crystals channels are close to those of the tripropylam-
monium fluoride. Thus, it can be expected that the TPA molecules are protonated
inside the channels of CoAPO-5 single crystal. The TPA molecules in the as-grown
CoAPO-5 crystal channels may have three predominating forms: (1) tripropylam-
monium hydroxide ((CH3 CH2 CH2 )3 NH+ OH− ), (2) tripropylammonium fluoride
((CH3 CH2 CH2 )3 NH+ F− ) when HF is used to restrain the nucleation and slow
down the crystallization during the crystal growth, and (3) tripropylammonium
cation compensating of the negatively charged framework.
A mass spectrometer was used to in-situ monitor the pyrolysis process of
the TPA molecules accommodated in the CoAPO-5 crystals. The mass spectra
recorded at temperatures ranging from 373 to 623 K are shown in Fig. 6.3. By
elevating temperature from 300 K up to 473 K, the signal intensity of water
molecules (m/z = 18 and 17) converted from the tripropylammonium hydroxide
was increased gradually. Then the signals at m/z=114, 86, 72, 41 and 30 resulted
from the neutral (CH3 CH2 CH2 )3 N molecules were observed. The intensity of these
signals increased when the sample was heated to 623 K, implying the conversion of
(CH3 CH2 CH2 )3 NH+ OH− to (CH3 CH2 CH2 )3 N occurred in the temperature region
of 473 ∼ 648 K:
(C H3 C H2C H2 )3 N H + O H − → (C H3 C H2C H2)3 N + H2 O, (373 − 473K )
It is worth pointing out that the water signals decreased in the temperature range
473 ∼ 598K, meanwhile the signal intensity of TPA increased. This result can be
attributed to the fact that the occluded tripropylammonium fluoride is more stable
than the hydroxide, which is comparable to the results of Soulard et al. (1987) who
Fig. 6.3 Mass spectra obtained at various points during the thermal pyrolysis of template occluded
within CoAPO-5 crystals in the temperature range 373 ∼ 623 K
investigated decomposition of tetrapropylammonium ions in MFI-type zeolites by

means of thermoanalytical techniques.
(CH3 CH2 CH2 )3 NH+ F− → (CH3 CH2 CH2 )2 N + HF, (598–648 K).
Figure 6.4 shows the mass spectra of the carbon precursors measured in a higher
temperature region (648–823K). The signal intensity of the TPA molecules started
Fig. 6.4 Mass spectra obtained at various points during the thermal pyrolysis of template occluded
within CoAPO-5 crystals in the temperature range 648 ∼ 823 K
to decrease when the temperature is increased to 648 K, meanwhile, the signals of

propylene at m/z = 42, 41, 39 and 27 and ammonia (NH3 ) at m/z = 17 and 16
were observed. The carbon precursor molecules were decomposed stepwisely into
lighter molecules through sequential abstraction of propylene. It is interesting to
note that, no signals of (CH3 CH2 CH2 )2 NH and (CH3 CH2 CH2 )NH2 were observed
in these mass spectra, although the signals of propylene and ammonia molecules
were clearly seen. This is due to the fact that the (CH3 CH2 CH2 )3 NH+ cations are
strongly attracted by the negative charged lattice (Z− ). The subsequent pyrolysis
reaction of the (CH3 CH2 CH2 )3 NH+ Z− can be described by the following reaction
equations:
(CH3 CH2 CH2 )3 NH+ Z− → (CH3 CH2 CH2 )2 NH+ Z− + C3 H6 , (623–673 K)

(CH3 CH2 CH2 )2 NH+ − + −
2 Z → (CH3 CH2 CH2 )2 NH3 Z + C3 H6 ,
(CH3 CH2 CH2 )NH+ − + −
3 Z → NH3 + C3 H6 + H Z .
A similar mechanism has been reported by Nowotny et al. (1991) and Park et al.
(1984) who investigated the decomposition of tetrapropylammonium cations inside
the channels of ZSM-5 crystals. The decomposition of the (CH3 CH2 CH2 )3 NH+ Z−
composites leads to a successive release of C3 H6 molecules. These C3 H6 molecules
are pyrolyzed into smaller molecules (such as CH4 and C2 H4 ) in the channels. This
was evidenced by the signals at m/z = 28, 27, 26 attributable to C2 H4 ; and the
signals at m/z = 16, 15 attributable to CH4 . These small hydrocarbon molecules
were subsequently carbonized. As an evidence, the intensity of H2 at m/z = 2 was
increased. All signals became weaker and finally undetectable when the sample was
heated to 823 K, indicating a complete carbonization of hydrocarbon precursor.
The carbonization process of the TPA molecules inside the channels of CoAPO-5
crystals was further analyzed by micro-Raman spectroscopy. Raman spectra of the
TPA@CoAPO-5 crystals measured at various temperatures from 373 to 773 K in
a vacuum of 10−3 mbar are shown in Fig. 6.5. When the CoAPO-5 crystals were
treated at a temperature below 573 K, the spectrum remained essentially the same
as that at 373 K, except that the intensities decreased. It shows typical characteris-
tic Raman-active modes of TPA molecules: The peaks near 2850–3050 cm−1 are
assigned to the CH3 and CH2 symmetric and anti-symmetric stretching modes, the
peak at about 1460 cm−1 is due to the CH3 anti-symmetrical deformation mode
and that near 967 ∼ 1150 cm−1 is attributed to C − N stretching mode (During,
1989). When the sample was pyrolyzed at temperature 623 K for 3 hours, all sig-
nals related to TPA molecules disappeared from the spectrum, indicating that the
carbon precursor TPA molecules start to decompose at this temperature. At even
higher temperature, the G-band at 1600 cm−1 appeared when the sample was heated
at temperature above 648 K, implying that carbon precursor TPA molecules was
carbonized inside the channels of CoAPO-5 single crystal. With still increasing
temperature, the relative intensity of G-band gradually strengthened. The sharp-
ness of the peaks is an indicator that carbon atoms are highly ordered in the chan-
nels. While the radial breathing mode (RBM) frequencies at 510 and 550 cm−1
(Li et al., 2004a) and the disorder-induced D line appear at temperature up to
Fig. 6.5 Raman spectra of

the TPA contained inside the
channels of CoAPO-5 crystal
were measured at various
temperatures in a vacuum of
10−3 mbar
673 K, an indicative of forming the 0.4 nm carbon nanotubes at the temperature.

As a result, the crystals turned homogeneous black with strong optical anisotropy
(Li et al., 2001). The relative intensity of RBM is increased with the increasing
temperature. All these results indicate that the 0.4 nm SWNTs inside the channels
of CoAPO-5 crystal started to be formed from small graphite sheet. Compared
with that of AlPO4 -5 single crystal, the temperature of TPA carbonization inside
the channels of CoAPO-5 single crystal is notably decreased (Sun et al., 1999).
It should be pointed out that there also exist Raman signals for the CoAPO-5 lat-
tice at frequencies of 450 cm−1 , due to the P-O-Al vibration (Holmes et al., 1994).
However, the Raman intensity of SWNTs is at least 20 times stronger than that
of CoAPO-5, owing to the resonant behavior. Thus the contribution of the vibra-
tions from the CoAPO-5 framework to the Raman spectra is negligible (Zhai et al.,
2006c).
6.4 Conclusions
In summary, the identification of the volatiles and the characterization of the inter-
mediate materials during pyrolysis of TPA carbon precursor molecules trapped in
as-grown CoAPO-5 crystals have been studied. The present data supported the
hypothesis that the TPA exists as tripropylammonium ions in crystals channels
and latter decomposes into dipropyl- and propylammonium cations by sequential
abstraction of propylene, in the temperature range 423–648 K. The result of micro-
Raman showed that the 0.4 nm nanotubes were formed at temperature about 673 K.
Compared with that of AlPO4 -5 single crystal, the temperature of TPA carbonization
inside the channels of CoAPO-5 single crystal is notably decreased because of the
catalyst effect of Co-ions in the AFI framework.
Acknowledgments This research was supported by Hong Kong CERGC Grants of 605003,
602807, RGC DAG04/05.SC24, RGC Central Allocation CA04/05.SC02, and HKUST President
direct allocation F0204-A.
References
During, J. R., Beshir, W. B., Godbey, S. E., Hizer, T. J., 1989, Raman and infrared spectra, confor-
mational stability and Ab initio calculations for n-propylamine. J. Raman. Spectrosc, 20, pp.
311–333.
Holmes, A. J., Kirby, S. J., Ozin, G. A., Young, D., 1994, Raman spectra of the unidimensional
aluminophosphate molecular sieves AlPO4 -11, AlPO4 -5, AlPO4 -8, and VPI-5. J. Phys. Chem,
98, pp. 4677–4682.
Li, Z. M., Tang, Z. K., Liu, H. J., Wang, N., Chan, C. T., Saito, R., 2001, Polarized adsorption
spectra of single-walled carbon nanotubes aligned in channels of AlPO4 -5 single crystals. Phys.
Rev. Lett, 87, pp. 127401.1–12701.4.
Li, Z. M., Tang, Z. K., Siu, G. G., Bozovic, I., 2004a, Raman characterization of 0.4 nm single-wall
carbon nanotubes using the full-symmetry line group. Appl. Phys.Lett, 84, pp. 4101–4103.
Li, Z. M., Zhai, J. P., Liu, H. J., Li, I. L., Chan, C. T., Sheng, P., Tang, Z. K., 2004b, Synthesis of
4 Å single-walled carbon nanotubes in catalytic Si-substituted AlPO4 -5 molecular sieves. Appl.
Phys. Lett, 85, pp. 1253–1255.
Mei, Y. F., Siu, G. G., Li, Z. M., Zhai, J. P., Liu, H. J., Tang, Z. K., Lai, C. W., Ong, H. C., 2005,
Visible cathodoluminescence of 4 Å single-walled carbon nanotubes. Appl. Phys.Lett, 87, pp.
213114.1–212114.3.
Nowotny, M., Lercher, J. A., Kessler, H., 1991, I.R. spectroscopy of single zeolite crystals. Part 1:
Thermal decomposition of the template in MFI-type materials. Zeolites, 11, pp. 454–459.
Parker, L. M., Bibby, D. M., Patterson, J. E., 1984, Thermal decomposition of ZSM-5 and silicalite
precursors. Zeolites, 4, pp. 168–174.
Qiu, S., Pang, W., Kessler, H., Guth, J. L., 1989, Synthesis and structure of the [AlPO4 ]12 Pr4 NH
molecular sieve with AFI structure. Zeolites, 9, pp. 440–444.
Soulard, M., Bilger, S., Kessler, H., Guth, J. L., 1987, Thermoanalytical characterization of MFI-
type zeolites prepared either in the presence of OH- or of F- ions. Zeolites, 7, pp. 463–470.
Sun, H. D., Tang, Z. K., Chen, J., Li, G., 1999, Synthesis and Raman characterization of mono-
sized single-wall carbon nanotubes in one-dimensional channels of AlPO4 -5 crystal. APPl.
Phys. A, 69, pp. 381–384.
Tang, Z. K., Sun, H. D., Wang, J., Chen, J., Li, G., 1998, Mono-sized single-wall carbon nanotubes
formed in channels of AlPO4 -5 single crystal. Appl. Phys. Lett, 73, pp. 2287–2289.
Tang, Z. K., Zhang, L. Y., Wang, N., Zhang, X. X., Wen, G. H., Li, G. D., 2001, Superconductivity
in 4 Å diameter single wall nanotubes. Science, 292, pp. 2462–2465.
Wang, N., Tang, Z. K., Li, G. D., Chen, J. S., 2000, Single-walled 4 Å carbon nanotube arrays.
Nature, 408, pp. 50–51.
Zhai, J. P., Li, Z. M., Liu, H. J., Li, I. L., Sheng, P., Hu, X., Tang, Z. K., 2006a, Catalytic effect
of metal cations on the formation of Carbon nanotubes inside the channels of AlPO4 -5 crystal.
Carbon, 44, pp. 1151–1157.
Zhai, J. P., Tang, Z. K., Lam, F. L. Y., Hu, X., 2006b, Thermal decomposition of carbon precursors
in decorated AFI zeolite crystals. J. Phys. Chem. B, 110, pp. 19285–19290.
Zhai, J. P., Tang, Z. K., Li, Z. M., Li, I. L., Jiang, F. Y., Sheng, P., Hu, X., 2006c, Carbonization
mechanism of Tetrapropylammonium-hydroxide in channels of AlPO4 -5 single crystals. Chem.
Mater, 18, pp. 1505–1511.
Part II
Quantum Dots and Molecular Spintronics
Chapter 7
Nanodevices and Maxwell’s Demon
Supriyo Datta∗
7.1 Introduction
Maxwell invented his famous demon to illustrate the subtleties of the second law
of thermodynamics and his conjecture has inspired much discussion ever since (see
for example, Leff and Rex, 1990, 2003; Nikulov and Sheehan, 2004 and references
therein). When the subject of thermodynamics was relatively new, it was not clear
that heat was a form of energy since heat could never be converted entirely into
useful work. Indeed the second law asserts that none of it can be converted to work
if it is all available at a single temperature. Heat engines can only function by oper-
ating between two reservoirs at two different temperatures. Maxwell’s demon is
supposed to get around this fundamental principle by creating a temperature dif-
ferential between two sides of a reservoir that is initially at a uniform temperature.
This is achieved by opening and closing a little door separating the two sides at just
the right times to allow fast molecules (white) to cross to the left but not the slow
molecules (black, Fig. 7.1a). As a result, faster molecules crowd onto the left side
making its temperature higher than that of the right side.
Technology has now reached a point where one can think of building an elec-
tronic Maxwell’s demon that can be interposed between the two contacts (labelled
source and drain) of a nanoscale conductor (Fig. 7.1b) to allow electrons to flow
preferentially in one direction so that a current will flow in the external circuit even
without any external source of power. Such a device can be built (indeed one could
argue has already been built) though not surprisingly it is expected to operate in
conformity with the second law. Here I would like to use this device simply to
illustrate the fundamental role of “contacts” and “demons” in transport and energy
conversion. I will try to keep the discussion at an academic level steering clear of real
world details. But it should be noted that the illustrative devices we will discuss are
very much within the capabilities of present-day technology. For example, recent
pioneering experiments on thermoelectric effects in molecular conductors (Reddy
et al., 2007) seem to agree well with the predictions from our model (Paulsson and
∗ School of Electrical and Computer Engineering, Purdue University, West Lafayette, IN
47907, USA

C Springer 2008
60 S. Datta
Datta, 2003). The device discussed in Section 7.3 (Figs. 7.5 and 7.6) incorporating
an elastic Maxwell’s demon (Datta, 2005b,c) is being investigated experimentally
by introducing manganese impurities into a GaAs spin-valve device with MnAs
contacts (Saha et al., 2006). The device incorporating an inelastic demon described
in Section 7.4 (Figs. 7.9 and 7.10) is very similar to the pentalayer spin-torque
device which has been studied by a number of groups both experimentally (Fuchs
et al., 2006; Huai et al., 2005) and theoretically (Salahuddin and Datta, 2006a,b)
though we are not aware of any discussion of the possibility of using the device
as a nanoscale heat engine or as a refrigerator as proposed here. We leave it to
future work to assess whether these possibilities are of any practical importance.
Here my objective is to lay out a simple transparent model showing how out-of-
equilibrium demons suitably incorporated into nanodevices can achieve energy con-
version.
In the last 20 years there has been significant progress in our ability to tackle the
problem of quantum transport far from equilibrium and a conceptual framework has
emerged through a combination of the Landauer approach with the non-equilibrium
Green function (NEGF) method (Datta, 1989, 1990; Meir and Wingreen, 1992),
which is being widely used in the analysis and design of nanoscale devices (see
Datta, 2005a and references therein). It provides a unified description for all kinds
of devices from molecular conductors to carbon nanotubes to silicon transistors
in terms of the Hamiltonian [H] describing the channel, the self-energies [1,2 ]
describing the connection to the contacts and the [s ] describing interactions inside
the channel (Fig. 7.3b). In each case the details are very different, but once these
matrices (whose size depends on the number of basis functions needed to describe
the channel) have been written down, the procedure for obtaining quantities of
interest such as current flow and energy dissipation are the same regardless of the
specifics of the problem at hand. The model covers different transport regimes from
the ballistic to the diffusive limit depending on the relative magnitudes of 1,2 and
s . In this paper I will use a particularly simple version of this approach (Fig. 7.3a)
where matrices like 1,2 and s are replaced with numbers like γ1,2 and γs hav-
ing simple physical interpretations: γ1,2/ represents the rate of transfer of channel
electrons in and out of the contacts while γs / represents the rate at which they
interact with any “demons” that inhabit the channel.
In the past it was common to have γ1,2 << γs so that transport was domi-
nated by the interactions within the channel, with contacts playing a minor enough
role that theorists seldom drew them prior to 1990! By contrast, today’s nanode-
vices have reached a point where γ1,2 >> γs , placing them in what we could
call the ballistic or “Landauer limit”. An appealing feature of this limit is that
it physically separates the dynamics from the dissipation: reversible dynamics
dominates the channel while dissipation dominates the contact. Usually these two
aspects of transport are conceptually so intertwined that it is difficult to see how irre-
versibility is introduced into a problem described by reversible dynamic equations
(Newton or Schrodinger) and this issue continues to spark debate and discussion
ever since the path-breaking work of Boltzmann many years ago (see for example,
McQuarrie, 1976).
7 Nanodevices and Maxwell’s Demon 61
Let me start with a brief summary of the basic framework shown in Fig. 7.3a that
we call the bottom-up viewpoint (Section 7.2). We will then use this approach to
discuss a specially designed device which is in the Landauer limit except for a partic-
ularly simple version of Maxwell’s demon that interacts with the channel electrons
but does not exchange any energy with them (Section 7.3). We then consider a more
sophisticated demon that exchanges energy as well and show how it can be used
to build nanoscale heat engines and refrigerators (Section 7.4). We conclude with
a few words about entangled demons and related conceptual issues that I believe
need to be clarified in order to take transport physics to its next level (Section 7.5).
Readers may find the related video lectures posted on the nano HUB useful (Datta,
2006) and I will be happy to share the MATLAB codes used to generate the figures
in this paper.
7.2 Bottom-up Viewpoint
Consider the device shown in Fig. 7.1b without the “demon” but with a voltage V
applied across two contacts (labelled “source” and “drain”) made to a conductor
(“channel”). How do we calculate the current I , as the length of the channel L is
made shorter and shorter, down to a few atoms? This is not just an academic question
since experimentalists are actually making current measurements through “chan-
nels” that are only a few atoms long. Indeed, this is also a question of great interest
from an applied point of view, since every laptop computer contains about one bil-
lion transistors, each of which is basically a conductor like the one in Fig. 7.1b with
L ∼ 50 nm, but with the demon replaced by a third terminal that can be used to
control the resistance of the channel.
As the channel length L is reduced from macroscopic dimensions (∼ millime-
ters) to atomic dimensions (∼ nanometers), the nature of electron transport that
is, current flow, changes significantly (Fig. 7.2). At one end, it is described by a
diffusion equation in which electrons are viewed as particles that are repeatedly
scattered by various obstacles causing them to perform a “random walk” from the
source to the drain. The resistance obeys Ohm’s law: a sample twice as long has
twice the resistance. At the other end, there is the regime of ballistic transport where
the resistance of a sample can be independent of length. Indeed due to wavelike
interference effects it is even possible for a longer sample to have a lower resis-
tance!
The subject of current flow is commonly approached using a “top-down” view-
point. Students start in high school from the macroscopic limit (large L) and seldom
reach the atomic limit, except late in graduate school if at all. I believe that this
is primarily for historical reasons. After all, 20 years ago, no one knew what the
resistance was for an atomic scale conductor, or if it even made sense to ask about
its resistance But now that the bottom-line is known, I believe that a “bottom-up”
approach is needed if only because nanoscale devices like the ones I want to talk
about look too complicated from the “top-down” viewpoint.
62 S. Datta
Fig. 7.1 (a) Maxwell’s demon opens and closes a trapdoor to separate fast (white) molecules
from slow (black) molecules making the left warmer than the right, thus creating a temperature
differential without expending energy. (b) Electronic Maxwell’s demon discussed in this talk lets
electrons go preferentially from right to left thus creating a current in the external circuit without
any external source of energy
In the top-down view we start by learning that the conductance G = I /V

is related to a material property called conductivity σ through a relation that
depends on the sample geometry and for a rectangular conductor of cross-section
A and length L is given by G = σ A/L. We then learn that the conductivity is
given by
σ = e2 nτ/m
where e is the electronic charge, n is the electron density, τ is the mean free time
and m is the electron mass. Unfortunately from this point of view it is very diffi-
cult to understand the ballistic limit. Since electrons get from one contact to the
other without scattering it is not clear what the mean free time τ is. Neither is
Fig. 7.2 Schematic representing the general approach used to model nanoscale devices: (a) Simple
version with numbers γ, D used in this talk and (b) Complete version with matrices , H (adapted
from Datta, 2005a)
it clear what one should use for ‘n’ since it stands for the density of free elec-
trons and with molecular scale conductors it may not be clear which electrons are
free. Even the mass is not very clear since the effective mass is deduced from the
bandstructure of an infinite periodic solid and cannot be defined for really small
conductors.
7.2.1 Conductance Formula: The “bottom-up” Version
A more transparent approach at least for small conductors is provided by the bottom-
up viewpoint (Datta, 2005a, Chapter 1) which leads to an expression for conduc-
tance in terms of two basic factors, namely the density of states D around the equi-
librium electrochemical potential and the effective escape rate γ /h from the channel
into the contacts ( = h/2π, h being Planck’s constant):
G = (e2 / h)2π Dγ (7.1a)
The escape rate appearing above is the series combination of the escape rates
associated with each contact:
γ = γ1 γ2 /γ1 + γ2 (7.1b)
This is an expression that we can apply to the smallest of conductors, even a

hydrogen molecule. Although it looks very different from the expression for con-
ductivity mentioned earlier, it is closer in spirit to another well-known expression
for the conductivity
σ = e2 Ñ D̃ (7.2)
in terms of the density of states per unit volume Ñ and the diffusion coefficient D̃.
Indeed we could obtain Eq. (7.1a) from Eq. (7.2) if we make the replacements
D → Ñ .AL and γ / → D̃/L 2 (7.3)
which look reasonable since the density of states for a large conductor is expected
to be proportional to the volume AL and the time taken to escape from a diffusive
channel is ∼ D̃/L 2 .
7.2.2 Current-Voltage Relation: Without Demons
The result cited above (Eq. (7.1a)) is a linear response version of a more general
set of equations that can be used to calculate the full current-voltage characteristics,
which in turn follow from the NEGF-Landauer formulation (Fig. 7.3b). For our
64 S. Datta
Fig. 7.3 Evolution of devices from the regime of diffusive transport to ballistic transport as the
channel length L is scaled down from millimetres to nanometres
purpose in this talk the simpler version (Fig. 7.3a) will be adequate and in this
version the basic equations are fairly intuitive:
i 1 (E) = (e/h) γ1 D (E) [ f 1 (E) − f (E)]

i 2 (E) = (e/h) γ2 D (E) f (E) − f 2 (E) (7.4)
These equations relate the currents per unit energy at contacts 1 and 2 to the
density of states D(E), the Fermi functions at the two contacts related to their elec-
trochemical potentials
1
f 1,2 (E) = (7.5)
1 + exp((E − μ1,2 )/k B T1,2 )
and the distribution function f (E) inside the channel.

The total currents at the source and drain contacts are obtained by integrating the
corresponding energy resolved currents
I1 = ∫ d Ei 1 (E) , I2 = ∫ d Ei 2 (E) (7.6)
If the electrons in the channel do not interact with any demons, we can simply
set i 1 (E) = i 2 (E), calculate f (E) and substitute back into Eq. (7.4) to obtain
i 1 (E) = i 2 (E)
γ1 γ2
= (e/) D(E) [ f 1 (E) − f 2 (E)] (7.7)
γ1 + γ2
The conductance expression stated earlier (Eq. (7.1a)) follows from Eqs. (7.6)
and (7.7) by noting that

d E [ f 1 (E) − f 2 (E)] = μ1 − μ2 (7.8)
and assuming the density of states D(E) to be nearly constant over the energy range
of transport where f1 (E) − f 2 (E) is significantly different from zero.
One-level conductor: Equation (7.7) can be used more generally even when the den-
sity of states has sharp structures in energy. For example, for a very small conductor
with just one energy level in the energy range of interest, the density of states is a
“delta function” that is infinitely large at a particular energy. Equations (7.6 and 7.7)
then yield a current-voltage characteristic as sketched below.
The maximum current is equal to eγ1 /2, assuming γ2 = γ1 . It might appear that
the maximum conductance can be infinitely large, since the voltage scale over which
the current rises is ∼ k B T , so that dI /dV can increase without limit as the temper-
ature tends to zero. However, the uncertainty principle requires that the escape rate
of γ /h into the contacts from an energy level also broadens the level by γ as
shown below. This means that the voltage scale over which the current rises is at
least ∼ (γ1 + γ2 )/e = 2γ1/e, even at zero temperature.
66 S. Datta
This means that a small device has a maximum conductance of
dI eγ1 /2 e2
∼ =
dV 2γ1 4
This is not too far from the correct result
G max = e2 / h ≈ 25.8 K (7.9)
which is one of the seminal results of mesoscopic physics that was not known
before 1988. One could view this as a consequence of the energy broadening
required by the uncertainty principle which comes out automatically in the full
Landauer-NEGF approach (Fig. 7.3b), but has to be inserted by hand into the sim-
pler version we are using where the density of states D(E) is an input parameter
(Fig. 7.3a).
For our examples in this paper we will use a density of states that is constant in
the energy range of interest, for which the current-voltage characteristic is basically
linear.
7.2.3 Current-Voltage Relation: With Demons
Defining i s (E) as the “scattering current” induced by interaction of the electrons

with the “demon”, we can write
i 1 (E) = i 2 (E) + i s (E) (7.10)

This current can be modelled in general as a difference between two processes

one involving a loss of energy ε from the demon and the other involving a gain of
energy by the demon.

i s (E) = (e/) γs D(E) dε D(E + ε)
[F(ε) f (E) (1 − f (E + ε)
− F(−ε) f (E + ε) (1 − f (E)] (7.11)
The basic principle of equilibrium statistical mechanics requires that if the demon
is in equilibrium at some temperature TD then the strength F(ε) of the energy loss
processes is related to the strength F(−ε) of energy gain processes by the ratio:
F(ε) = F(−ε) exp (−ε/k B TD ) (7.12)
With a little algebra one can show that this relation ensures that if the electron
distribution f (E) is given by an equilibrium Fermi function with the same temper-
ature T the two terms in Eq. (7.11) will cancel out. This result is independent of the
detailed shape of the function F(ε) describing the spectrum of the demon, as long as
Eq. (7.9) is true. This means that if the demon is in equilibrium with the electrons
with the same temperature, there can be no net flow of energy either to or from
the demon. Indeed one could view this as the basic principle of equilibrium statisti-
cal mechanics and work backwards to obtain Eq. (7.12) as the condition needed to
ensure compliance with this principle.
To summarize, if the electrons in the channel do not interact with any “demons”,
the current voltage characteristics are obtained from Eq. (7.7) using the Fermi func-
tions from Eq. (7.5). For the more interesting case with interactions, we solve for
the distribution f (E) inside the channel from Eqs. (7.4), (7.10) and (7.11) and then
calculate the currents. Usually the current flow is driven by an external voltage that
separates the electrochemical potentials μ1 and μ2 in Eq. (7.7). But thermoelectric
currents driven by a difference in temperatures T1 and T2 can also be calculated
from this model (Paulsson and Datta, 2003) as mentioned in the introduction. Our
focus here is on a different possibility for energy conversion, namely through out-
of-equilibrium demons.
7.2.4 Where is the Heat?
Before we move on, let me say a few words about an important conceptual issue that
caused much argument in the early days: Where is the heat dissipated in a ballistic
conductor? After all, if there is no demon to take up the energy, there cannot be any
dissipation inside the channel. The answer is that the transiting electron appears as
a hot electron in the drain (right) contact and leaves behind a hot hole in the source
68 S. Datta
(left) contact (see Fig. 7.4). The contacts are immediately restored to their equilib-
rium states by unspecified dissipative processes operative within each contact. These
processes can be quite complicated but are usually incorporated surreptitiously into
the model through what appears to be an innocent boundary condition, namely that
the electrons in the contacts are always maintained in thermal equilibrium described
by Fermi distributions (Eq. (7.5)) with electrochemical potentials μ1 and μ2 and
temperatures T1 and T2 .
To understand the spatial distribution of the dissipated energy it is useful to look
at the energy current which is obtained by replacing the charge ‘e’ with the energy
E of the electron:

I E1 = (1/e) d E E i 1 (E), I E2 = (1/e) d E E i 2 (E) (7.13)
The energy currents at the source and drain contacts are written simply as
I Es (E) = (μ1 /e) I1 , I Ed (E) = (μ2 /e) I2 (7.14)
assuming that the entire current flows around the corresponding electrochemical
potentials.
Figure 7.4 shows a spatial plot of the energy current from the source end
to the drain end for a uniform 1-D ballistic conductor with a voltage of 50 mV
Fig. 7.4 Plot of energy current in a 1-D ballistic conductor with G = e2 / h ≈ 40 μS and an applied
voltage of V = 0.05 volts. Energy dissipated is given by the drop in the energy current, showing that
the Joule heating V 2 G = 0.1 μW is divided equally between the two contact-channel interfaces
applied across it. For a conductor with no demon for electrons to exchange energy
with, i 1 (E) = i 2 (E) making the energy current uniform across the entire chan-
nel implying that no energy is dissipated inside the channel. But the energy cur-
rent entering the source contact is larger than this value while that leaving the
drain contact is lower. Wherever the energy current drops, it means that the rate
at which energy flows in is greater than the rate at which it flows out, indicat-
ing a net energy dissipation. Clearly in this example, 0.05 μW is dissipated in
each of the two contacts thus accounting for the expected Joule heating given by
V 2 G = 0.1 μW.
Real conductors have distributed demons throughout the channel so that dissipa-
tion occurs not just in the contacts but in the channel as well. Indeed we commonly
assume the Joule heating to occur uniformly across a conductor. But there are now
experimental examples of nanoscale conductors that would have been destroyed if
all the heat were dissipated internally and it is believed that the conductors survive
only because most of the heat is dissipated in the contacts which are large enough
to get rid of it efficiently. The idealized model depicted in Fig. 7.4 thus may not be
too far from real nanodevices of today.
As I mentioned in the introduction, what distinguishes the Landauer model
(Fig. 7.4) is the physical separation of dynamics and dissipation clearly showing
that what makes transport an irreversible process is the continual restoration of
the contacts back to equilibrium. Without this repeated restoration, all flow would
cease once a sufficient number of electrons have transferred from the source to
the drain. Over a 100 years ago Boltzmann showed how pure Newtonian dynam-
ics could be supplemented to describe transport processes, and his celebrated
equation stands today as the centerpiece for the flow of all kinds of particles.
Boltzmann’s approach too involved “repeated restoration” through an assumption
referred to as the “Stohsslansatz” (see for example, McQuarrie, 1976) Today’s
devices often involve Schrodinger dynamics in place of Newtonian dynamics
and the non-equilibrium Green function (NEGF) method that we use (Fig. 7.3b)
supplements the Schrodinger equation with similar assumptions about the repeated
restoration of the surroundings that enter the evaluation of the various self-energy
functions or the corresponding quantities γ in the simpler model that we
are using.
“Contacts” and “demons” are an integral part of all devices, the most common
demon being the phonon bath for which the relation in Eq. (7.12) is satisfied by
requiring that energy loss ∼ N and energy gain ∼ (N + 1), N being the number
of phonons given by the Bose Einstein factor if the bath is maintained in equilib-
rium. Typically such demons add channels for dissipation, but our purpose here is
to show how suitably engineered out-of-equilibrium demons can act as sources of
energy.
For this purpose, it is convenient to study a device specially designed to
accentuate the impact of the demon. Usually the interactions with the demon
do not have any clear distinctive effect on the terminal current that can be easily
detected. But in this special device, ideally no current flows unless the channel
electrons interact with the demon. Let me now describe how such a device can be
engineered.
70 S. Datta
7.3 Elastic Demon
Let us start with a simple 1-D ballistic device but having two rather special kinds of
contacts. The source contact allows one type of spin, say “black” (drawn pointing
to the right), to go in and out of the channel much more easily than the other type,
say “white” (drawn pointing to the left, Fig. 7.5a). Devices like this are called spin-
valves and are widely used to detect magnetic fields in magnetic memory devices
(see for example the articles in Heinreich and Bland, 2004).
Although today’s spin-valves operate with contacts that are far from perfect,
since we are only trying to make a conceptual point, let us simplify things by
assuming contacts that are perfect in their discrimination between the two spins.
One only lets black spins while the other only lets white spins to go in and out
of the channel. We then have the situation shown in Fig. 7.5b and no current can
flow since neither black nor white spins communicate with both contacts. But if we
introduce impurities into the channel (the demon) with which electrons can interact
and flip their spin, then current flow should be possible as indicated: black spins
come in from the left, interact with impurities to flip into white spins and go out the
right contact (Fig. 7.5c).
Consider now what happens if the impurities are say all white (Fig. 7.6a). Elec-
trons can now flow only when the bias is such that the source injects and drain
collects (positive drain voltage), but not if the drain injects and the source collects
(negative drain voltage). This is because the source injects black spins which interact
with the white impurities, flip into whites and exit through the drain. But the drain
only injects white spins which cannot interact with the white impurities and cannot
cross over into the source. Similarly if the impurities are all black, current flows
Fig. 7.5 Anti-parallel (AP) Spin-valve: (a) Physical structure, (b) no current flows if the contacts
can discriminate between the two spins perfectly, (c) current flow is possible if impurities are
present to induce spin-flip processes
Fig. 7.6 Perfect AP Spin-valve with (a) white (down spin, drawn as pointing to the left) impurities
and (b) black (up spin, drawn as pointing to the right) impurities
only for negative drain voltage (Fig. 7.6b). At non-zero temperatures the cusps in
the current-voltage characteristics get smoothed out and we get the smoother curves
shown in Fig. 7.6 a, b.
Note the surprising feature of the plots at T = 300 K: there is a non-zero cur-
rent even at zero voltage! This I believe is correct. Devices like those shown in
Fig. 7.6 a, b with polarized impurities could indeed be used to generate power
and one could view the system of impurities as a Maxwell’s demon that lets elec-
trons go preferentially from source to drain or from drain to source. The second
law is in no danger, since the energy is extracted at the expense of the entropy
of the system of impurities that collectively constitute the demon. Assuming the
spins are all non-interacting and it takes no energy to flip one, the polarized state
of the demon represents an unnatural low entropy state. Every time an electron
goes through, a spin gets flipped raising the entropy and the flow will eventually
stop when demon has been restored to its natural unpolarized state with the highest
entropy of N I k B ln 2, where N I is the number of impurities. To perpetuate the flow
an external source will have to spend the energy needed to maintain the demon in
its low entropy state.
7.3.1 Current Versus Voltage: Summary of Equations
Let me now summarize the equations that I am using to analyze structures like the
one in Fig. 7.6 quantitatively. Basically it is an extension of the equations described
72 S. Datta
earlier (see Eq. (7.6), Section 7.2) to include the two spin channels denoted by the
subscripts “u” (up or black) and “d” (down or white):
i 1,u (E) = (e/)γ1,u Du (E) [ f 1 (E) − fu (E)]

i 1,d (E) = (e/)γ1,d Dd (E) [ f 1 (E) − f d (E)] (7.15)
i 2,u (E) = (e/)γ2,u Du (E) [ f u (E) − f 2 (E)]
i 2,d (E) = (e/) γ2,d Dd (E) [ f d (E) − f 2 (E)] (7.16)
For the scattering current caused by the demon we write (see Eq. (7.11))
i s,u (E) = − i s,d (E) = (e/) (2π J 2 N I ) Du (E) Dd (E)

[Fd f u (E) (1 − f d (E) − Fu fd (E) (1 − f u (E)] (7.17)
noting that we are considering an elastic demon that can neither absorb nor give
up energy (ε = 0). The first term within parenthesis in Eq. (7.17) represents an
up electron flipping down by interacting with a down-impurity while the second
term represents a down electron flipping up by interacting with an up-impurity.
The strengths of the two processes are proportional to the fractions Fd and Fu of
down and up-spin impurities. The overall strength of the interaction is governed by
the number of impurities N I and the square of the matrix element J governing the
electron-impurity interaction.
Equation (7.16) can be visualized in terms of an equivalent circuit (Fig. 7.7)
if we think of the various f ’s as “voltages” since the currents are proportional to
differences in ‘f’ just as we expect for conductances defined as
g1,u = (e2 /) γ1,u Du etc.
Fig. 7.7 An anti-parallel (AP) spin-valve with spin-flip impurities and a simple equivalent circuit
to help visualize the equations we use to describe it, namely Eqs. (7.15) through (7.17) (adapted
from Datta, 2005 b,c)
The scattering current (Eq. (7.17)) too could be represented with a conductance
gs if we set Fd = Fu = 0.5
i s = (e/)(π J 2 N I )Du (E)Dd (E)( f d − fu ) (7.18a)
However, this is true only if the impurities are in equilibrium, while the interest-
ing current-voltage characteristics shown in Fig. 7.6 require an out-of-equilibrium
demon with Fd = Fu . So we write the total scattering current from Eq. (7.17) as
a sum of two components, one given by Eq. (7.18a) and another proportional to
(Fd − Fu ) which we denote with a subscript ‘dem’:
i dem = (Fd − Fu )N I /τ0

1
= (e/) (π J 2 ) Du (E)Dd (E) [ f d (1 − f u ) + f u (1 − f d )] (7.18b)
τ0
Equation (7.16) and (7.17) can be solved to obtain the distribution functions
f u (E) and f d (E) by imposing the requirement of overall current conservation (cf.
Eq. (7.11)):
i 1,u (E) = i 2,u (E) + i s,u (E)

i 1,d (E) = i 2,d (E) + i s,d (E) (7.19)
The currents are then calculated and integrated over energy to obtain the terminal
currents shown in Fig. 7.6. We can also find the energy currents using equations like
Eqs. (7.14) and (7.15) and the results are shown in Fig. 7.8 for Fd − Fu = −1 and
for Fd − Fu = 0 each with a voltage difference of 2k B T = 50 mV between the two
terminals. With Fd − Fu = 0 the direction of current is in keeping with an external
battery driving the device. But with Fd − Fu = −1, the external current flows against
the terminal voltage indicating that the device is acting as a source of energy driving
a load as shown in the inset. This is also borne out by the energy current flow which
shows a step up at each interface indicating that energy is being absorbed from the
contacts (∼ 10 nW from each) and delivered to the external load.
But isn’t this exactly what the second law forbids? After all if we could just
absorb energy from our surroundings (the contacts) and do useful things, there
would be no energy problem. The reason this device is able to perform this impos-
sible feat is that the impurities are assumed to be held in a non-equilibrium state
with very low entropy. A collection of N I impurities can be unpolarized in 2 N I
different ways having an entropy of S = N I k B ln 2. But it can be completely
polarized (Fd − Fu = ±1) in only one way with an entropy of S = 0. What this
device does is to exchange entropy for energy. Many believe that the universe too is
evolving the same way, with constant total energy, from a particularly low entropy
state continually towards a higher entropy one. But that is a different matter.
To have our device continue delivering energy indefinitely we will need an exter-
nal source to maintain the impurities in their low entropy state which will cost
74 S. Datta
Fig. 7.8 Energy currents with Fd − Fu = −1 and with Fd − Fu = 0 each with a voltage difference
of 2k B T = 50 mV between the two terminals. With Fd − Fu = 0 the direction of current is in
keeping with an external battery driving the device. But with Fd − Fu = −1, the external current
flows against the terminal voltage indicating that the device is acting as a source of energy driving a
load as shown in the inset. This is also borne out by the energy current flow which shows a step up
at each interface indicating that energy is being absorbed from the contacts (∼ 10 nW from each)
and delivered to the external load
energy. The details will depend on the actual mechanism used for the purpose but we
will not go into this. Note that if we do not have such a mechanism, the current will
die out as the spins get unpolarized. This depolarization process can be described
by an equation of the form:

1 1
d PI /dt + + PI = i s (7.20)
τI τ0
where τ0 and i s are related to the scattering current as defined in Eqs. (7.18), while
the additional time constant τ I represents processes unrelated to the channel elec-
trons by which impurities can relax their spins.
7.4 Inelastic Demon
We have argued above that although one can extract energy from polarized impu-
rities, energy is needed to keep them in that state since their natural high entropy
state is the unpolarized one. It would seem that one way to keep the spins nat-
urally in a polarized state is to use a nanomagnet, a collection of spins driven
by a ferromagnetic interaction that keeps them all pointed in the same direction.
Could such a magnet remain polarized naturally and enable us to extract energy
from the contacts forever? The answer can be “yes” if the magnet is at a differ-
ent temperature from the electrons. What we then have is a heat engine working
between two reservoirs (the electrons and the magnet) at different temperatures and
we will show that its operation is in keeping with Carnot’s principle as required by
the second Law.
To model the interaction of the electrons with the nanomagnet we need to modify
the expression for the scattering current (Eq. (7.17)) for it now takes energy to flip a
spin. We can write

i s,u (E) = − i s,d (E) = (e/) (2π J N I )
2
dε Du (E) Dd (E + ε)
[F(ε) f u (E) (1 − f d (E + ε)) − F(−ε) f d (E + ε) (1 − f u (E)]

(7.21)
where F(ε) denotes the magnon spectrum obeying the general law stated ear-
lier (see Eq. (7.13)) if the magnet has a temperature TD . Equation (7.21) can
be solved along with Eqs. (7.15), (7.16) and (7.19) as before to obtain currents,
energy currents etc. But let us first try to get some insight using simple approxima-
tions.
If we assume that the electron distribution functions f u (E) and fd (E) are
described by Fermi functions with electrochemical potentials μu and μd respec-
tively and temperature T , and make use of Eq. (7.13) we can rewrite the scattering
current from Eq. (7.21) in the form
i s,u (E) = −i s,d (E) = (e/)(2π J 2 N1 ) ∫ dε Du (E)Dd (E + ε)

F(−ε) fd (E + ε) 1 − fu (E)
F(ε) f u (E)(1 − f d (E + ε) 1 −
F(+ε) 1 − f d (E + ε) f u (E)
= (e/)(2π J 2 N1 ) ∫ dε Du (E)Dd (E + ε)F(ε) fu (E)(1 − f d (E + ε)

ε μd − ε − μu
1 − exp + (7.22)
k B TD kB T
If we further assume the exponent to be small, so that 1 − exp(−x) ≈ x, we

can write

μu − μd ε ε
i s,u = −i s,d ∼ + − (7.23)
kB T kB T k B TD
76 S. Datta
The first term represents a dissipative current proportional to the potential

difference and can be represented by a conductance like the gs in Fig. 7.7, while
the second term isg the demon induced source term that can be harnessed to do
external work. It vanishes when the demon temperature TD equals the electron
temperature T .
To be specific, let us assume that the demon is cooler than the rest of the device
(TD < T ) so that
ε ε
− <0
kB T k B Td
giving rise to a flow of electrons out of the up spin node and back into the downspin
node. If we use this to drive an external load then μu − μd > 0, which will tend to
reduce the net current given by Eq. (7.23) and the maximum output voltage one can
get corresponds to “open circuit” conditions with zero current:
μu − μd ε ε
≤ − (7.24)
kB T k B TD kB T
This expression has a simple interpretation in terms of the Carnot principle.

Every time an electron flows around the circuit giving up energy ε to the demon,
it delivers energy μu − μd to the load. All this energy μu − μd + ε is absorbed from
the contacts and the Carnot principle requires that
Energy from contacts / T ≤ Energy to demon/TD

that is,
μu − μd + ε ε
≤
T TD
which is just a restatement of Eq. (7.24). Note that usual derivations of the Carnot
principle do not put the system simultaneously in contact with two reservoirs at
Fig. 7.9 Heat engine: (a) AP Spin-valve with a cooled magnet as Maxwell’s demon controlling the
flow of electrons. (b) Output current and output power versus output voltage as the load is varied
from short circuit (V = 0) to open circuit conditions (I = 0). The spectrum of the magnet is
assumed to consist of a single energy ε = 2k B T = 50 meV. (c) Energy current profile assuming a
load such that the output voltage is 50 mV
78 S. Datta
Fig. 7.10 (a) Heat engine from Fig. 7.9 operated as a refrigerator by applying an external battery
to inject downspin (white) electrons from the drain that flip down the thermally created up spins
in the magnet, thus cooling it. (b) Energy current profile showing that energy is absorbed from the
external battery and the demon and dissipated in the source and drain contacts
different temperatures as we have done. Our treatment is closer in spirit to the classic
discussion of the ratchet and pawl in the Feynman Lectures (Feynman et al., 1963).
Figure 7.9 shows numerical results obtained by solving Eqs. (7.21), (7.19), (7.15)
and (7.16). The extracted power is a maximum (Fig. 7.9d) when the output voltage is
somewhere halfway between zero and the maximum output voltage of 80 mV (a few
k B T ). Figure 7.9b shows the energy current profile at an output voltage of 50 mV:
energy is absorbed from the source and drain contacts and given up partly to the
demon and the rest is delivered to the external load. The efficiency (energy given to
load/energy absorbed from contact) is a maximum close to open circuit conditions,

but the energy delivered is very small at that point as evident from Fig. 7.9d.
As one might expect, one can also operate the same device as a refrigerator
by using an external source that seeks to inject downspins (white) from the drain
contact that flip back thermally created up spins in the magnet thus cooling it. It is
evident from the energy current profile shown in Fig. 7.10 that in this case, energy is
absorbed from the demon and from the battery and given up to the source and drain
contacts.
7.5 Entangled Demon
In this talk I have tried to introduce a simple transparent model showing how out-
of-equibrium demons suitably incorporated into nanodevices can achieve energy
conversion. At the same time this model illustrates the fundamental role played by
“contacts” and “demons” in these processes. I would like to end by pointing out
another aspect of contacts that I believe is important in taking us to our next level of
understanding. The basic point can be appreciated by considering a simple version
of the spin capacitor we started with (see Fig. 7.5) but having just one impurity
(Fig. 7.11). No current can flow in this structure without spin-flip processes since
the source injects black (up) electrons while the drain only collects white (down)
electrons. But if the black electron interacts with the white impurity (A) to produce
a black impurity then the white electron can be collected resulting in a flow of
current.
The process of conversion
From A: Black electron ⊗ White impurity
To B: White electron ⊗ Black impurity
is incorporated into our model through the scattering current (see Eq. (7.17)).
A more complete quantum transport model involving matrices (Fig. 7.2a) rather
Fig. 7.11 We have described spin-flip processes in terms of a direct conversion from state A to
state B. Quantum mechanics, however, requires an intermediate state consisting of a superposition
of A and B before wavefunction collapse reduces it to a B. This state may have a significant role in
devices with weak contacts and strong interactions in the channel, requiring a model that can deal
with entangled states
80 S. Datta
than numbers (Fig. 7.2b) could be used to describe this effect, but the essential
underlying assumption in either case is that the state of the electron-impurity sys-
tem changes from A to B upon interaction. Quantum mechanics, however, paints
a different picture of the process involving an intermediate entangled state. It says
that the system goes from A into a state consisting of a superposition of A and B and
it is only when the electron is collected by the drain that the wavefunction collapses
to a B. If the collection rates γ1,2 / are much larger than the interaction rate per
impurity γs /N I , we expect the entangled state to play a minor role. But this may
not be true of devices with weak contacts and strong interactions in the channel,
requiring a model that can deal with entangled states.
Entangled states are difficult to describe within the conceptual framework we
have been using where both the electrons and the impurity are assumed to exist in
independent states. It is hard to describe a “conditional state” where the electron
is black if the impurity is white or vice versa, let alone a superposition of the two.
To account for this entangled state we need to treat the electron and impurity as
one big system and write rate equations for it, in the spirit of the many-electron
rate equations widely used to treat Coulomb blockade but the standard approach
(Beenakker, 1991; Likharev, 1999) needs to be extended to include coherences
and broadening. This is an area of active research (see for example Braun et al.,
2004; Braig and Brouwer, 2005) where an adequate general approach does not
yet exist.
Actually correlated states (classical version of entanglement) were an issue even
before the advent of quantum mechanics. Boltzmann ignored them through his
assumption of “molecular chaos” or “Stohsslansatz”, and it is believed that it is
precisely this assumption that leads to irreversibility (see for example, McQuarrie,
1976). An intervening entangled or correlated state is characteristic of all “channel”-
“contact” interactions, classical or quantum, and the increase in entropy character-
istic of irreversible processes can be associated with the destruction or neglect of
the correlation and/or entanglement generated by the interaction (see for example,
Zhang, 2007, Datta, 2007). This aspect is largely ignored in today’s transport theory,
just as even the presence of a contact was barely acknowledged before the advent
of mesoscopic physics. But new experiments showing the effect of entanglements
on current flow are on the horizon and will hopefully lead us to the next level of
understanding.
Acknowledgments This work was supported by the Office of Naval Research under grant no.
N00014-06-1-0025.
References
Beenakker, C.W.J., 1991, Theory of Coulomb blockade oscillations in the conductance of a quan-
tum dot. In Physical Review, B44, 1646.
Braig, S. and Brouwer, P.W., 2005, Rate equations for Coulomb blockade with ferromagnetic leads.
In Physical Review, B71, 195324.
Braun, M., Koenig, J. and Martinek, J., 2004, Theory of transport through quantum-dot spin valves
in the weak coupling regime. In Physical Review, B70, 195345.
Datta, S., 1989, Steady-state quantum kinetic equation. In Physical Review, B40, 5830.
Datta, S., 1990, A simple kinetic equation for steady-state quantum transport. In Journal of Physics:
Condensed Matter, 2, 8023.
Datta, S., 2005a, Quantum transport: Atom to transistor (Cambridge University Press).
Datta, S., 2005b, Proposal for a spin capacitor. In Applied Physics Letters, 87, 013115.
Datta, S., 2005c, Spin dephasing and hot spins. In Proceedings of the International School of
Physics “Enrico Fermi” Course CLX edited by A. D’Amico, G. Balestrino and A. Paoletti
(IOS Press, Amsterdam and SIF Bologna).
Datta, S., 2006, Concepts of quantum transport, a series of video lectures, http://www.nanohub.org/
courses/cqt
Datta, S., 2007, Quantum Transport as a Process of Repeated Disentanglement, Preprint.
Feynman, R.P., Leighton, R.B. and Sands, M., 1963, Lectures on Physics, vol.1, Chapter 46
(Addison-Wesley, Boston).
Fuchs, G.D., Krivotorov, I.N., Braganca, P.M., Emley, N.C., Garcia, A.G.F., Ralph, D.C., Buhrman,
R.A., 2005, Adjustable spin torque in magnetic tunnel junctions with two fixed layers, In
Applied Physics Letters, 86, 152509.
Huai, Y., Pakala, M., Diao, Z., Ding, Y., 2005, Spin transfer switching current reduction in magnetic
tunnel junction based dual spin filter structures. In Applied Physics Letters, 87, 222510.
Heinreich, B. and Bland, J.A.C., (eds) 2004, Introduction to micromagnetics in ultrathin magnetic
structures, vol. IV (Springer-Verlag, Berlin).
Leff, H. S. and Rex, A.F., (eds) 1990, Maxwell’s demon, entropy, information and computing,
Princeton Series in Physics.
Leff, H. S. and Rex, A. F., (eds) 2003, Maxwell’s Demon 2: entropy, classical and quantum infor-
mation, computing (IOP Publishing, Bristol).
Likharev, K., 1999, Single electron devices and their applications. In Proceedings IEEE, 87, 606.
McQuarrie, D.A., 1976, Statistical Mechanics (Harper and Row).
Meir, Y. and Wingreen, N.S., 1992, Landauer formula for the current through an interacting elec-
tron region. In Physical Review Letters 68, 2512.
Nikulov, A. and Sheehan D. (eds.) 2004, Entropy, 6.
Paulsson, M. and Datta, S., 2003, Thermoelectric effect in molecular electronics. In Physical
Review, B67, 241403(R).
Reddy, P., Jang, S.Y., Segalman, R. and Majumdar, A., 2007, Thermoelectricity in molecular junc-
tions. In Science, 315, 1568.
Saha, D., Holub, M., Bhattacharya, P. and Liao, Y.C., 2006, Epitaxially grown MnAs/GaAs lateral
spin valves. In Applied Physics Letters, 89, 142504.
Salahuddin, S. and Datta, S., 2006a, Electrical detection of spin excitations. In Physical Review,
B73, 081301(R).
Salahuddin, S. and Datta, S., 2006b, Self-consistent simulation of quantum transport and magneti-
zation dynamics in spin-torque based devices. In Applied Physics Letters, 89, 153504.
Zhang, Qi-Ren 2007, A General information theoretical proof for the second law of thermodynam-
ics, arXiv: quant-ph/0610005v3.
Chapter 8
Manipulating Electron Spins
in an InGaAs/InAlAs Two-Dimensional
Electron Gas
C. L. Yang, X. D. Cui, S. Q. Shen, H. T. He, Lu Ding, J. N. Wang, F. C. Zhang

and W. K. Ge∗
8.1 Introduction
With the basic idea of the spin field effect transistor, as proposed by Datta and Das
(1990), the electric field tunable spin-orbit interaction has been an ideal candidate
for manipulating the electron spin-polarization. The structural inversion asymme-
try (SIA) induced spin-orbit coupling (Rashba interaction) (Bychkov and Rashba,
1984) in a two-dimensional electron gas (2DEG) system has attracted more and
more attention due to its potential applications in spintronics devices. A two-
dimensional electron gas (2DEG) in a semiconductor hetero-structure with C2v
symmetry and vertical structural inversion asymmetry (SIA) would induce a Rashba
spin-orbit coupling and lead to spin splitting of the conduction band in the momen-
tum space (Bychkov and Rashba, 1984; Winkler, 2003). This system provides a
good platform for electrical control of electron spins in semiconductors. Such exam-
ples include spin photocurrent (Ganichev et al., 2002), electric-dipole-induced spin
resonance (Duckheim and Loss, 2006), and spin coherent transport (Crooker et al.,
2005; Murakami et al., 2003; Sinova et al., 2004; Shen et al., 2004). With spin
splitting of the energy bands, optical excitation of quantum well (QW) structures by
circularly polarized radiation will lead to a current whose direction and magnitude
depend on the helicity of the incident light. This is named the circular photogal-
vanic effect (CPGE) and has been demonstrated using inter-subband (Ganichev
et al., 2001; Ganichev and Prettl 2003) or inter-band (Bel’kov et al., 2003) exci-
tations. A related effect, in which the current is driven by the spin-flip process of the
non-equilibrium population of electrons in the spin-split bands, is called the spin-
galvanic effect (SGE) (Ganichev et al., 2002). The most interesting consequence of
the k-dependent spin splitting is the implication that an applied electric field would
induce not only a charge current but also a spin polarization (Aronov et al., 1991;
∗ Department of Physics and Institute of Nano-Science and Technology, The Hong Kong
University of Science and Technology, Hong Kong, China


C Springer 2008
84 C. L. Yang et al.
Edelstein, 1990; Chaplik et al., 2002; Inoue et al., 2003; Kato et al., 2004; Silov
et al., 2004; Ganichev et al., 2004) along the direction perpendicular to the current.
This paper provides the experimental evidence of circularly polarized optical-
excitation-induced spin photocurrent in (001) grown InGaAs/InAlAs 2DEGs under
oblique incidence of radiation for inter-band excitation, which is 2 orders of
magnitude stronger than similar observations using far-infrared excitation for inter-
subband transitions. The theoretically predicted spectral inversion of the CPGE spin
photocurrent was demonstrated experimentally. Furthermore, we have measured
the current-induced spin-polarization in the same samples using Kerr rotation
experiment. The spin splitting of the energy bands was examined by the beating of
the Shubnikov–de Haas (SdH) oscillations, which reveals a unified picture for the
spin photocurrent, current-induced spin polarization and spin orbit coupling.
While optically injected spin current has been evidenced both in theory and
experiments (Ganichev et al., 2001, 2002; Ganichev and Prettl, 2003; Bhat et al.,
2005; Li et al., 2006), but detecting spin current, which is of great interest to spin-
tronics physics as well as device applications, remains to be a challenging problem
(Hübner et al., 2003; Stevens et al., 2003; Kato et al., 2004; Sih et al., 2005; Wunder-
lich et al., 2005; Saitoh et al., 2006). Coherent transport and control of electron spins
in semiconductor hetero-junctions have been studied intensively with the ultimate
aim of implementing spintronic devices (Ganichev and Prettl, 2003; Wolf et al.,
2001; Awschalom et al., 2002; Zutic et al., 2004). Among these efforts, optical injec-
tion of electron spin into 2DEG can be regarded as an efficient source to implement
spin transport. In the spin photogalvanic effect, electrons in the valence band are
pumped into the conduction band by irradiation of circularly polarized light, and
the nonuniform distribution of the excited electrons due to the spin-orbit coupling
will circulate a spin polarized electric current. In the case that the incident light
is normal to the 2D plane, electrons with opposite spins would travel in opposite
directions. As a result, a pure spin current circulates while the electric photo-current
vanishes (Ganichev et al., 2001; Bhat et al., 2005).
In this paper, we report an observation of onward or outward electric currents in
a crossbar-shaped (001) InGaAs/InAlAs quantum well when linearly polarized light
irradiates on the sample at normal incidence near the junction at 77 K. Though no
electric photo-current was measured when the incident light spot is far away from
the crossing region of the crossbar channels, which is consistent with the conven-
tional spin photogalvanic effect, electric currents were detected when the light spot
is close to the crossing region. The electric current was demonstrated to share the
same light-polarization dependence as the spin current. We attribute the observed
electric current to be converted from the optically injected spin current by scattering
in the crossing region.
8.2 Experimental and Results
Two samples (named D and E) studied here were Inx Ga1−x As/In0.52Al0.48 As
2DEGs grown on semi-insulating (001) InP substrate with well thickness of 14 nm.
The SIA was achieved by δ–doping of only one side of the barrier layer (on top
8 Manipulating Electron Spins in an InGaAs/InAlAs Two-Dimensional Electron Gas 85
Fig. 8.1 (a) and (b) are the beating patterns of the SdH oscillations at 1.6 K under low mag-
netic field as shown in the first derivative of the magneto-resistance Rxx of samples D and E,
respectively. (c) and (d) are the calculated carrier concentrations based on the FFT spectrum from
(a) and (b)
of the well). To enhance the SIA, sample E was grown with a graded indium com-
position from 0.53 to 0.75 for the quantum well, instead of the uniform indium
composition of 0.70 for sample D. Hall measurements at 1.6 K showed that the
carrier concentration was 1.5 × 1012 cm−2 and 1.4 × 1012 cm−2 for samples D
and E, respectively. The SdH oscillations as well as their pronounced beating
patterns at low magnetic field were indicated in Fig. 8.1a and b using the first
derivative of Rxx . The beating pattern arises from two closely separated oscil-
lation frequency components caused by the presence of two kinds of carriers in
the system. The beating leads to double-peak structures with similar height in the
fast Fourier transform (FFT) spectra and gives the density of the two kinds of
carriers as shown in Fig. 8.1c and d. The calculated carrier density is in excel-
lent agreement with the Hall concentration if a spin degeneracy of 1 is taken. In
our samples, the distinct beating pattern in the SdH oscillations are believed to
come from the spin splitting in 2DEGs, which is further supported by the fol-
lowing spin photocurrent and current-induced spin polarization experiment. The
obtained spin-resolved concentrations allow us to determine the Rashba spin-orbit
interaction parameter α (Engels et al., 1997; Schapers et al., 1998), which gives
the value of 3.0 × 10−12 eVm and 6.3 × 10−12 eVm for samples D and E, respec-
tively, indicating that the stronger the inversion asymmetry, the larger the spin-
orbit interaction in a system. The measured strongly structure-related spin-orbit
coupling coefficient suggests that the SIA induced spin-splitting is the dominating
mechanism.
The experimental setup for the interband-transition-induced CPGE is schemat-
ically shown in the inset of Fig. 8.2. A tunable Ti:sapphire femtosecond laser was
employed for the band-to-band excitation with its polarization modified by either
a crystalline λ 4 waveplate which yields right (σ − ) or left (σ + ) hand circularly
Fig. 8.2 (a) and (b) are the angle φ dependence of the spin photocurrent in the 2DEG sample
D and E, respectively, with oblique incidence angle of 300 and laser power of 100 mW (λ =
880 nm) at 10 K, showing the helicity dependence of the photocurrent. The solid lines are the fit
using sin 2φ. (c) is a schematic diagram with spin-split valence and conduction bands to show the
microscopic origin of the CPGE induced photocurrent. The vertical arrows show the permitted
transitions with σ − excitation and the horizontal arrow indicates the final direction of the induced
photocurrent. (d) is the incidence angle Θ dependence of the photocurrent. The solid line is the
fitted curve based on Eq. (8.1). The experimental setup of the spin photocurrent is depicted in
the inset
polarized light, or a PEM which yields a periodically oscillating polarization

between σ − and σ + . The electrodes are made at the [110] or [11̄0] sample edges to
lead the current along the laboratory x-direction. The photocurrent Ix was measured
in the unbiased structures at various temperatures via a low-noise current amplifier
and a lock-in amplifier. The only difference in our experimental data between those
using the λ 4 waveplate and those using the PEM is that the former presents a
background current even at normal incidence, while the latter does not. This is
most likely due to the Damber effect when interband excitation was employed
(Ganichev and Prettl, 2003). To show only the polarization dependent signal, all
the data presented are measured by PEM, which gives the difference between the
σ − and σ + polarizations. The incident light beam is polarized in the z − y plane,
with an angle of Θ away from the normal. By changing the angle φbetween the
polarization plane of the incident light and the optical axis of the λ 4 waveplate
or PEM, we can measure the helicity dependence of the photocurrent, which is the
hallmark of spin photocurrent that distinguishes it from other photocurrent effects.
Figure 8.2a and b show the dependence of the photocurrent of the two samples on
the laser beam polarization, represented by angle φ. between the optical axis of the

λ 4 modulator and the polarization plane of the incident light, with λ = 880 nm and

Θ = 300 . It clearly shows that when φ changes from −π 4 to +π 4, the current
varies from one maximum to the other maximum of opposite direction, in good
agreement with the fitting using sin 2φ (see the discussion below). Figure 8.2d
shows the Θ (incidence angle) dependence of the spin photocurrent. When
Θ. = 0, since there is no y-component of optically induced electron spin polarization,
no current is created. With the increase of Θ, leading to a larger y-component of

the electron spin polarization, the spin photocurrent first gets larger but finally gets
smaller because of the increased reflection.
Ganichev and Prettl (2003) have systematically used inter-subband excitation to
study the CPGE induced spin photocurrent. Inter-band excitation using circularly
polarized light will also produce a spin photocurrent, for which the microscopic
mechanism can be schematically described as shown in Fig. 8.2c. Considering the
SIA induced Rashba interaction, there will be non-parabolic terms in the Hamilto-
nian, such as the linear (in k// ) spin splitting for electrons (or light holes), and spin
3
splitting of the heavy-hole (Winkler, 2000) states proportional to k// . The coupling
between the spin and wave vector of the carriers, as well as the spin related selec-
tion rules, yield a non-uniform distribution of carriers in k-space upon circularly
polarized optical excitation. The imbalanced momentum relaxation of electrons in
the conduction band results in a net current due to the absorption of the circularly
polarized light.
For quantum well structures of symmetry C2v grown along the principal axis
[001] with structural inversion asymmetry, a photocurrent can be generated only
under oblique incidence of irradiation. If the incidence is in the (y, z) plane, then
the photocurrent is induced along the x direction (Ganichev et al., 2001; Ganichev
and Prettl 2003) and can be phenomenologically estimated to be
jx = γx y t p ts sin θ E 02 Pcirc , (8.1)
where γx y is a pseudotensor component related to the Rashba coupling coefficient,

tp and ts are the transmission coefficients from Fresnel’s formula for linear p and s
the electric field amplitude in vacuum, θ is the angle of refrac-
polarizations, E o is √
tion (sin θ = sin Θ/ ε∗ , ε∗ ∼ 13). Since
Iσ − − Iσ +
Pcirc = = sin 2φ, (8.2)
Iσ − + Iσ +
jx is expected to be proportional to sin 2φ at a fixed incidence angle Θ, just as we

have observed. Considering that tp and ts are Θ dependent, we can also fit the Θ
dependence of the experimental jx based on Eq. (8.1) and the Fresnel’s formula
4 cos2 Θ
t p ts = . (8.3)
cos Θ + ε∗ − sin2 Θ ε∗ cos Θ + ε∗ − sin2 Θ
The solid line in Fig. 8.2d is the fitted result, which is in excellent agreement with
the experimental data. Furthermore, we find that the photocurrent is almost linearly
dependent on the laser power. For sample D, the current drops faster when tempera-
ture is higher than 200 K, leaving a very small signal at 300 K. But for sample E, the
current only drops about a factor of 4 from 11 K to room temperature. We believe
that the larger photocurrent of sample E is due to its much larger spin splitting, as
also revealed by the SdH experiment. We also find that the inter-band excitation
induced spin photocurrent is up to 2 orders of magnitude stronger than that of the

inter-subband transitions in similar experiments. In addition to the incident direction
and the polarization dependence of the spin photocurrent, we have found that it also
changes direction when the laser wavelength is changed, as indicated in Fig. 8.3 for
the spectra dependence of the CPGE photocurrent. The photoreflectance spectra of
the samples are shown to clearly indicate the quantized energy levels of electrons
and holes as marked by the arrows. For the two samples, we have explicitly shown
that the CPGE spin photocurrent will change its sign when the laser wavelength
changes, which is in agreement with the theoretical prediction by Golub (2003). The
spectral inversion in the CPGE spin photocurrent is thought to be a characteristic
to distinguish the spin splitting mechanisms between the SIA and bulk inversion
asymmetry (BIA). From the big dip in the spectra response, we can tell that the
SIA should be the dominating mechanisms for the spin splitting in the samples.
This is also consistent with the dramatic change of the Rashba coefficient with our
modification of the sample structure as mentioned in the SdH measurement. There
are mainly two contributions for spectral inversion of the spin photocurrent. The
first one comes from the transitions from the light hole to the electron subband. Due
to the selection rules for light hole and heavy hole (such as −3/2 → −1/2 for
heavy hole, −1/2 → +1/2 for light hole with j = +1), electrons excited from
the heavy hole and the light hole bands by the same circularly polarized light will
jump to opposite spin splitting branches (+1/2 or −1/2) in the conduction band,
which will result in reversed current in the SIA case. Secondly, as pointed out by
Golub (2003), the negative contribution can also come from the heavy hole related
Fig. 8.3 Spectral response of the spin photocurrent for (a) sample D and (b) sample E. The pho-
toreflectance spectra of the two samples are also shown to determine the electronic structures. The
arrows indicate the heavy hole (hh) and light hole (lh) related transitions
transitions at larger wavevector due to the mixing or anti-crossing of the hole bands.
The spectra inversion in the CPGE spin photocurrent is very simple and useful in
distinguishing the spin splitting mechanisms between the SIA and BIA. We intend
to carry out further studies along this direction, to involve ground states excitations
by changing the sample composition and/or expand the laser wavelength range, and
make careful calculations for comparison.
There have been theoretical predictions for spatially homogeneous spin polar-
ization resulting from an electrical current (Edelstein, 1990) in systems such as
2DEGs, in additional to the spin Hall effect caused by the spin accumulation at the
edges of a sample due to the existence of a spin current perpendicular to a charge
current. Indeed we can simply interpret this current-induced spin polarization as the
converse process of the spin photo-current. Since the Rashba coupling term in the
Hamiltonian can be expressed as

H S O = α σ · z × k// = α k// · (
σ × z ) , (8.4)
one can interpret this coupling as if the momentum (carried by a current) induces
an effective magnetic field, or the spin induces an effective momentum. Inoue
et al. (2003) first derived the diffusive conductance tensor for a disordered 2DEG
with spin-orbit interaction and showed that the applied bias (E x ) induces a spin
accumulation

Sy = 4πeτ DλE x , (8.5)
where D = m e /2π2 is the 2DEG density of states per spin, the lifetime τ is the
momentum relaxation time, and λ = α E Z / represents the Rashba interaction.
From Eq. 8.3 we find that the intensity of the current-induced spin polarization
will also give a measure of the spin-orbit interaction parameter α. Kerr effect was
used to monitor the spin polarization along the y-direction under an external elec-
tric field in the x-direction, as shown in the inset of Fig. 8.4b. In the Kerr rotation
experiment, a polarizer and an analyzer were placed in the incident and reflected
laser beams, respectively. A PEM situated before the analyzer on the reflected beam
was employed to sensitively monitor any optical polarization rotation induced by
the spin polarization in the samples. To get rid of any possible reflectivity changes
induced by heating effects due to the applied electric field, the Kerr rotation was
extracted from the signal difference between positive and negative electric fields.
The laser wavelength was tuned to be close to the transition energy between the
third excited state of heavy hole and the electron, which was identified from pho-
toreflectance measurements. Figure 8.4a and b show the Kerr rotation against the
external electric field at 11 K for samples D and E, respectively. The detected
Kerr signal increases with increase of the external field, which is in agreement
with the theoretical prediction that the spin polarization Sy is proportional to E x .
However, we didn’t observe linear relationship between the spin polarization and
external field, which can originate from two possibilities. Firstly, nonlinearity at low
field mostly comes from the experimental uncertainty due to the limited sensitivity.
Fig. 8.4 Electric field (Ex ) dependence

of the Kerr rotation of sample D (a) and E (b) to show the
current induced spin polarization S y at 10 K. The temperature dependence of the Kerr rotation of
sample D (Ex = 50 V/cm) is shown in the inset of (a). The experimental setup is schematically
shown in the inset of (b), of which P1 and P2 represent the polarizer and analyzer, respectively
The second reason is the electric field dependent spin relaxation time. The spin
relaxation time is found to decrease (Kato et al., 2004) with increase of the electric
field, resulting in a reduced spin polarization or spin density, which is the reason that
Kerr rotation signal gets saturated at high field. Also we find that sample E gives a
Kerr rotation about 3–4 times larger than that of sample D, which is consistent with
the much larger spin-orbit coupling coefficient in the former as determined by the
SdH oscillation. The temperature dependence of the Kerr rotation, as presented in
the inset of Fig. 8.4a for sample D, shows that the spin polarization drops quickly
when temperature is above 100 K, which is due to the reduced spin and momentum
relaxation time at elevated temperatures. The experimental setup for converting spin
current into electric current is shown in Fig. 8.5b, with a crossing bar carved on the
above sample E. The Rashba coupling coefficient is about α = 3.0 × 10−12 eVm
determined by Shubnikov-de Hass oscillation (Golub, 2003). The time-resolved
Kerr rotation measurement with standard pump probe technique gives the spin
decoherence time τs ≈ 106 ps at 77 K as shown in Fig. 8.5a. The Fermi velocity
of the sample is estimated to be v F ≈ 5.0 × 105 m/s (taking the effective mass as
0.07 m e). Thus the spin coherence length is roughly estimated to be ls ≈ 70 μm
at 77 K. Taking consideration of these parameters we designed a setup with two
electric channels of widths WS = 20 μm and WL = 200 μm, respectively, one being
narrower and the other wider than the spin coherence length, i.e., W S < ls < W L,
as shown in Fig. 8.5b. The planar electric channels were carved out by standard
photo-lithography and wet etching along [110] and [1− 10], defined as lab x- and
y-axis respectively (see Fig. 8.5b). Electric contacts were made to the 2DEG by
Ni/Au metal electrodes at far ends of the electric channels.
Linearly polarized laser light of wavelength 880 nm and power 40 mW produced
by a tunable Ti:sapphire laser was used as the exciting source for the inter-band
Fig. 8.5 (a) The data of time resolved Kerr rotation for spin decoherence was fitted by an exponen-
tial function exp[−t/τs ] (the red line) with the spin decoherence time τs = 106 ps. (b) Geometry
of the asymmetric crossbar-shaped 2DEG sample, where the spin coherence length is expected to
lie between the width of the narrow and wide channels
transition. The incident light was firstly guided through a photo-elastic modula-
tor (PEM) which modulated the light polarization to be oscillating between two
orthogonal directions with a frequency of 50 KHz. Then the light was focused
through a 10× objective lens before being normally shed upon a 10 μm diameter
spot on the sample which was mounted in a cryostat at liquid nitrogen temperature.
The electric currents passing through the terminals were monitored by voltage drops
at symmetrically loaded resistors of 633 Ω. The voltages were read out with a lock-
in amplifier which was triggered by the PEM so that the electric currents induced by
non-polarization related phenomena e.g. Dember effect and thermoelectric effects
were clearly eliminated.
No measurable current along the wide channels (y-axis) was observed on the
background noise of tens of picoamperes under normal incidence, though a signifi-
cant photocurrent was observed if the circular polarized light was shed at an oblique
angle on the same samples (Yang et al., 2006). However, when light with linear
polarization along the x- or y-axis was normally incident onto the plane near the
electric channel junction, we observed reproducible currents along the narrow chan-
nels under unbiased conditions. Figure 8.6c–e show the currents passing through
the terminals as functions of the spot position when the incident light is along the
20 μm-wide channel (x-axis). The current peaks at x = 0 and 200 μm corresponding
to the positions of the two channel junctions, and vanishes at a distance of 50 μm
away from the peak position. Note that the current curves of IC−A and IC−B and
of I A−D and I D−B almost overlap, indicating that the current flows inward or out-
ward simultaneously through the channel junction. The spatial distribution of the
observed current is substantially different from the current due to Hall effect or the
reciprocal spin Hall effect, where the Hall current is unidirectional. The flow pattern
of the currents and the absence of a net electric current along the wide channel are
the signatures of the observed effect as summarized in Fig. 8.6b. The vanishing of
Fig. 8.6 (a) Schematic view of the 2DEG sample and the experimental setup. The 20 and 200 μm-
wide planar channels are carved out by standard photolithography and wet etching. A linearly
polarized light beam, polarized along x-axis as indicated by red color or along y-axis as indicated
by blue color, scans along the narrow channel (x-axis) at normal incidence. (b) Flow pattern of
measured electric currents for light spot along the narrow channel. (c), (d), and (e) show the typical
charge currents through all the terminals as a function of the light spot position at x-axis
all currents at the middle point of the C−D channel, x = 100 μm, evidently indicates
that the observed phenomena are not due to conventional photocurrents. Note that
the spatial location of the light spot depicted in Fig. 8.6 is hundreds of microns
away from the metal electrodes, therefore any contribution to the measured elec-
tric currents from the interfaces between the metal electrodes and semiconductor
channels is negligible. As the current measurements were locked at the polariza-
tion oscillating frequency, the obtained current measures the difference between the
electric currents generated by the two orthogonally polarized lights in the device.
To further examine the relationship between the observed currents and polarization
of the incident light, we rotate the polarization plane of the incident light while the
light spot is fixed at the peak position of IC−A and IC−B in Fig. 8.6c. The current
varies with Φ, the angle between the light polarization plane and the y-axis, and
a good fit is obtained with the form cos 2Φ, as shown in Fig. 8.7. Obviously the
polarization dependence of the current is in excellent agreement with the calculated
angle dependence of the spin current (Li et al., 2006; Zhou and Shen, 2007). Note
that the lock-in technique we used does not measure any Φ-independent current
because of cancellation.
The result of the laser scan along the edge of channel A−B at x = 10 μm is
shown in Fig. 8.4. Electric currents I A−C and I B−C were observed near the chan-
nel junction, and they were almost identical. The current I A−B remains negligible
Fig. 8.7 The charge current IA−C (black dot with error bars) at channel junction x = 0 varies with
the polarization angel ∼ of light relative to the lab frame, as shown in the inset. The red curve is a
fit of cos 2Φ
despite the laser spot position moving. The spatial dependence of the current are
drastic in the vicinity of Y = 0, or along the central line of channel C−D. When the
laser scans along the central line of channel A−B, i.e., at x = 100 μm, no current was
observed, for the light spot is located well beyond the spin coherence distance. This
is consistent with the conventional spin photogalvanic effect, where the current van-
ishes when the linearly polarized light falls normally on the plane. We thus conclude
that the observed electric current near the junction with a narrow width (20 μm) is
not a conventional photocurrent (Ganichev et al., 2001, 2002; Yang et al., 2006).
The equality of I A−C and I B−C obviously excludes the possibility of a diffusion
mechanism for generating the photo-current.
To eliminate the possible effect of the specific sample, we also carried out
measurements on sample D, with a stronger Rashba coupling about α = 6.3 ×
10−12 eVm. Essentially the same results were observed. To understand the experi-
mental results, we interpret the observed electric current as a result of the scatter-
ing of the spin current near the junction of the crossing bars. Firstly, the optical
excitation serves as the source of a spin current with its spin polarization lying in
the plane and perpendicular to the direction of the electron motion. In detail, the
linearly polarized light selectively pumps electrons in the valence band up to the
conduction band which is split due to the strong spin-orbit coupling. As a result
the excited electrons with non-zero spin-dependent momenta form a spin current
before recombining with holes in the valence band. The effective distance of this
spin current is estimated to be within the spin coherence length ls , which has the
relation of WL/2 > ls > WS/2. Keeping this picture in mind the observed effect
can then be well understood as follows. In the case that the light spot scans along
the central line of the narrow channel C−D in Fig. 8.6, the geometric boundaries
of the sample set a limit to the motion of the excited electrons and lead them to
two opposite directions with opposite in-plane spin polarizations. Two beams of
the excited electrons compose the same spin current, which is then scattered by the
edges of the crossing region to produce a transverse electric current. On the other
hand, when the light spot scans along the central line of the wide channel A−B in
Fig. 8.8 (a) Schematic view of the light scanning along the 200 μm-wide channel. The red and
green scanning lines are parallel to y-axis and are 10 and 100 μm away from the channel edge
respectively. (b) Flow pattern of measured electric current for light spot scanning along the edge.
(c) The corresponding charge currents IA−C show strong position dependence near the junction
center
Fig. 8.8, the spin current decays to zero before it reaches the edges of the crossing
region, therefore no electric current circulates. While in the case that the light spot
moves along the edge, also shown in Fig. 8.8, when the distance between the spot
and the edge is within the distance of ls , the edge may again impose a constraint
to the movement of the electrons and lead to an outward (inward) electric current
along the edge, together with an inward (outward) electric current perpendicular
to the edge, and thus a measurable current between C−A/C−B (in the case the
edge is close to terminal C) or D−A/D−B (in the case that the edge is close to D).
As it is known from the optical Bloch equation (Zhou et al., 2007). that the spin
current depends on the angle between the edge and the polarization direction of
light, Jx y = J0 + J1 cos 2, the same dependencies on the light polarization of
the electric current as shown in Fig. 8.7 can be considered as a direct evidence to
strongly support the above scenario. Furthermore, the sample is carved on a uniform
strip of InGaAs/InAlAs, so that the spin injection problem is clearly avoided when
the spin current passes through the scattering region at the junction. Also as the spin
coherence length at liquid nitrogen temperature is longer than the width WS = 20 μm
of the narrow channel, the impurity effect or other interactions would not be crucial
in this experiment. Finally, as the phase coherence length of electrons in this system
is much shorter than WS, the possibility that this effect being a mesoscopic one
should be excluded.
8.3 Discussion and Summary
In summary, we have shown two converse effects, spin photocurrent and current-
induced spin polarization, in a specially designed Rashba system, i.e. a 2DEG sys-
tem with asymmetric potential barriers. It is worth noticing the comparison of the
two samples used in this work. The Rashba coefficient measured from SdH oscilla-
tions has a ratio of about 2 between samples E and D. For the spin photocurrent, the
efficiency ratio is a bit less than 3 between these two samples. For the Kerr rotation,
the relative efficiency ratio is about 3 to 4. This evidently shows that the physical
mechanisms behind the three effects are uniformly based on the Rashba coupling,
and any of them can be taken as an indication of the Rashba effect. Though the gen-
eral trends are consistent, we do not expect to find precise values for these ratios,
as so many parameters are involved in any quantitative comparison. In particular,
as shown in Eq. 8.5, the momentum relaxation time is involved in evaluating the
current-induced spin polarization, so that more work in evaluating the lifetime will
be helpful to further elaborate the correlation between various spin-splitting induced
effects. As the CPGE was observed under obliquely incident excitation of inter-
band transitions, the spin photocurrent was found to have efficiency much larger
than the similar effect induced by inter-subband transitions. Meanwhile, the Kerr
effect induced by a current flow in the same samples is quite pronounced as well,
indicating the possibility of driving spin polarization with a current. The current-
induced spin polarization based on the Rashba effect may be a useful spin source
for spintronic devices, in addition to magnetization under magnetic field. Finally,
the spectral inversion in the CPGE spin photocurrent provides us an easy way to
distinguish the SIA and the BIA in the spin-orbit coupling.
The conversion of spin current to electric current has been discussed extensively
in the context of the reciprocal spin Hall effect where the seperated spins are out of
plane (Hirsch et al., 1999). In the present case the spin current is polarized within the
plane, as the linearly polarized light does not generate out-of-plane spin polariza-
tion due to conservation of the total angular momentum. As examined in (Li et al.,
2006), an electron of spin-y polarization will be equally scattered along the x and
−x axes, and hence the resulted scattered electric current has such a pattern: either
both outward or both inward, in the transverse directions of the spin current. Our
experimental observation of the electric currents is in excellent qualitative agree-
ment with this picture. In the Landauer-Büttiker formalism, numerical simulations
were performed in a symmetric crossbar shaped setup with the tight binding approx-
imation. The numerical results and symmetry analysis show that the resulted electric
current depends strongly upon the spin polarization of the spin current. For the case
of the in-plane-polarized spin current as in the present experiment, the conversion

rate for the spin current to electric current is estimated to be 0.3 ∼ 1% (Li et al.,
2006). In the present work, although there are some unknown parameters of the
sample, the amplitude of the injected spin current Jx y by the linearly polarized
light can be roughly estimated by using the measurable photocurrent Jx injected
by the circularly polarized light at a small oblique angle Θ away from normal. If
the powers of the applied two laser lights are equal, a theoretical estimation gives
Jx ()/Jx y ≈ 0.17(2e/) by solving of semiconductor Bloch optical equation
near the Γ point of the InGaAs 2DEG (Zhou and Shen, 2007). Taking the light spot
size, the geometry of the setup, and the laser power into account, we estimate from
the experimental data that Jx (Θ) ≈ 94ΘnA (Yang et al., 2006), hence the injected
spin current is estimated to be approximately 550 nA (h/2e). The observed electric
current in the present experiment is several nA, say 4.0 nA for the peaks in Fig. 8.6c.
The conversion rate from the spin current to charge current is roughly about 0.8%,
close to the numerically simulated values.
Briefly speaking, we observed onward or outward electric currents in a crossbar-
shaped (001) InGaAs/InAlAs quantum well when linearly polarized light irradiates
on the sample at normal incidence near the junction. Both the spatial pattern and
polarization dependence of the electric current are in agreement with the scenario
that the electric current is produced by the optically injected spin current scattered
near the crossing edges. Our observation provides a realistic technique to detect
spin currents, and opens a new route to study spin-related science and engineering
in semiconductors.
Acknowledgments The authors thank S. J. Xu, L. Ding and X. Z. Ruan for assistance in lab-
oratory facility and X. Dai, J. Wang, B. Zhou, S. C. Zhang and Q. Niu for helpful discussions.
This work was supported by the Research Grant Council of Hong Kong under Grant No.: HKU
7039/05P, the Large-Item-Equipment Funding and the University Development Funding of HKU
on “Nanotechnology Research Institute” program under contract No. 00600009. The university
grant DAG 04/05.SC 21 and 05/06.SC30 from HKUST. After completion of the present work
we were beware of relevant work by Valenzuela and Tinkham (2006), where the electric current
induced by spin current of the out-of-plane polarization in metal Al was detected experimen-
tally. We also greatly appreciate Prof. B. A. Foreman for his help and his critical reading of the
manuscript.
References
Aronov, A.G., Lyanda-Geller, Y.B., Pikus, G.E., 1991, Soviet Physics - JETP, 73, pp. 537–541.
Awschalom, D.D., Loss, D., Samarth, N. (eds) 2002, Semiconductor Spintronics and Quantum
Computation (Springer, Berlin).
Bel’kov, V.V., Ganichev, S.D., Schneider, P., Back, C., Oestreich, M., Rudolph, J., Hägele,
D., Golub, L.E., Wegscheider, W., Prettl, W., 2003, Circular photogalvanic effect at
inter-band excitation in semiconductor quantum wells. Solid State Communications, 128,
pp.283–286
Bhat, R.D.R., Nastos, F., Najmaie, A., Sipe, J.E., 2005, Pure spin current from one-photon absorp-
tion of linearly polarized light in noncentrosymmetric semiconductors. Physical Review Letters,
94, p. 096603.
Bychkov, Y.A., Rashba, E. I., 1984, Oscillatory effects and the magnetic susceptibility of carriers
in inversion layers. Journal of physics C: Solid state physics, 17, pp. 6039–6045.
Chaplik, V., Entin, M.V., Magarill, L.I., 2002, Spin orientation of electrons by lateral electric
field in 2D system without inversion symmetry. Physica E: Low-dimensional Systems and
Nanostructures, 13, pp. 744–747.
Crooker, S.A., Furis, M., Lou, X., Adelmann, C., Smith, D.L., Palmstrom, C.J., Crowell, P.A.,
2005, Imaging spin transport in lateral ferromagnet/semiconductor structures. Science, 309,
pp. 2191–2195.
Datta, S., Das, B., 1990, Electronic analog of the electro-optic modulator. Applied Physics Letter,
56, pp. 665–667.
Duckheim, M., Loss, D., 2006, Electric-dipole-induced spin resonance in disordered semiconduc-
tors. Nature Physics, 2, pp. 195–199.
Edelstein, V.M., 1990, Spin polarization of conduction electrons induced by electric current in
two-dimensional asymmetric electron systems. Solid State Communications, 73, pp.233–235.
Engels, G., Lange, J., Schapers, Th., Luth, H., 1997, Experimental and theoretical approach to
spin splitting in modulation-doped Inx Ga1−x As/InP quantum wells for B → 0. Physical Review
B, 55, pp. R1958–R1961.
Ganichev, S.D., Prettl, W., 2003, Spin photocurrents in quantum wells. Journal of physics C: Solid
state physics, 15, pp. R935–R983.
Ganichev, S.D., Danilov, S.N., Eroms, J., Wegscheider, W., Weiss, D., Prettl, W., Ivchenko, E.L.,
2001, Conversion of spin into directed electric current in quantum wells. Physical Review
Letters, 86, pp. 4358–4361.
Ganichev, S.D., Danilov, S.N., Schneider, P., Bel’kov, V.V., Golub, L.E., Wegscheider, W.,
Weiss, D., Prettl, W., 2004, Can an electric current orient spins in quantum wells?
arXiv:cond-mat/0403641v1.
Ganichev, S. D., Ivchenko, E.L., Bel’kov, V.V., Tarasenko, S.A., Sollinger, M., Weiss, D.,
Wegscheider, W., Prettl, W., 2002, Spin-galvanic effect. Nature, 417, pp. 153–156.
Golub, L.E., 2003, Spin-splitting-induced photogalvanic effect in quantum wells. Physical
Review B, 67, p. 235320.
Hirsch, J.E., 1999, Spin Hall Effect. Physical Review Letters, 83, pp. 1834–1837.
Hübner, J., Rühle, W.W., Klude, M., Hommel, D., Bhat, R.D.R., Sipe, J.E., Driel, H.M.van, 2003,
Direct observation of optically injected spin-polarized currents in semiconductors. Physical
Review Letters, 90, pp. 216601.
Inoue, J., Bauer, G.E.W., Molenkamp, L.W., 2003, Diffuse transport and spin accumulation in a
Rashba two-dimensional electron gas. Physical Review B, 67, p. 033104.
Kato, Y.K., Myers, R.C., Gossard, A.C., Awschalom, D.D., 2004a, Current-induced spin
polarization in strained semiconductors. Physical Review Letters, 93, p. 176601.
Kato, Y.K., Myers, R.C., Gossard, A.C., Awschalom, D.D., 2004b, Observation of the spin Hall
effect in semiconductors. Science, 306, pp. 1910–1913.
Li, J., Dai, X., Shen, S.Q., Zhang, F.C., 2006, Transverse electric current induced by optically
injected spin current in a cross-shaped InGaAs/InAlAs system. Applied Physics Letter, 88,
p. 162105.
Murakami, S., Nagaosa, N., Zhang, S.C., 2003 Dissipationless quantum spin current at room
temperature. Science, 301, pp. 1348–1351.
Saitoh, E., Ueda, M., Miyajima, H., Tatara, G., 2006, Conversion of spin current into charge
current at room temperature: Inverse spin-Hall effect. Applied Physics Letter, 88, p. 182509.
Schapers, Th., Engels, G., Lange, J., Klocke, Th., Hollfelder, M., Luth, H., 1998, Effect of the
heterointerface on the spin splitting in modulation doped Inx Ga1−x As/InP quantum wells for
B → 0. Journal of Applied Physics, 83, pp. 4324–4333.
Shen, S.Q., Ma, M., Xie, X.C., Zhang, F.C., 2004, Resonant spin Hall conductance in
two-dimensional electron systems with a Rashba interaction in a perpendicular magnetic
field. Physical Review Letters, 92, p. 256603.
Sih, V., Myers, R.C., Kato, Y.K., Lau, W.H., Gossard, A.C., Awschalom, D.D., 2005, Spatial
imaging of the spin Hall effect and current-induced polarization in two-dimensional electron
gases. Nature Physics, 1, pp. 31–35.
Silov, A. Yu., Blajnov, P.A., Wolter, J.H., Hey, R., Ploog, K.H., Averkiev, N.S., 2004, Current-
induced spin polarization at a single heterojunction. Applied Physics Letter, 85, pp. 5929–5931.
Sinova, J., Culcer, D., Niu, Q., Sinitsyn, N.A., Jungwirth, T., MacDonald, A.H., 2004, Universal
intrinsic spin Hall effect. Physical Review Letters, 92, p. 126603.
Stevens, M.J., Smirl, A.L., Bhat, R.D.R., Najmaie, A., Sipe, J.E., Driel, H.M.van, 2003, Quantum
interference control of ballistic pure spin currents in semiconductors. Physical Review Letters,
90, p. 136603.
Valenzuela, S.O., Tinkham, M., 2006, Direct electronic measurement of the spin Hall effect.
Nature, 442, pp. 176–179.
Winkler, R., 2000, Rashba spin splitting in two-dimensional electron and hole systems. Physical
Review B, 62, pp. 4245–4248.
Winkler, R., 2003, Spin-Orbit Coupling Effects in Two-Dimensional Electron and Hole Systems
(Springer, Berlin).
Wolf, S.A., Awschalom, D.D., Buhrman, R.A., Daughton, J.M., Molnár, S. von, Roukes, M.L.,
Chtchelkanova, A.Y., Treger, D.M., 2001, Spintronics: A spin-based electronics vision for the
future. Science, 294, pp. 1488–1495.
Wunderlich, J., Kaestner, B., Sinova, J., Jungwirth, T., 2005, Experimental observation of the
spin-Hall effect in a two-dimensional spin-orbit coupled semiconductor system. Physical
Review Letters, 94, p. 047204.
Yang, C.L., He, H.T., Ding, L., Cui, L.J., Zeng, Y.P., Wang, J.N., Ge, W.K., 2006, Spectral
dependence of spin photocurrent and current-induced spin polarization in an InGaAs/InAlAs
two-dimensional electron gas. Physical Review Letters, 96, p. 186605.
Zhou, B., Shen, S.Q., 2007, Deduction of pure spin current from the linear and circular spin
photogalvanic effect in semiconductor quantum wells. Physical Review B, 75, p. 045339.
Zutic, I., Fabian, J., Sarma, S.D., 2004, Spintronics: Fundamentals and applications, Reviews of
Modern physics. 76, pp. 323–410.
Chapter 9
Continuum Modelling of Nanoscale
Hydrodynamics
Ping Sheng∗ , Tiezheng Qian and Xiaoping Wang
9.1 Introduction
It may come as a surprise to most physicists and material scientists that up until
recently, continuum hydrodynamics can not accurate model fluid(s) flow at the
nanoscale. The problem lies not in the Navier Stokes (NS) equation, which expresses
momentum balance and therefore must be valid, but in the boundary condition at
the fluid-solid interface, required for the solution of the NS equation. Traditionally
the no-slip boundary condition (NSBC) has always been used to solve hydrody-
namic problems, which states that there can be no relative motion at the fluid-solid
interface. The purposes of this article are to delineate the problem and to present
its resolution through the application of Onsager’s principle of minimum energy
dissipation, which underlies almost all the linear response phenomena in dissipative
systems.
Even though lacking in first principles support, NSBC was widely regarded as a
cornerstone in continuum hydrodynamics, owing to its broad applicability in diverse
fluid-flow problems. However, an important exception was pointed out some time
ago in the so-called the moving contact line (MCL) problem in immiscible flows.
Here the contact line is defined as the intersection of the two-phase immiscible
fluid-fluid interface with the solid wall. In 1974, Dussan and Davis showed with
rigor that under the assumptions of (1) fluid incompressibility, (2) rigid and flat
solid wall, (3) impenetrable fluid-fluid interface, and (4) NSBC, there is a velocity
discontinuity at the MCL, and the tangential force exerted by the fluids on the solid
wall in the vicinity of the MCL is infinite. Subsequently, by employing molecular
dynamic (MD) simulations, Koplik et al. (1988) and Thompson and Robbins (1988)
have shown that near-complete slip occurs at the MCL, in direct contradiction to
the NSBC. The failure of the NSBC and the lack of a viable alternative mean that
the usual continuum hydrodynamics can not accurately model fluids flow on the
micro/nanoscale. This has often been referred to as an example of the so-called
“breakdown of continuum.”
∗ Department of Physics and Institute of Nano Science & Technology, Hong Kong University of
Science and Technology, Clear Water Bay, Kowloon, Hong Kong, China.

C Springer 2008
100 Sheng et al.
Faced with this challenge to the NSBC, there are two choices. The first is to
retain the NSBC and regard the slip at the MCL to arise from the “fracturing” of the
fluid-solid interface. In this scenario, the very large tangential stress in the vicinity of
the MCL can break the fluid-solid bonds that are responsible for the NSBC, and the
resulting slip hence lies outside the realm of hydrodynamics, just as solid fractures
can not be described by linear elasticity. Thus the slip occurs only in the immediate
molecular vicinity of the MCL, while NSBC holds elsewhere. The problem with
this scenario is that since the conclusion of Dussan and Davis is independent of the
fluid velocity, at very low velocities the tangential stress can be very small. And if
the molecular dimension offers a lower cutoff to the divergence in force, that force
may not be large enough to break any molecular bond, and the scenario becomes
inconsistent with itself.
The other possible scenario is that hydrodynamics holds everywhere, including
the MCL. But in that case the obvious question is: What is the relevant boundary
condition that would resolve the force divergence and yield the near-complete slip
at the MCL? Of the two choices, this second scenario is generally regarded as more
daunting, since success would mean that the NSBC is at best only an approximation.
In other words, success in finding the alternative boundary condition(s) is tanta-
mount to a direct challenge to the validity of the NSBC.
We offer a resolution to the MCL problem as posed by Dussan and Davis, along
the line of the second scenario (Qian et al., 2003, 2004, 2006). What differentiates
the new boundary conditions from the NSBC is their derivation, in conjunction with
the Stokes equation (from which the Navier-Stokes equation can be obtained by
including inertial forces), from the basic principle that underlies the linear response
theories in dissipative systems—Onsager’s principle of minimum energy dissipation
(Onsager, 1931a,b). The viewpoint expounded by this derivation is that fluid and
solid are composed of molecules, albeit with different interactions. Hence if there
is viscous dissipation in bulk fluid, there should be a similar form of dissipation at
the fluid-solid interface, characterized by some parameter that would play the role
of viscosity in the bulk. By doing so the present approach derives both the equations
of motion and the boundary conditions within a unified framework. The relevance
to the nanoscale is that the liquid-solid interfacial dissipation is characterized by a
coefficient β with the dimension of [viscosity]/[length]. In particular, if one divides
the bulk viscosity by β, the resulting length scale, denoted the slip length s , is
generally in the nanoscale. Hence the ratio s /L, where L is some length scale
intrinsic to the system, e.g., solid particle size in colloids, determines whether the
NSBC is a good approximation (when s /L 1), or the new boundary conditions
are needed (when s /L ∼ 1).
In what follows, a brief recapitulation of the MCL history in Section 9.2 is fol-
lowed by the statement of results and comparison with MD simulations in Sec-
tion 9.3. The derivation of the relevant boundary conditions will be preceded by a
short introduction to the principle of minimum energy dissipation in Section 9.4. In
Section 9.5 the continuum hydrodynamics formulation is shown to emerge varia-
tionally as the Euler-Lagrange equations from the dissipative energy functional. In
Section 9.6 we compare the numerical predictions of our continuum hydrodynamics
with MD simulation results, and discuss the implications.
9 Continuum Modelling of Nanoscale Hydrodynamics 101
9.2 History of the MCL Problem
Since the MCL represents our incision point on the question of hydrodynamic
boundary condition, it is only appropriate to acknowledge the many prior and
contemporary works on this problem. As shown in Fig. 9.1, when one fluid dis-
places another immiscible fluid, the contact line moves relative to the solid wall.
The incompatibility of the NSBC with the MCL has been pointed out by Moffat
(1964), Huh and Scriven (1971), Dussan and Davis (1974), Dussan (1976), and
de Gennes (1985). But it was Dussan and Davis (1974) who showed that in the
continuum formulation of the problem with the usual assumptions, the tangential
viscous stress would vary as ηV /x along the wall, where η is the viscosity, V
the wall speed, and x is the distance along the wall away from the MCL. This
variation directly leads to a non-integrable singularity, implying infinite dissipation.
There have been many models aimed at resolving the problem. These include the
kinetic adsorption/desorption model by Blake and Haynes (1969), the various slip
models of Hocking (1977), Huh and Mason (1977), and Zhou and Sheng (1990),
the interface formation/disappearance model of Shikhmurzaev (1997), the diffuse-
interface models by Seppecher (1996), Jacqmin (2000), Chen et al. (2000), Pismen
and Pomeau (2000), and Briant and Yeomans (2004), plus the variational model
of Glasner (2005). It should be noted that none of these models can reproduce the
so-called slip profile (see Fig. 9.2) obtained from MD simulations.
In a separate approach, Cox divided the MCL problem into an inner region and
an outer region (Cox, 1986). By carrying out an asymptotic analysis, he found to the
leading order of the capillary number Ca = ηV /γ a relation between the apparent
contact angle (the angle of the fluid-fluid interface at some mesoscopic distance
away from the solid boundary) in the outer region and the microscopic contact angle,
the capillary number, and the distance from the contact line. What happens in the
Fig. 9.1 When one fluid displaces another immiscible fluid, the contact line is moving relative to
the solid wall. Owing to the contact-line motion, the dynamic contact angle θd deviates from the
static contact angle θs which is determined by the Young equation γ cos θs + γ2 = γ1
102 Sheng et al.
Fig. 9.2 Inset shows two immiscible fluids under shear between two parallel plates. The relative
fluid-solid velocity, normalized to that of the plate, is shown as a function of horizontal position
along the fluid-solid boundary. Zero is the position of the static contact line, and vx = 0 means total
slip. Here open circles are from MD simulations, and solid line is calculated from Eqs. (9.5–9.7).
σ and ε are the molecular length and energy parameters from MD simulations
inner region is packaged into a model-dependent constant in the asymptotic relation.

An interesting observation is that the asymptotic behavior is independent of the
microscopic boundary conditions at the MCL (Dussan, 1976). But this conclusion
only deepens the mystery of what actually happens at the contact line.
A break on the MCL problem was offered by the MD simulation results (Koplik
et al., 1988; Thompson and Robbins, 1989). In Fig. 9.2 we show a typical MD
slip profile. Here the near-complete slip is indicated by the peak of the slip pro-
file, where zero means complete slip in the laboratory frame where the flat top and
bottom surfaces are moving in opposite directions with speed V , inducing Cou-
ette flow of the two immiscible fluids. The slip profile is measured at a molecular
distance away from the solid wall, through time-averaging the tangential (to the
wall) velocities of the molecules within the discretized cells. It is seen that away
from the MCL, the slip profile does not go to −1, which would mean no-slip, but
instead reaches a constant value that indicates partial slip. Such particle slip can be
well-accounted for by the Navier boundary condition (NBC), proposed nearly two
centuries ago: v slip = ls ς̇, where v slip is relative tangential velocity between the
fluid and solid at the fluid-solid interface, ls is the slip length, and ς̇ is the shear
rate at the interface (Navier, 1823). The small value of the slip length explains why
the NBC is practically indistinguishable from the NSBC in single-phase flows. The
fact that the partial slip can be readily visualized in MD is due to the large fluid
velocity and its associated large shear rate generic to such simulations, since other-
wise it would be extremely difficult to extract the mean behavior from the Brownian
thermal noise.
While the NBC can account for the partial slip in single-phase flows, it fails to
account, by an order of magnitude, for the near-complete slip of the MCL. Recent
MD-continuum hybrid simulations have been applied to investigate this problem
by Hadjiconstantinou (1999), and by Ren and Weinan (2007). But such approaches
leave unresolved the problem of the MCL boundary conditions.
The recent discovery of the generalized Navier boundary conditions (GNBC)

(Qian et al., 2003, 2004), as stated in the next section, has resolved the MCL
conundrum by reproducing for the first time the MD slip profile through continuum
hydrodynamic calculations (solid line in Fig. 9.2). Its derivation via the principle
of minimum energy dissipation in Section 9.5 shows the form of GNBC to be
uniquely necessitated by the thermodynamics of immiscible two-phase flows (Qian
et al., 2006).
9.3 Statement of Results
For two-phase immiscible flows, a free energy functional is required to stabilize the
interface. The simplest Landau free energy functional with a double-well potential
has the Cahn-Hilliard form (Cahn and Hilliard, 1958):
s
K u 4
FC H [φ(r )] = d r (∇φ) + − φ + φ
2 2
, (9.1)
2 2 4
where φ is the field variable defined as (n 2 − n 1 )/(n 2 + n 1 ), n 1(2) being the number
density of fluid 1(2) at the spatial point r, and K , s, u are material parameters
whose values can be fixed by the fluid-fluid interfacial thickness, interfacial tension,
and the condition that the double well has minima at φ = ±1. In addition to the
fluid-fluid interfacial free energy, there is also the interfacial free energy per unit
area at the fluid-solid interface, γ f s (φ), which changes continuously from the value
γ f s (−1) (fluid 1 and the solid) to γ f s (+1) (fluid 2 and the solid). The chemical
potential μ for the fluid-fluid interface and the similar quantity L for the fluid-solid
interface can be obtained from the variation of the total free energy:

δ FC H [φ] + d S[γ f s (φ)] = d r[μδφ] + d S[Lδφ], (9.2)
where
μ = −K ∇ 2 φ − sφ + uφ 3 , (9.3)
L = K ∂n φ + ∂γ f s (φ)/∂φ. (9.4)
Here the subscript n means normal to the fluid-solid interface. In terms of the
phase field variable φ and velocity v, the two equations of motion are
∂φ
+ v · ∇φ = M∇ 2 μ, (9.5)
∂t
∂ v
ρ + v · ∇ v = −∇ p + ∇ · σ vis + μ∇φ + fe , (9.6)
∂t
104 Sheng et al.
with the subsidiary fluid incompressibility condition ∇ · v = 0 and the boundary

conditions
vn = 0, (9.7a)
∂n μ = 0, (9.7b)
∂φ
+ vτ ∂τ φ = −Γ L(φ), (9.7c)
∂t
β(φ)vτslip = −η(∂n vτ + ∂τ vn ) + L(φ)∂τ φ. (9.7d)
Here M (Eq. (9.5)) and Γ (Eq. (9.7a)) are two phenomenological parame-
ters characterizing the mobility and boundary relaxation of the two fluid species,
ρ denotes mass density, p denotes pressure, σ vis = η[(∇ v) + (∇ v)T ] is the viscous
stress tensor, fe denotes external body force, and subscript τ denotes the component
tangential to the fluid-solid interface. In Eq. (9.7d), β(φ) stands for the fluid-solid
interfacial slip coefficient, it has the dimension of (viscosity)/(length) and can be
different for the two fluids, hence the dependence on φ.
Equations (9.7a) and (9.7b) are the physical boundary conditions stating the
impenetrability of the solid boundary, while Eqs. (9.7c) and (9.7d) are denoted the
generalized Navier boundary conditions (GNBC) (Qian et al., 2003, 2006), because
the two emerge consistently from the principle of minimum energy dissipation (see
Section 9.5). Equations (9.5–9.7) form a well-defined mathematical system. All the
parameter values can be obtained from MD simulations, although for M and Γ the
effort would be significantly more than that, e.g., for the viscosity. Their numerical
solution directly yields the solid curve shown in Fig. 9.2. In that case the M and Γ
values have been obtained by optimizing the fit, while all the other parameter values
were obtained from MD simulations. Once the M and Γ values were fixed, excellent
no-adjustable-parameter agreement has been achieved between MD simulations of
Couette and Poiseuille flows and their continuum counterparts (Qian et al., 2003,
2006). These will be presented in Section 9.6. We just note here that the last term
on the right hand side of Eq. (9.7d), denoted the uncompensated Young stress, is
responsible for the near-complete slip of the MCL. These and other points will be
further discussed in Section 9.6.
9.4 Onsager’s Principle of Minimum Energy Dissipation
It has been noted by Helmholtz that if the fluid velocity is specified on the bound-
aries, then the incompressible stationary flow (with negligible inertia forces) that
minimizes the viscous dissipation represents the solution to the Stokes equation
(Batchelor, 1967). The variational principles involving dissipation have been fur-
ther developed by Rayleigh (1873), Onsager (1931a,b), Edwards and Freed (1974),
and Doi (1983). We attribute to Onsager for the version that foreshadowed many
later developments in statistical mechanics of dissipative systems, including the
reciprocity theorem, the fluctuation-dissipation theorem, etc.
Onsager’s principle of minimum energy dissipation is about the rules governing

the optimal paths of deviation and restoration to equilibrium. As a variational prin-
ciple, it differs from the principle of minimum free energy in its application as an
approach for deriving the equations of motion (and in the present case the boundary
conditions as well), but not for determining the global dissipative state. The latter is
in contrast to the principle of minimum free energy, which is often used to determine
the global ground state of matter.
To see how the principle of minimum energy dissipation arises, it is instructive to
use a one-variable example for illustration. Let α be the displacement from equilib-
rium. In an overdamped dissipative system, the dynamics may be described by the
Langevin equation
∂ F(α)
ηα̇ = − + ς (t), (9.8)
∂α
where α̇ denotes the rate of change of α, F(α) is the relevant free energy, and
ς (t) is the white noise with a zero mean, satisfying the correlation ς (t)ς (t )
=
2ηk B T δ(t − t ), where k B denotes the Boltzmann constant and T the temperature.
The left hand side of Eq. (9.8) is simply the dissipative force, which is balanced by
the conservative force plus a stochastic force on the right. With the white noise term
on the right hand side of Eq. (9.8), the dynamics of α is no longer deterministic and
its trajectory is best described by a probability density P(α, t) that is governed by
the Fokker-Planck equation
2

∂P ∂ P 1 ∂ ∂F
=D + P , (9.9)
∂t ∂α 2 k B T ∂α ∂α
where the diffusion constant D satisfies the Einstein relation ηD = k B T . It is simple

to verify that the stationary solution of Eq. (9.9) is given by the Boltzmann distribu-
tion Peq ∝ exp[−F(α)/k B T ]. The transition probability for α at t to α at t + Δt is
given by

1 (α − α)2 F(α ) − F(α)
P(α , t + Δt|α, t) = √ exp − exp − ,
4π DΔt 4DΔt 2k B T
(9.10)
for α in the vicinity of α and small Δt. By using the Einstein relation, the two
exponents can be combined:

1 A
P(α , t + Δt|α, t) = √ exp − , (9.11)
4π DΔt 2k B T
where

η(α − α)2 η 2 ∂ F(α)
A= + [F(α ) − F(α)] ≈ α̇ + α̇ Δt (9.12)
2Δt 2 ∂α
106 Sheng et al.
is the quantity to be minimized if we want to maximize the probability of transition

with respect to α . For small Δt, it is seen that instead of minimizing A with respect
to the target state α , the same is achieved my minimizing with respect to the rate α̇.
Indeed, if we carry out the simple minimization on the right hand side of Eq. (9.12),
we obtain the force balance equation
∂ F(α)
ηα̇ = − , (9.13)
∂α
i.e., the Langevin equation without the stochastic force term. This is reasonable,
since the stochastic force has a zero mean, so Eq. (9.13) is true on the average.
Thus we learn from the above that (a) there can be a variational functional, of
which the quantity A is the one-variable version, which should be minimized with
respect to the rates, (b) the result of the such minimization would guarantee the
force balance on average, and (c) the minimization would also yield the equations
of motion and the related boundary conditions, which represent the most probable
course of a dissipative process. The last statement essentially guarantees that in the
statistical sense, the most probable course will be the only course of action observed
macroscopically.
For the general case of multivaribles, the so-called dissipative variational func-
tional can be simply generalized from Eq. (9.12) as
1 ∂ F(α1 , · · ··, αn )
n
A= ηi j α̇i α̇ j + α̇i , (9.14)
2 ∂αi
i, j i=1
where in the case of αi s being field variables, the summation should be replaced
by integrals, and partial derivatives by functional derivatives. In Eq. (9.14) the dissi-
pation coefficient matrix elements ηi j must be symmetric with respect to the inter-
change of the two indices, as shown by Onsager (1931a,b) based on microscopic
reversibility.
9.5 Variational Derivation of the Fluid-Solid Boundary

Conditions
Armed with the variational principle, in this section we intend to first construct
the energy dissipation functional A, from which desired boundary conditions and
the equations of motion can be derived. As seen from Eq. (9.14), there are two
parts to the functional. For the first part, we have to identify the various sources
of dissipation. For the second part, we have already written down the relevant free
energies in Section 9.3, what is required is to write down their variation with time.
There are basically four dissipation sources: the bulk viscous dissipation, the
dissipation due to inter-diffusion at the fluid-fluid interface, the frictional dissipation
at the fluid-solid interface, and the relaxational dissipation around the contact line,
also due to inter-diffusion (Qian et al., 2006). For the viscous dissipation, it is well
known that the total dissipation rate may be written as
η
Rvis = d r (∂i v j + ∂ j vi )2 . (9.15)
2
Since we regard the fluid and solid similarly as collections of molecules but with
different interactions, it is natural to assume that the qualitative form of the energy
dissipation at the fluid-solid interface is similar to the viscous dissipation in bulk
fluid. That is, if we start with Rvis = d r[η(∂z vx )2 ] for a shear flow with ∂z vx = 0
only, then for a fluid layer along the solid boundary with thickness Δz normal to
the surface, we can write Rvis = d S[(η/Δz)(Δvx )2 ], where d S = d r/Δz is the
surface differential. Since Δvx is the relative (tangential) velocity between the fluid
layer and the solid boundary, it is precisely what we would call the slip velocity.
That directly suggests the form of frictional dissipation rate at the fluid-solid
interface to be

Rslip = d S[β(vτslip )2 ], (9.16)
where the slip coefficient β has the dimension of [viscosity]/[length]. Hence a

slip length may be defined as ls = η/β. It should be emphasized that while Rslip
arises from the assumption of relative slip at the fluid-solid interface; there is no
specification of how much slipping there should be. Even an infinitesimal amount
of slipping would lead to the form of Eq. (9.16). In both Eqs. (9.15) and (9.16), the
rate variable [α̇i in Eq. (9.14)] is noted to be the fluid velocity.
For dissipations related to inter-diffusion and relaxation, the relevant variable
is φ. For small perturbations away from the equilibrium (concentration of the two
fluid species), there can be dissipation related to the diffusive current in restor-
ing the equilibrium. Since φ is locally conserved, it must satisfy the continuity
equation
∂φ
φ̇ = + v · ∇φ = −∇ · J, (9.17)
∂t
where J is the diffusive current and also the rate variable. Since the Onsager dissi-
pation function must be quadratic in rates, we thus have

| J|2
Rdi f f = d r , (9.18)
M
where M is the mobility coefficient. Equation (9.18) is exactly in the form of dif-
fusive dissipation in general. At the fluid-solid interface, φ is no longer conserved
because diffusive transport normal to the solid boundary is possible (∂n Jn = 0 in
108 Sheng et al.
general). The only rate variable here is φ̇, hence boundary relaxation would induce
a dissipation rate

φ̇ 2
Rrel = dS , (9.19)
Γ
where Γ is a constant related to the boundary mobility coefficient.

For the rate of change of the free energy (second term on the right hand side of
Eq. (9.14)), we may write from Eq. (9.2) that since δφ = ∇φ · δ x + (∂φ/∂t)δt, the
time variation of the free energy is given by

∂φ ∂φ
d r μ + dS L = d r μ(φ̇ − v · ∇φ) + d S L(φ̇ − v · ∇φ)
∂t ∂t

v · ∇φ] + d S[L φ̇ − Lvτ ∂τ φ]
= d r[μφ̇ − μ

= d r[−μ∇ · J − μ
v · ∇φ] + d S[L φ̇ − Lvτ ∂τ φ]

= d r[∇μ · J − μ
v · ∇φ] + d S[L φ̇ − Lvτ ∂τ φ], (9.20)
where the first term of the last line is obtained from the first term of the previous
line through the use of integration by parts, with

d r[∇ · (μ J)] = d S[μJn ] = 0
because Jn = 0 at the solid boundary. The last line of Eq. (9.20) may be decom-
posed into the entropy part and the work part according to general thermodynamics,
i.e., Ḟ = −T Ṡ + Ẇ , where

−T Ṡ = d r[∇μ · J] + d S[L φ̇], (9.21)
and

Ẇ = d r[−
v · (μ∇φ)] + d S[−vτ (L∂τ φ)]. (9.22)
Whereas the entropy part reflects the dissipations caused by diffusion and bound-
ary relaxation of the two species of fluid, the work part is seen to be the “elastic”
force/stress exerted by the interface on the flow. Here μ∇φ is the capillary force
density and L∂τ φ the stress of the fluid-fluid interface at the solid boundary. For the
stationary state, it is easy to see from the first line of Eq. (9.20) that Ḟ = 0. That is,
in steady state the diffusive transport of the fluid-fluid interface must be balanced by
the kinematic transport by the flow.
By collecting all the terms, the variational functional A can be written as

η
v (r ), J(r ), φ̇(r )] =
A[ (∂i v j + ∂ j vi )2
d r
4
2 2
β slip 2 J φ̇
+ dS (vτ ) + d r + dS
2 2M 2Γ

+ d r[∇μ · J] + d S[L φ̇] + d r[− v · (μ∇φ)]

+ d S[−vτ (L∂τ φ)]. (9.23)
The above functional should be minimized with respect to the three rates
v, J, φ̇ , supplemented with the incompressibility condition ∇ · v = 0. We first
minimize with respect to J. There are only two terms in Eq. (9.23) where the
diffusive current appears. The relevant Euler-Lagrange equation is given by
J = −M∇μ. (9.24)
When combined with the continuity Eq. (9.17), Eq. (9.24) directly leads to one
of the two equations of motion (9.5). Similar minimization with respect to φ̇ at the
solid boundary leads directly to the relaxation boundary condition (9.7c), one of the
two boundary conditions denoted as the GNBC.
To minimize with respect to v, it is necessary to first impose the incompress-
ibility
condition by the use of a Lagrange multiplier λ(r ), leading to an extra term
d r[λ∂i vi ]. Then the total variation with respect to the fluid velocity can be divided
into a group of terms for the bulk:

−η d r ∂ j ∂ j vi + ∂i v j δvi − d r [∂i λδvi ] − d r [μ∂i φδvi ] = 0, (9.25)
from which we get the Stokes equation
−∇ p + η∇ 2 v + μ∇φ = 0, (9.26)
where we have identified the Lagrange multiplier λ = − p. In the presence of inertial

forces, the Stokes equation (force = 0) is immediately generalized to the Navier-
Stokes equation (force = mass × acceleration) through momentum conservation.
That yields Eq. (9.6).
The other group of terms, resulting from the variation of A with respect to (tan-
gential) velocity at the solid-fluid boundary, is given by

η d S[∂n vτ δvτ + ∂τ vn δvτ ] + β d S[vτslip δvτ ] − d S[L∂τ φδvτ ] = 0, (9.27)
110 Sheng et al.
from which the boundary condition (9.7d) is obtained. This completes the
derivation.
It should not be a surprise to note that Eqs. (9.24), (9.7c) and (9.7d) are examples
of linear response, i.e., the response of a dissipative system is linearly proportional
to the forcing function. Thus the GNBC is a direct reflection of this general behavior.
Also, from our derivation it is clear that Eqs. (9.7c) and (9.7d) constitute a consistent
pair. It can be shown that both are important in the transport of the contact line across
chemically patterned surfaces.
Also, the GNBC does not exclude the NSBC. In fact, NSBC represents the
special case where the slip length is zero, or the slip coefficient β is infinite.
9.6 Implications for Nanoscale Hydrodynamics
Here we compare the numerical predictions of the continuum hydrodynamic system

(as given by Eqs. (9.5–9.7)) with MD simulations on molecular-scale fluid(s) flow.
In order to carry out meaningful comparisons, the parameters of the system must
first be fixed. This can be done easily for the density ρ, viscosity η, slip coefficient
β, interfacial tension γ , fluid-fluid interface thickness ξ , and static contact angle
θs by directly using those measured values from MD simulations. That leaves the
two phenomenological parameters M and Γ that theoretically can also be obtained
through MD simulations, but here we choose to treat them as parameters to be fixed
by fitting one set of MD data, shown in Fig. 9.2. Once fixed, we can predict the
velocity profiles, pressure profiles, and interfacial shapes under different flow con-
figurations with no adjustable parameters. Below we show three cases. The first two
cases are two-phase immiscible Couette flows between two parallel plates separated
by H , with different static contact angles. In the first case, denoted the symmetric
case, the fluid-solid interaction on two sides of the contact line (the fluid-solid inter-
facial tension γ f s in the contuum model) is the same. Hence θs is 90◦ . In the second
case, denoted the asymmetric case, the fluid-solid interactions are different on two
sides of the contact line, with θs = 64◦ . In the third case, we consider immiscible
Poiseuille flow between two parallel plates with separation H and the same static
contact angle as the asymmetric case. We will use√ the length and velocity units from
MD simulations. Namely, the velocity unit is ε/m and the length unit is σ . Here
m denotes the molecular mass, assumed to be the same for both liquids, ε is the
energy unit in the Lennard-Jones interaction between the liquid molecules, and σ is
the attendant molecular interaction range.
In Fig. 9.3 we show the comparison of the tangential (to the wall) and normal
velocity profiles for the symmetric case, at different distances z away from the solid
wall. In Fig. 9.4 we show the same comparison for the asymmetric case. In Fig. 9.5
we show the comparison of the interfacial profiles, and in Fig. 9.6 we show both the
interfacial profile (9.6a) and velocity (9.6b) comparisons for the Poiseuille flow. In
all cases the lines are the numerically evaluated continuum predictions and the sym-
bols are the MD results. Excellent quantitative agreements are seen. In particular,
for the tangential velocity profile closest to the solid wall, the near-total slip of the
MCL is clearly evident. This is a direct consequence of the uncompensated Young
Fig. 9.3 Comparison of the MD (symbols) and continuum (lines) velocity profiles (vx (x) and
vz (x) at different z levels) for a symmetric case of immiscible Couette flow (V=0.25(ε/m)1/2 and
H = 13.6σ). The profiles are symmetric about the centre plane z = H/2, hence only the lower
half is shown at z = 0.425σ (circles and solid lines), 2.125σ (squares and dashed lines), 3.825σ
(diamonds and dotted line), and 5.525σ (triangles and dot-dashed lines)
Fig. 9.4 Comparison of the MD (symbols) and continuum (lines) velocity profiles (vx (x) and vz (x)
at different z levels) for an asymmetric case of immiscible Couette flow (V=0.2(ε/m)1/2 and H =
13.6σ), shown at z = 0.425σ (circles and solid lines), 2.975σ (squares and long-dashed lines),
5.525σ (diamonds and dotted line), 8.075σ (up-triangles and dot-dashed lines), 10.625σ (down-
triangles and dashed lines), 13.175σ (left-triangles and solid lines). Although the solid lines are
used to denote two different z levels, for each solid line, whether it should be compared to circles
or left-triangles is self-evident
112 Sheng et al.
Fig. 9.5 Comparison of the

MD (symbols) and
continuum (lines) fluid-fluid
interface profiles, defined by
φ = 0. The circles and dotted
line denote the symmetric
immiscible Couette flow with
V = 0.25(ε/m)1/2 and
H = 13.6σ; the squares and
dashed line denote the
asymmetric immiscible
Couette flow with
V=0.2(ε/m)1/2 and
H = 13.6σ
stress (Qian et al., 2003, 2006) (the last term on the right hand side of Eq. (9.7d)),
which is a peaked function centered at the contact line. The integral of the uncom-
pensated Young stress (across the contact line) represents the tangential stress that
results from the deviation of the dynamic (microscopic) contact angle from the static
one. While such deviations are usually small (measured in MD simulations to range
from 0.5 to 2◦ ), owing to the relatively large magnitude of γ as compared to the
viscous stress (as measured by the capillary number Ca, generally in the range of
0.01–0.1 or even smaller), even a small deviation of the contact angle can imply a
Fig. 9.6 Comparison of the MD (symbols) and continuum (lines) results for an asymmetric case
of immiscible Poiseuille flow. An external force mgext = 0.05ε/σ is applied on each fluid molecule
in the x-direction, and the two walls, separated by H = 13.6σ, move at a constant speed V =
0.51(ε/m)1/2 in the −x-direction to maintain a stationary steady-state interface. (a) Fluid-fluid
interface profiles, defined by φ = 0. (b) vx (x) at different z levels. The profiles are symmetric about
the center plane z = H/2, hence only the lower half is shown at z = 0.425σ (circles and solid line),
2.125σ (squares and dashed line), 3.825σ (diamonds and dotted line), and 5.525σ (triangles and
dot-dashed line)
very large tangential projection of the capillary force. This is the basic reason why
the Navier boundary condition, which involves just the tangential viscous stress,
can not possibly explain the near-total slip of MCL. However, due to the peaked
nature of the uncompensated Young stress, away from the contact line the GNBC is
identical to the NBC.
An important point to note is that the phenomenon of near-total slip at the MCL
is clearly independent of the average fluid velocity. Even at low velocities there is
still near-total slip at the MCL. The reason is that while the total tangential stress
becomes smaller as the fluid velocity decreases, the absolute magnitude of the slip
velocity also decreases. The two are linearly proportional to each other, as stated by
Eq. (9.7d). That is, Eq. (9.7d) is in the form of a linear friction law.
Another important piece of information is in regard to the transition behavior
from total slip at the MCL to the regions far away, deep in the single phase regime.
Is the slip behavior abruptly stopped at some point away from the MCL? Does the
amount of relative slip decay exponentially away from the MCL? In fact both of
these possibilities were previously considered in the form of ad hoc fixes to the
stress singularity (if the NSBC were applied). So what is the true behavior? We
Fig. 9.7 Log-log plot for the slip profiles. Here vx /V + 1 is the scaled slip velocity at the lower
fluid-solid interface z = 0, and x/σ measures the distance from the MCL in units of σ. The wall
is moving at −V, hence vx /V = 0 means complete slip and vx /V = −1 means no slip. The vx
profiles were obtained for five symmetric cases of Couette flow, with different values for H but
the same value for V (= 0.05) and also the same parameters for densities and interactions. The
symbols represent the MD results and the solid lines represent the continuum results, obtained for
H = 6.8σ (black circles and line), H = 13.6σ (red squares and line), H = 27.2σ (green diamonds
and line), H = 54.4σ (blue up-triangles and line), H = 68σ (orange down-triangles and line).
There are two solid curves for each color, one for the slip profile to the left of the MCL and the
other to the right of the MCL. The dashed line has the slope of −1, indicating that the 1/x behavior
is approached for increasingly larger H. For H = 68σ, the 1/x behavior extends from to, where ls
was measured to be 2σ. Inset: The scaled tangential velocity vx /V at z = 0, plotted as a function
of x/σ
114 Sheng et al.
have carried out immiscible Couette flows in increasingly wider channels (Qian
et al., 2004). The inset to Fig. 9.7 shows the tangential velocity profiles at the wall.
Immediately next to the MCL, there is a small core region where the slip profiles
show a sharp decay within a few slip lengths. As the channel width H increases,
a much more gentle variation of the slip profiles becomes apparent. In order to
reveal the nature of this slow variation, we plot in Fig. 9.7 the same data in the
log-log scale. The dashed line has a slope −1, indicating a 1/x behavior of the slip
profile, where x is the distance from the MCL. Because of the finite H , there is
always a plateau in each of the single-phase flow regions, where the constant small
amount of slip is given by 2V s /(H + 2s ), which acts as an outer cutoff on the 1/x
profile. This amount of slip is simply derived from the NBC and the NS equation
for uniform shear flow. The continuum results shown in Fig. 9.7 were obtained on
a uniform mesh, using the same set of parameter values corresponding to the same
local properties in all the five MD simulations. We have extended the MCL sim-
ulations, through continuum hydrodynamics, to lower flow rates and much larger
systems, using the iterative grid redistribution approach (Ren and Wang, 2000). The
1/x partial slip profile is verified to several hundred slip lengths. That means even if
the slip length is in the nanoscale, partial slip can extend to mesoscopic distances.
9.7 Concluding Remarks
The most significant implication of our results is the fact that the fluid-solid bound-
ary conditions can be consistently derived from the same statistical mechanic princi-
ple that underlies the linear response theories. Making this possible is the viewpoint
that both fluid and solid are collections of molecules, so their interfaces, while spe-
cial, can still be treated within the same theoretical framework. This is the reason
why our continuum calculations agree so well with the MD simulations, because
the latter inherently treat both fluid and solid as composed of molecules. With
the present continuum formulation, it becomes possible now to carry out accurate
numerical calculations of nanofluidic/microfluidic phenomena that are beyond the
reach of MD simulations.
We note here that the use of Onsager’s principle of minimum energy dissipa-
tion represents a departure from the traditional theoretical approach of using the
kinetic theory to derive the boundary conditions. While the kinetic theory can be
useful in the case of rarified gases where the mean free path of collision is large,
its application to liquid-solid interface, where the interaction between liquid and
solid molecules is strong and the relevant mean free paths very small, can involve
difficulties in estimating the effects of the required approximations. Therefore even
though the kinetic theory can usually provide a very physical picture of the various
processes, it may not be as fruitful as the dissipation point of view in deriving the
relevant boundary conditions. It would be most helpful if the two approaches can be
combined in some way in the future.
What the resolution of the MCL problem tells us is that the NSBC is at best an
approximation. While the partial slip in single phase flows may be quite small, its
physical significance can not be minimized. First of all, we can see that without the
small partial slip at the fluid-solid interface, there can not be the total slip at the
MCL. The two are consistent and can not be separately treated under the framework
of minimum energy dissipation principle. Second, the slip coefficient is a parame-
ter whose statistical mechanical underpinning should be similar to that of the bulk
viscosity coefficient (Barrat and Bocquet, 1999a,b). Its theoretical and experimen-
tal studies remain a fruitful area to be further explored. This is in contrast to the
NSBC, which is just a clean statement which lies outside the realm of statistical
mechanics.
There are some clear physical implications associated with the GNBC and its
derivation framework. Here we would like to mention just two. The existence of a
length scale, the slip length, means that the ratio of particle size to the slip length is
now an important dimensionless parameter. When that ratio is large (i.e., for large
particles), the slip effect is small, and one can approximate the particle transport
dynamics (in fluid) by using the NSBC. In particular, the Cauchy-Born principle
holds, i.e., the macroscopic deformation is equal to the microscopic deformation.
But when the ratio of the particle size to slip length is moderate to small (e.g., for
colloids consisting of nanoparticles) then the slip effect can not be neglected, and
the Cauchy-Born principle in its traditional form would have to be modified. The
fact that some threshold particle size must exist (for the slip effect to be manifest)
is not in doubt, since in the limiting case of fluid molecules common sense tells us
that relative slip (between the molecules) occurs. A second implication is that when
the same framework of minimum energy dissipation principle is applied to complex
fluids, there can be order parameters other than the number density. The additional
order parameters would on the one hand introduce complexities into the boundary
conditions, but on the other hand also imply richness one can explore for targeted
purposes, e.g., lubrication. These and other problems are being actively pursued
at present.
References
Barrat, J.-L., Bocquet, L., 1999a, Large slip effect at a nonwetting fluid–solid interface. Physical
Review Letters, 82, pp. 4671–4674.
Barrat, J.-L., Bocquet, L., 1999b, Influence of wetting properties on hydrodynamic boundary con-
ditions at a fluid/solid interface. Faraday Discussions, 112, pp. 119–128.
Batchelor, G. K., 1967, An Introduction to Fluid Dynamics. Cambridge University Press.
Blake, T. D., Haynes, J. M., 1969, Kinetics of liquid/liquid displacement. Journal of Colloid and
Interface Science, 30, pp. 421–423.
Briant, A. J., Yeomans, J. M., 2004, Lattice Boltzmann simulations of contact line motion. II.
Binary fluids. Physical Review E, 69, p. 031603.
Cahn, J. W., Hilliard, J. E., 1958, Free energy of a nonuniform system. I. Interfacial free energy.
Journal of Chemical Physics, 28, pp. 258–267.
Chen, H. Y., Jasnow, D., Vinals, J., 2000, Interface and contact line motion in a two phase fluid
under shear flow. Physical Review Letters, 85, pp. 1686–1689.
Cox, R. G., 1986, The dynamics of the spreading of liquids on a solid surface. Part 1. Viscous flow.
Journal of Fluid Mechanics, 168, pp. 169–194.
116 Sheng et al.
Doi, M., 1983, Variational principle for the Kirkwood theory for the dynamics of polymer solutions
and suspensions. Journal of Chemical Physics, 79, pp. 5080–5087.
Dussan V. E. B., 1976, The moving contact line: the slip boundary condition. Journal of Fluid
Mechanics, 77, pp. 665–684.
Dussan V. E. B., Davis, S. H., 1974, On the motion of a fluid–fluid interface along a solid surface.
Edwards, S. F., Freed, K. F., 1974, Theory of the dynamical viscosity of polymer solutions. Journal
of Chemical Physics, 61, pp. 1189–1202.
De Gennes, P. G., 1985, Wetting: Statics and dynamics. Reviews of Modern Physics, 57,
pp. 827–863.
Glasner, K. B., 2005, Variational models for moving contact lines and the quasi-static approxima-
tion. European Journal of Applied Mathematics, 16, pp. 1–28.
Hadjiconstantinou, N. G., 1999, Hybrid atomistic-continuum formulations and the moving contact
line problem. Journal of Computational Physics, 154, pp. 245–265.
Hocking, L. M., 1977, A moving fluid interface. Part 2. The removal of the force singularity by a
slip flow. Journal of Fluid Mechanics, 79, pp. 209–229.
Huh, C., Mason, S. G., 1977, The steady movement of a liquid meniscus in a capillary tube.
Huh, C., Scriven, L. E., 1971, Hydrodynamic model of steady movement of a solid/liquid/fluid
contact line. Journal of Colloid and Interface Science, 35, pp. 85–101.
Jacqmin, D., 2000, Contact-line dynamics of a diffuse fluid interface. Journal of Fluid Mechanics,
402, pp. 57–88.
Koplik, J., Banavar, J. R., Willemsen, J. F., 1988, Molecular dynamics of Poiseuille flow and
moving contact lines. Physical Review Letters, 60, pp. 1282–1285.
Moffatt, H. K., 1964, Viscous and resistive eddies near a sharp corner. Journal of Fluid Mechanics,
18, pp. 1–18.
Navier, C. L. M. H., 1823, Memoire sur les lois du movement des fluides. Mem. l’Acad. R. Sci.
l’Inst. France, 6, pp. 389–440.
Pismen, L. M., Pomeau, Y., 2000, Disjoining potential and spreading of thin liquid layers in the
diffuse-interface model coupled to hydrodynamics. Physical Review E, 62, pp. 2480–2492.
Onsager, L., 1931a, Reciprocal relations in irreversible processes. I. Physical Review, 37,
pp. 405–426.
Onsager, L., 1931b, Reciprocal relations in irreversible processes. II. Physical Review, 38,
pp. 2265–2279.
Qian, T. Z., Wang, X. P., Sheng, P., 2003, Molecular scale contact line hydrodynamics of
immiscible flows. Physical Review E, 68, pp. 016306.
Qian, T. Z., Wang, X. P., Sheng, P., 2004, Power-law slip profile of the moving contact line in
two-phase immiscible flows. Physical Review Letters, 93, p. 094501.
Qian, T. Z., Wang, X. P., Sheng, P., 2006, A variational approach to moving contact line
hydrodynamics. Journal of Fluid Mechanics, 564, pp. 333–360.
Rayleigh, L., 1873, Some general theorems relating to vibrations. Proceedings of the London
Mathematical Society, 4, pp. 357–368.
Ren, W., E, W. 2007, Boundary conditions for the moving contact line problem. Physics of Fluids,
19, p. 022101.
Ren, W., Wang, X. P., 2000, An iterative grid redistribution method for singular problems in
multiple dimensions. Journal of Computational Physics, 159, pp. 246–273.
Shikhmurzaev, Y. D., 1997, Moving contact lines in liquid/liquid/solid systems. Journal of Fluid
Mechanics, 334, pp. 211–249.
Seppecher, P., 1996, Moving contact lines in the Cahn-Hilliard theory. International Journal of
Engineering Science, 34, pp. 977–992.
Thompson, P. A., Robbins, M. O., 1989, Simulations of contact-line motion: Slip and the dynamic
contact angle. Physical Review Letters, 63, pp. 766–769.
Zhou, M. Y., Sheng, P., 1990, Dynamics of immiscible-fluid displacement in a capillary tube.
Physical Review Letters, 64, pp. 882–885.
Chapter 10
Defect in Zinc Oxide Nanostructures
Synthesized by a Hydrothermal Method
A. B. Djurišić∗ , K. H. Tam, C. K. Cheung, Y. H. Leung, C. C. Ling,

C. D. Beling, S. Fung and W. K. Chan
10.1 Introduction
Zinc oxide (ZnO) which is a wide band gap (∼ 3.37 eV) semiconductor with large
exciton binding energy (∼ 60 meV), is expected to have great potential for electronic
and optoelectronic applications. Recently, a great variety of ZnO morphologies,
such as nanowires, nanorods, tetrapods and nanoribbons, has been reported (Wang,
2006). Among various morphologies, fabrication of ZnO nanorods (nanowires) have
attracted much attention because of its great potential to realized the new generation
of optoelectronic devices. ZnO nanorods arrays have been fabricated by various
methods, such as metal-organic chemical vapor deposition (MOCVD), vapor phase
deposition, pulsed laser deposition and hydrothermal growth.
10.1.1 Fabrication of ZnO Nanorod arrays
10.1.1.1 Vapor Phase Deposition
Vapor phase deposition has been extensively used for the fabrication of the ZnO
nanostructures. In this method, source materials are firstly evaporated under vac-
uum, then transported and condensed onto the substrates, which were placed in
a lower temperature zone. Well aligned ZnO nanorod arrays can be produced by
proper control of the supersaturated pressure. Vapor-solid-liquid (VSL) process is
one of the well known vapor phase techniques to fabricate ZnO nanorod array. It
was developed by Wagner and Ellis, 1964. Metal catalysts are introduced into the
synthesis process to direct the crystal growth in a specific direction. Au, Cu, Ni
and Sn are commonly used as typical catalysts. This synthesis method is simple,
however it involves relatively high growth temperature. The high temperature
environment may cause decomposition and undesired reactions in the reactor
(Yi et al., 2006).
∗
Department of Physics, The University of Hong Kong, Pokfulam Road, Hong Kong

C Springer 2008
118 Djurišić et al.
10.1.1.2 Metal-Organic Chemical Vapor Deposition (MOCVD)
Metal-organic chemical vapor deposition (MOCVD) is commonly used to fabricate

semiconductor film. Recently, ZnO nanorods arrays have been successfully prepared
using catalyst free MOCVD by Park et al. (2002). This work has provided a lower
temperature environment (∼ 400 ◦C) to fabricate high purity ZnO nanorods which
are favourable for the realization of the ZnO photonic and optoelectronic nanode-
vice. Further works need to be done to reduce the anisotropic growth, which may be
caused by the high speed laminar gas flow in the growth condition.
10.1.1.3 Pulsed Laser Deposition (PLD)
Pulsed laser deposition (PLD) is a deposition process which target material is vapor-
ized using a pulsed laser and deposited as a thin film on the substrate surface. It
is recognized that the ZnO nanostructure synthesized by this method usually have
excellent electronic and optical properties. ZnO nanorod arrays grown by PLD have
been firstly reported by Sun et al. (2004). Related research of ZnO nanorod arrays
fabrication via PLD has been reported by other groups. It was found that the ZnO
rod usually capped by a nano size tip Sun et al. (2006).
10.1.1.4 Hydrothermal Growth
Methods mentioned above can produce crystalline ZnO nanorods with high aspect
ratio. However, these synthesis methods usually required relatively high fabri-
cation temperature which limited the choice of substrates and lowered product
yield. Hydrothermal synthesis is of considerable interest since it is environmen-
tally friendly, has low production cost and involves low fabrication temperature
(Tam et al., 2006). Recently, lots of works have been done on the fabrication of
ZnO nanorods array on large (or flexible) substrate for electronic device applica-
tions. However, while low growth temperature is an advantage for the fabrication of
nanorods on flexible plastic substrates, it also results in a large defect concentration
in the nanorods. As-grown rods typically exhibit large concentrations of defects
and pronounced defect emission. The defect emission is strongly dependent on the
substrate preparation and post-fabrication treatment, such as annealing (Djurišić and
Leung, 2006).
10.1.2 Visible Defect Emission in ZnO
Green and yellow defect emissions are commonly observed in the photolumines-
cence spectra of ZnO. However, the origins of these emissions are still controversial.
Green emission is the most commonly observed defect emission in ZnO nanos-
tructures. Numbers of hypotheses have been proposed for the mechanism of this
10 Defect in Zinc Oxide Nanostructures Synthesized by a Hydrothermal Method 119
emission. The green emission is often attributed to singly ionized oxygen vacancies,
as suggested by Vanheusden et al. (1996). Vanheusden’s group investigated the cor-
relation between the green emission and the singly ionized oxygen vacancies in ZnO
by electron paramagnetic resonance (EPR) and photoluminescence. However, this
hypothesis is highly controversial, and has been questioned based on the absence of
EPR peak at g ∼ 1.96 in green emitting ZnO nanostructures (Djurišić et al., 2004).
Other hypotheses, such as antisite oxygen, zinc vacancy, complex defects including
zinc interstitials, surface defect and Cu impurity, have also been proposed. However,
the exact type of defect responsible for green emission has not been conclusively
identified. The origins of this green defect emission require further study. However,
there is convincing evidence that the defect is located at the surface. It was shown
that the defect emission is depressed by coating a surfactant on ZnO nanostructure
(Djurišić et al., 2004). Polarized luminescence experiments from the aligned ZnO
nanorods by Hsu et al. (2004) also indicated the defect emission is originated from
the surface.
Yellow defect emission is commonly observed in the ZnO nanostructures pre-
pared by hydrothermal method or electrodeposition. This emission is often assigned
to oxygen interstitial (Heo et al., 2005), due to the recombination of a delocalized
electron close to the conduction band with a deeply trapped hole in the oxygen
interstitial centers. This hypothesis has been supported by Lin et al. (2006), with the
observation that an increase of acceptor-like defects would enhance the intensity of
yellow emission. However, there is still no consensus on the origin of yellow defect
emission. Other hypotheses, such as hydroxyl groups, have also been proposed.
Zhou et al. (2002) suggested that the presence of Zn(OH)2 on the ZnO surface
led the quenching of the ZnO excitonic emission and was strongly correlated to the
yellow emission by investigating the annealed samples.
In addition to the intrinsic defects in the ZnO, Al, Li or H impurities have also
been suggested to be responsible for yellow emission. Ohashi et al. (1999) have
observed that the intensity of the yellow emission was proportional to the concentra-
tion of the paramagnetic center and that the concentration of the paramagnetic center
is increased by doping Al in the ZnO samples. Ortiz et al. (1997) proposed that the
yellow luminescence was probably related to Li, which was often unintentionally
doped in the hydrothermally grown ZnO. Similar experiments have been done by
Borseth et al. (2006), who showed that the weak yellow luminescence was caused by
Li-related defect. Experiments have been done to suppress the contribution of intrin-
sic defects by annealing the ZnO samples and yellow emission was still observed
from the Li-doped samples. In additional, yellow emission was also observed from
the hydrogen implanted single crystal ZnO by Ip et al. (2002).
10.1.3 Defect Identification
Apart from PL, several techniques have been used to identify the origin of the
defect emission in ZnO. Raman spectroscopy is commonly employed to study
the impurities, dopants and defect complexes in materials. This technique can be
used as a nondestructive measurement to determinate the growth mechanism of

ZnO nanostructures and defect identification. Kaschner and co-workers (2002) have
invsetigated that the influence of nitrogen on the lattice dynamic of ZnO by local
vibrational mode. Recently, Raman spectroscopy is used as an assistant tool for
determinating the defects in ZnO nanostructures.
Electron paramagnetic resonance (EPR) is another well-established tool for iden-
tifying paramagnetic defects in ZnO. As mentioned above, the EPR peak at g ∼ 1.96
is highly correlated to the defect emission in ZnO. Although the assignment of the
peaks is in question, this technique can still give more information about the defects.
In addition, it is straightforward for measuring the nanostructures, which can be
treated as any sample in powder form.
The deep-level transient spectroscopy (DLTS) is commonly employed to detect
the electrically active defects, which are located below the conduction band. This
technique is more applicable to thin films or single crystals. However, the results
can still be applicable to the origin of defect in nanostructures as defect position and
behavior are similar for the both nano and bulk cases.
X-ray photoelectron spectroscopy is a surface chemical analysis technique that
can be used to study the doping concentration and origin of defect emission in doped
ZnO. Photoelectron spectrum can be fitted in several peaks which corresponding
to the bonding energy between the atoms. Those peaks can give the information
about the origin of defect on surface by combining the study of photoluminescence.
Wang et al. (2006) have found that the green defect emission originates from surface
defects in ZnO nanoparticles by study the relation between the O 1s spectra and the
green emission.
Positron annihilation spectroscopy (PAS) is a useful technique for investigating
vacancy type defects in a material. As positron is implanted into the sample, it may
get trapped and then finally annihilated in open-volume-type defects of the sample.
The annihilation process produces two photons carrying information about the elec-
tronic environment of the defect position, thus this technique is generally suited for
detecting delocalized state and localized defect state in the materials. Early studies
by de la Cruz et al. (1992) characterized some vacancy-type defects in ZnO single
crystal.
In the positron beam experiment, positrons with variable energy (0–30 keV)
are implanted into the material. The implanted positrons would be thermalized
rapidly (within 1–10 ps) and undergo diffusion. Finally, the positrons are trapped
at the defect sites and annihilate with the electrons nearby. The variation of the
positron energy allows the detection of defects as a function of the penetration
depth. The positron implantation profile can be described by the Makhovian
distribution:
mz m−1 z
P(z) = exp[−( )m ] with
z 0m z0
z̄
z0 =
(1 + 1/m)
where the mean implantation depth z̄ is given by:
AE n
z̄ =
ρ
where ρ is the density of the sample in g cm−3 , E is the implanted positron energy
in keV and A, m and n are empirical parameters with values 3.6 × 10−6 gcm−2 ,
2.0 and 1.62 respectively.
In addition, the time-dependence of the probability, n(z)dz, to find positron in
depth dz can be described as the following ordinary linear second-order differential
equation:

d 2 n(z) d n(z) n(z) n 0 (z)
− − 2 + =0
dz 2 dz Λ(z) L D+
where L = (D+ τeff )1/2 is the positron diffusion length and = kT /ξ e is the
dimensions of length which characterizes the electric field ξ(z). With D+ is the
positron diffusion coefficient, τeff is the probability per second that a given position
will be removed from the freely diffusion state, T is temperature, ξ is the electric
field and e is the positron charge.
The effective diffusion length L eff + can be obtained by solving the above equa-
tions using the computer program VEPFIT (van Veen et al., 1990). This value is
related to the mobility of the positron and offers information in the positron trapping
process.
Annihilation of the positron-electron pair produce a pair of photons with total
energy of 2m o c2 (where m o c2 = 511 keV). Doppler shift can be observed from the
annihilation radiation. The Doppler shift depends on the momentum of annihilating
pair. Since the participating positrons are predominantly thermalized, the Doppler
shift depends on the motion of the electrons that located on the defect site. This
broadening of the annihilation peaks caused by the Doppler shift is used in the
PAS study; two parameters S and W are usually employed. Figure 10.1 shows the
definition of the S and W parameter schematically.
The S parameter is defined as the ratio of the area of central low-momentum part
of the 511 keV annihilation photopeak to the total area under the whole curve and
Fig. 10.1 S and W parameter

diagram
the W parameter is defined as the ratio of the area in the high-momentum region to
the total area under the annihilation peak.
∞ ∞
−∞ −∞ d p x d p y ρ( p)
1
S, W = ∞ ∞
−∞ −∞ d p x d p y ρ( p)
t ot
where Σ1 is over the appropriate energy window for S and W definition, Σtot is over
the full annihilation photopeak.
From the definition, it can be obtained that annihilations with valence electrons
are reflected in S parameter, and W parameter contains information from the core
electron. The S parameter is more sensitive to the vacancy state of the annihilation
site, because the valence electrons have a low momentum that is contributing in the
central region of the Doppler-broadening spectra. On the other hand, W parameter
is usually used to study the chemical surrounding of the annihilation site as the core
electrons have higher momentum.
In this paper, ZnO nanorods were prepared by hydrothermal method on silicon
substrates. The nanorods were studied by scanning electron microscopy (SEM),
photoluminescence (PL), X-ray photoelectron spectroscopy (XPS), positron annihi-
lation spectroscopy (PAS) and I-V curve measurements before and after annealing
in different environments and temperatures. It was found that the yellow emission
in hydrothermally grown nanorods can be attributed to the presence of OH groups.
The defect emission was strongly affected by the annealing conditions. The origins
of the defect emissions were discussed.
10.2 Experimental Details
The ZnO nanorod arrays were prepared from 25 mM aqueous solutions of zinc
nitrate hydrate and hexamethylenetetramine. Before the growth of nanorods, silicon
substrates were cleaned by sonication in acetone, ethanol, and deionized water, and
then dried in an oven. Then, a seed layer was prepared using zinc acetate solution,
as described by Greene et al. (2003). After that, the Si substrates with acetate seed
layer were placed in the solution at 90 ◦ C for 5 hours. For increasing the aspect ratio
of the nanorods, polyethyleneimine was added to the solution (Greene et al., 2005).
Zinc nitrate hydrate, hexamethylenetetramine, zinc acetate and polyethyleneimine
were obtained from Aldrich. The morphology of the nanorods was examined by
scanning electron microscopy (SEM) using a Leo 1530 field emission SEM.
Annealing has been performed in a tube furnace under different conditions.
The gas flow rate was 0.1 Lpm, and the pressure was ∼ 1 Torr. The optical prop-
erties of the annealed ZnO nanorods were investigated using room temperature
photoluminescence. The PL spectra were obtained using a fiberoptic spectrometer
PDA-512-USB (Control Development Inc.). The excitation wavelength was 325 nm
from a He-Cd laser with power of ∼ 15 mW.
In order to have more comprehensive study on the defects, PAS and XPS were
used to examine the ZnO nanorods before and after annealing. Low-energy positrons
were focus to a ∼ 1 mm diameter spot on the sample under magnetic guidance in
the positron annihilation experiment. The implantation energy of the positrons was
varied between 0 and 25 keV. Computer programme VEPFIT was used to fit the
experimental data, including positron implantation, positron diffusion, and positron
annihilating at different possible sites, for low-momentum S parameter. While XPS
spectra were obtained using PAL 102 XPS and the position of the carbon C 1s peak
(285.0 eV) was taken as a reference.
Furthermore, ZnO nanorods were grown on the p-type GaN to from a p-n hetero-
junction. The I-V characteristics of the junction with various annealing conditions
were investigated by Keithley 2400 Source Meter.
10.3 Results and Discussions
10.3.1 SEM Characterisation
Figure 10.2 shows the morphology of the as-grown ZnO nanorods. It was observed
that the rods were highly oriented. The rod was ∼ 500 nm long with diameter range
25–50 nm, while the density of the rods were on the order of several hundred rods
per μm2 . In addition, morphology of the rods did not show significant change on
different substrates and after annealing.
10.3.2 Photoluminescence
PL spectrum of ZnO nanorods on Si substrates before and after annealing under

different conditions are shown in Fig. 10.3. As-grown ZnO nanorods exhibit a UV
emission peak and a broad defect emission centre at yellow spectral range. The
defect emission was significantly depressed after annealing at 200 ◦C in all gas flows
which were tried in this work (Fig. 10.3a). The results were consistent with the
previous report by Hsu et al. (2006). PL spectra for ZnO nanorods prepared by
hydrothermal method showed large quenching in the yellow defect emission after
Fig. 10.2 SEM images of

ZnO nanorods (a) side view,
(b) top view. Scale bar is
200 nm
Fig. 10.3 PL spectra of ZnO nanorods annealed in different environments at (a) 200 ◦ C, (b) 400 ◦ C,
and (c) 600 ◦ C
Fig. 10.4 TEM image of

ZnO nanorods annealed in
forming gas at 600 ◦ C. Scale
bar is 50 nm
annealing in reducing environment. This decrease of intensity was likely due to

the removal of the hydroxyl groups after annealing (Zhou et al., 2002; Tam et al.,
2006). Since the desorption rate of hydroxyl groups peaks at ∼ 150 ◦C, annealing
the samples at 200 ◦C was likely to remove the Zn(OH)2 and water on the surface
of the nanorods, corresponding to the depression in the defect emission.
Figure 10.3b and 10.3c show the ZnO nanorods annealed in different environ-
ments at 400 ◦C and 600 ◦C, respectively. Different visible defect emissions could
be observed depending on the annealing environment. Nanorods showed green and
orange emissions after annealing in air. Green defect emission could be observed
after annealing in N2 and forming gas (5% hydrogen+95% nitrogen). Surface dam-
age of the ZnO nanorods can be found after annealing in forming gas at higher
temperature. The TEM image of the ZnO nanorods annealed at 600 ◦C is shown
in Fig. 10.4. It can be observed that the green emission increases with the surface
damage. Relation between the PL and surface damage of ZnO need further study.
The origin of the defect emission will be further discussed in Section 10.3.3.
ZnO samples grown by Pal et al. (2006) using low-temperature hydrothermal
synthesis also showed similar optical properties. The highest UV to visible ratio
of the ZnO samples could be achieved after annealing at ∼ 200 ◦C (the optimum
annealing temperature). In order to verify that hydroxyl groups are responsible for
the reduction of yellow defect emission, PL was measured on ZnO nanorods at
different time after annealed at 200 ◦C. Figure 10.5a shows the PL spectra of ZnO
nanorods stored in air after annealing in Ar at 200 ◦C. It could be observed that
the depressed defect emission reappeared after putting the sample in air. The defect
emission increased with an increasing time of storage in air, and became dominant
after 96 hours.
ZnO nanorods were coated with 3-aminopropyltrimethoxysilane immediately
after annealing at 200 ◦C for further studies. The inset of Fig. 10.5b shows that simi-
lar PL spectrum can be obtained after the nanorods coated with silane. Figure 10.5b
shows that almost no change of the PL spectra after storage in air for 96 hours for
nanorods coated with silane. The possible reason may be that the hydroxyl groups
and water molecules in air were re-adsorbed on the surface of ZnO nanorods and
caused the reappearance of the yellow emission. However, ZnO nanorods coated
(a)
(b)
Fig. 10.5 (a) PL spectrum of ZnO nanorods stored in air after annealed at 200 ◦ C. (b) PL spectrum
of ZnO nanorods coated with silane stored in air after annealed at 200 ◦ C. Inset: PL spectrum of
as-grown ZnO nanorods with and without silane
with silane were inert to the storage environment, thus no significant change in PL
spectrum could be found.
10.3.3 PAS and XPS Measurements
In order to obtain more information about the nature of defects in ZnO nanorods,
PAS and XPS were used to examine the ZnO nanorods before and after anneal-
ing. Table 10.1 shows a summary of the obtained results. In PAS measurement, the
effective positron diffusion length (Leff +) represents a measure of the probability of
positrons being trapped into all kind of defects in a sample, while the S-parameter
is only sensitive to certain types of defects, such as Zn vacancy and vacancy cluster
or complex in ZnO (Tam et al., 2006). It could be observed that Leff + of the as-
grown samples was much shorter than the annealed samples, which was consistent
Table 10.1 Summary on the PAS and XPS results of ZnO nanorods before and after annealing
Annealing Leff + S O1 /Zn O1 /O2 (O1 + O2 )/Zn
conditions (nm)
None 2.8 0.547 0.44 0.69 1.07
200 ◦ C Ar 22 0.559 0.43 0.68 1.07
400 ◦ C Ar 24 0.558 0.44 0.54 1.24
600 ◦ C Ar 28 0.554 0.46 0.55 1.31
200 ◦ C N2 11 0.560 0.38 0.53 1.11
600 ◦ C N2 42 0.556 0.42 0.46 1.35
with the low UV to visible ratio in the as-grown samples in PL measurement. The
increase of S-parameter upon annealing implied that an increase in Zn vacancy and
vacancy cluster. However, the increase of Leff + with increasing annealing temper-
ature showed that the total defect concentration decreased with higher annealing
temperature. Therefore, it could only be explained that ZnO annealed in N2 would
have the effect of increasing the concentration of Zn vacancy or vacancy cluster
and decreasing the concentration of oxygen vacancy, ionized acceptor or disloca-
tion. The results show that the origin of the green emission exhibits after 600 ◦C
annealing is highly correlated to the Zn vacancy.
Photoelectron spectra of O 1s curves are fitted by two peaks. The lower energy
peak is attributed to oxygen ions in the wurtzite ZnO structures, surrounded by Zn
atoms, and the higher one is typically assigned to the loosely bound oxygen on the
surface of the ZnO. In all of the examined samples, O1 /Zn ratio was lower than 0.5,
indicating oxygen deficiency in the bulk of the nanorods. The increase of O1 /Zn
and (O1 + O2 )/Zn ratio with higher annealing temperature indicated that there may
have an increase of Zn vacancy or a decrease of oxygen vacancy in the samples upon
annealing. This result exhibited a good agreement with the PAS measurement. The
decrease of O1 /O2 ratio and increase of Leff + and S parameter suggested that the
origin of green emission is involved Zn vacancy or Zn vacancy complexes, as well
as surface trapping by chemisorbed oxygen which suggested by van Dijken et al
(2000).
10.3.4 I-V Measurement
ZnO nanorods were also grown on the p-GaN on sapphire to perform room-
temperature I-V measurement. Figure 10.6 shows the I-V characteristics of het-
erojunction annealed at various temperatures in Ar flow. It can be observed that all
curves show typical diode behaviour. The highest current at the same bias voltage
is obtained after annealing in Ar at 200 ◦C. As-grown sample has the worst perfor-
mance. It should be noticed that sample with less defect emission in PL seems to
have higher current. The correlation between the PL spectrum and the I-V charac-
teristic need to be further studied. However there is evidence that the native defects
in ZnO play a very complex role in the performance of the ZnO/GaN diode.
Fig. 10.6 I-V characteristics of as-grown and annealed ZnO nanorods/GaN diodes
10.4 Conclusions
In conclusion, ZnO nanorods have been fabricated by hydrothermal synthesis.

As-grown nanorods showed large yellow defect emission. This defect emission
could be decreased by annealing the samples at 200 ◦C in all gases which were
tried in this work. However, defect emission reappeared after storing the samples in
air. The presence of the OH groups on the hydrothermally grown ZnO nanorods was
likely responsible for the yellow emission. The results obtained from PAS and XPS
showed that the defects which appeared after annealing at higher temperature possi-
bly involved Zn vacancy. I-V characteristics of the heterojunction of the ZnO/GaN
device showed that native defects in ZnO play an important role in the performance.
Acknowledgments The work reported in this paper has been supported by the research grants
council of the Hong Kong special administrative region, China (project numbers hku 7008/04p,
7010/05p, and 7019/04p). Financial support from the strategic research theme, university develop-
ment fund, seed funding grant (administrated by the University of Hong Kong) and outstanding
young researcher award are also acknowledged.
References
Borseth, T. M., Svensson, B. G., Kuznetsov, A. Y., Klason, P., Zhao, Q. X. and Willander, M., 2006,
Identification of oxygen and zinc vacancy optical signals in ZnO. Applied Physics Letters, 89,
p. 262112.
de la Cruz, R. M., Pareja, R., Gonzalez, R., Boatner, L. A. and Chen, Y., 1992, Effect of thermo-
chemical reduction on the electrical, optical-absorption, and positron-annihilation characteris-
tics of ZnO crystals. Physical Review B, 45, pp. 6581–6586.
Djurišić, A. B. and Leung, Y. H., 2006, Optical properties of ZnO nanostructures. Small, 2,
pp. 944–961.
Djurišić, A. B., Choy, W. C. H., Roy, V. A. L., Leung, Y. L., Kwong, C. Y., Cheah, K. W., Rao, T. K.
G., Chan, W. K., Lui, H. F. and Surya, C., 2004, Photoluminescence and electron paramagnetic
resonance of ZnO Tetrapod structures. Advanced Functional Materials, 14, pp. 856–864.
Greene, L. E., Law, M., Goldberger, J., Kim, F., Yanfeng, J. C. J., Richard, Z., Saykally, J. and
Yang, P., 2003, Low-temperature Wafer-scale production of ZnO nanowire arrays. Angewandte
Chemie International Edition, 42, pp. 3031–3034.
Greene, L. E., Law, M., Tan, D. H., Montano, M., Goldberger, J., Somorjai, G. and Yang, P.,
2005, General route to vertical ZnO nanowire arrays using textured ZnO seeds. Nano Letter, 5,
pp. 1231–1236.
Heo, Y. W., Norton, D. P. and Pearton, S. J., 2005, Origin of green luminescence in ZnO thin film
grown by molecular-beam epitaxy. Journal of Applied Physics, 98, pp. 073502(1–6)
Hsu, J. W. P., Tallant, D. R., Simpson, R. L., Missert, N. A. and Copeland, R. G., 2006, Luminescent
properties of solution-grown ZnO nanorods. Applied Physics Letters, 88, p. 252103.
Hsu, N. E., Hung, W. K. and Chen, Y. F., 2004, Origin of defect emission identified by polarized
luminescence from aligned ZnO nanorods. Journal of Applied Physics, 96, pp. 4671–4673.
Ip, K., Overberg, M. E., Heo, Y. W., Norton, D. P., Pearton, S. J., Kucheyev, S. O., Jagadish, C.,
Williams, J. S., Wilson, R. G. and Zavada, J. M., 2002, Thermal stability of ion-implanted
hydrogen in ZnO. Applied Physics Letters, 81, pp. 3996–3998.
Kaschner, A., Haboeck, U., Martin, S., Matthias, S., Kaczmarczyk, G., Hoffmann, A., Thomsen,
C., Zeuner, A., Alves, H. R., Hofmann, D. M. and Meyer, B. K., 2002, Nitrogen-related local
vibrational modes in ZnO:N. Applied Physics Letters, 80, pp. 1909–1911.
Lin, Y. J., Tsai, C. L., Lu, Y. M. and Liu, C. J., 2006, Optical and electrical properties of undoped
ZnO films. Journal of Applied Physics, 99, p. 093501(1–4).
Ohashi, N., Nakata, T., Sekiguchi, T., Hosono, H., Mizuguchi, M., Tsurumi, T., Tanakaand, J.,
Haneda, H., 1999, Yellow emission from zinc oxide giving an electron spin resonance signal at
g = 1.96. Japanese Journal of Applied Physics, 38, pp. L113–L115.
Ortiz, A., Falcony, C., Hernandez, J., Garcia, M. and Alonso, J. C., 1997, Photoluminescent char-
acteristics of lithium-doped zinc oxide films deposited by spray pyrolysis. Thin Solid Films,
293, pp. 103–107.
Pal, U., Serrano, J. G., Santiago, P., Xiong, G., Ucer, K. B. and Williams, R. T., 2006, Synthesis
and optical properties of ZnO nanostructures with different morphologies. Optical Materials,
29, pp. 65–69.
Park, W. I., Kim, D. H., Jung, S. W. and Gyu-Chul Yi, 2002, Metalorganic vapor-phase epitaxial
growth of vertically well-aligned ZnO nanorods. Applied Physics Letters, 80, pp. 4232–4234.
Sun, Y., Fuge, G. M. and Ashfold, M. N. R., 2004, Growth of aligned ZnO nanorod arrays by
catalyst-free pulsed laser deposition methods, Chemical Physics Letters, 396, pp. 21–26.
Sun, Y., Fuge, G. M. and Ashfold, M. N. R., 2006, Growth mechanisms for ZnO nanorods formed
by pulsed laser deposition. Superlattices and Microstructures, 39, pp. 33–40.
Tam, K. H., Cheung, C. K., Leung, Y. H., Djurišić, A. B., Ling, C. C., Beling, C. D., Fung,
S., Kwok, W. M., Chan, W. K., Phillips, D. L., Ding, L. and Ge, W. K., 2006, Defects in
ZnO nanorods prepared by a hydrothermal method. Journal of Physical Chemistry B, 110,
pp. 20865–20871.
Van Dijken, A., Meulenkamp, E. A., Vanmaekelbergh, D. and Meijerink, A., 2000, The Kinetics of
the radiative and nonradiative processes in nanocrystalline ZnO particles upon photoexcitation.
J. Phys. Chem. B, 104, pp. 1715–1723.
Vanheusden, K., Seager, C. H., Warren, W. L., Tallant, D. R. and Voigt, J. A., 1996,Correlation
between photoluminescence and oxygen vacancies in ZnO phosphors. Applied Physics Letters,
68, pp. 403–405.
van Veen, A., Schut, H., De Vries, J., Hakvoort, R. A., Ijpma, M. R., Schultz, P. J., Massoumi, G.
R., Simpson, P. J. (eds), 1990, Positron beams for solids and surfaces, AIP Conf. Proc. 218,
AIP, New York, p. 171.
Wagner, R. S. and Ellis, W. C., 1964, Vapor-Liquid-Solid mechanism of single crystal growth.
Applied Physics Letters, 4, pp. 89–90.
Wang, Z. L., 2004, Zinc oxide nanostructures: growth, properties and applications. Journal of
Physics: Condensed Matter, 16, pp. R829–R858.
Wang, Z. G., Zu, X. T., Zhu, S. and Wang, L. M., 2006, Green luminescence originates from surface
defects in ZnO nanoparticles. Physica E: Low-dimensional Systems and Nanostructures, 35,
pp. 199–202.
Yi, G.-C., Wang, C. and Park, W. I., 2006, ZnO nanorods: synthesis, characterization and applica-
tions. Semiconductor Science and Technology, 20, pp. S22–S34.
Zhou, H., Alves, H., Hofmann, D. M., Kriegseis, W., Meyer, B. K., Kaczmarczyk, G. and Hoff-
mann, A., 2002, Behind the weak excitonic emission of ZnO quantum dots: ZnO/Zn(OH)2
core-shell structure. Applied Physics Letters, 80, pp. 210–212.
Part III
Nano Materials Design and Synthesis
Chapter 11
Towards Surface Science Studies of Surfaces
Formed by Molecular Assemblies Using
Scanning Tunneling Microscopy
Chen Wang∗ and Shengbin Lei
11.1 Introduction
The adsorption of molecules on surfaces are the central issues for classical surface
science studies since the beginning of this important field. It could be noted that
the fundamental investigations have been dedicated to the surfaces of metals, semi-
conductors and oxides which have great implications in catalysis, microelectronic
devices etc. With the fast evolving research field of constructing molecular architec-
tures with expectations on various functions, the knowledge at single molecule level
on the interacting mechanisms are keenly needed. The development of the molec-
ular nanostructures has also ushered in an interesting studying topic relating to the
interfaces between dissimilar molecular species. Studies on heterogeneous molecu-
lar interfaces could have multiple implications from understanding of fundamental
aspects of inter-molecular interactions, to the working mechanisms of molecular
devices. This category of studies could also form a natural extension of the classical
surface science principles for metals, semiconductors and oxides. To this aspect,
scanning tunnelling microscopy (STM), with the demonstrated resolution capabil-
ity at single molecule level, could provide unique approaches to reveal interactions
between dissimilar molecules.
Numerous STM studies have helped gain important insight on the interactions
among dissimilar molecules adsorbed on supporting surfaces, such as those in the
molecular assemblies prepared under various conditions. In a less noted aspect, the
STM can also provide direct evidence on the molecular adsorptions occurred on
top of the close packed molecular layers serving as the support surfaces in the
way exactly as traditional pristine surfaces of metals, semiconductors and oxides.
The experimental foundations enabling such studies are due to the known STM
resolution capability within one nanometer or so in the surface normal direction on
conductive surfaces. This detection range is well suited for construction of two or
even more molecular layers of similar or dissimilar species on top of the support
conducting solid surfaces. From molecular adsorption’s perspective, the similarities
of the investigations on such molecular surfaces with traditional solid surfaces are
∗ National Center for Nanoscience and Technology, Beijing, 100080, China

C Springer 2008
134 C. Wang, S. Lei
conceptually identical, such as the nature of adsorption (either chemical or phys-

ical adsorptions), adsorption induced reconstruction, adsorption selectivity, diffu-
sion behavior etc. In essence, such molecular surfaces represent the category of
organically functionalized surfaces with the properties dominated by the constitut-
ing molecular building units. The main differences of the molecular surfaces from
the traditional solid surfaces are originated with the chemical and structural hetero-
geneity of the molecular building units, which also offers a rich ground for system-
atic investigations. The results on the surface properties of the molecular surfaces
should surely help enrich the scope of surface science studies by incorporating the
tremendous synthetic capabilities of organic-molecule-based building blocks, and
also provide much needed knowledge critical for achieving controlled designs of
molecular functional devices. This work is intended to present some results showing
the possible effects of molecular adsorption on above specified molecular surfaces
using STM. These results are derived from a limited number of systems and could
provide useful references for broadened scope of studies.
11.2 Construction of Molecular Surfaces for Heterogeneous

Molecular Adsorptions
The realization of molecularly decorated surfaces can be directly achieved by two-

dimensional assembling of molecular building units. The general principles for
assembling the organic molecules are themselves being extensively explored and
provide important guidelines for designing molecular surfaces with periodically
distributed functional groups. A wide range of molecular assemblies have been ana-
lyzed and the revealed results stimulated much interest in the driving mechanisms
behind molecular architectures (Rabe and Buchholz, 1991; Claypool et al., 1997;
Cyr et al., 1996; Gimzewski and Joachim, 1999; De Feyter and De Schryver, 2003).
The scope of undertaken investigations includes various conditions (ultra-high vac-
uum, ambient, solutions) and surfaces (metal, graphite, silicon, etc.).
The formation of well-ordered two-dimensional assembling structures is jointly
affected by intermolecular interactions and molecule-substrate interactions (Rabe
and Buchholz, 1991; Claypool et al., 1997; Cyr et al., 1996; Hentschke et al.,
1992). When multi-component mixtures are involved in the self-assembly process,
the weak interactions between self-assembled molecules are very important in the
assembling of supramolecular architectures. It should be noted that hydrogen bond
plays an important role in building supramolecular nanostructures with its advan-
tage of selectivity and directionality, (Lei et al., 2001; De Feyter et al., 2005) and is
widely used and studied in molecular self-assemblies. Examples for the application
of hydrogen bond could be found in the construction of the 2D networks of car-
boxyl derived porphyrin and copper phthalocyanine (CuPc), (Lei et al., 2001) and
also the coded assembling of donor-acceptor supramolecular assembly of perylene
bisimide and diamino triazine derivative. (De Feyter et al., 2005) Other important
interacting forces, i.e. van der Waals and electrostatic interactions, are also prevalent
in constructing molecular surfaces or networks. As an example, it is well-known
that ordered two-dimensional monolayers of alkanes and alkane derivatives can be
11 Towards Surface Science Studies Using Scanning Tunneling Microscopy 135
formed through physisorption on highly oriented pyrolytic graphite (HOPG) (Rabe

and Buchholz, 1991; Hentschke et al., 1992). The crystallization effect of linear
alkane chains due to van der Waals interactions is dominant in such systems. The
functional terminal groups of the alkane derivatives are periodically arranged within
the alkane lamella templates and form a chemically decorated molecular surface.
11.3 Selective Adsorption Behavior of Molecular Surfaces
The above-mentioned molecular surfaces have been shown to have a number of

effects, such as selective adsorption, modulated assembling, immobilization effect
of adsorbates, etc. With rich variety of functional groups and intermolecular inter-
actions, investigations on the novel effects of assembled molecular architectures
should be highly rewarding.
The reported work have identified that both physical and chemical adsorptions
could occur on molecular surfaces, depending on the functional groups of the con-
stituting molecules of the surface and adsorbates. A readily available example of
molecular surface can be found with linear alkane derivatives (such as stearic acid,
1-octadecanol and 1-iodooctaecane) which are known to form parallel lamella pat-
terns on graphite surface in ambient conditions. When pre-covered on the support
surfaces (such as graphite or metals), the molecular lamella structures introduce
inherently the heterogeneous adsorption sites and anisotropic diffusion barriers in
association with the functional groups. The presence of heterogeneous adsorption
sites could result in selective adsorption of single molecules. The adsorbed species
would also experience the anisotropic diffusion barrier and organize in a restricted
manner.
It was observed that single or clusters of copper phthalocyanine (CuPc)
molecules could be stably adsorbed on alkane derivatives serving as the support
surface (Xu et al., 2000; Lei et al., 2004a). In the case of single CuPc molecule
adsorption on the lamella structure formed by stearic acid and 1-iodooctadecane,
the Pc molecules were found to be exclusively located on top of the alkane part of
stearic acid and 1-iodooctadecane lamellae, and the location of functional groups
are not the preferred adsorption sites. Possibly due to the coherence in geometry,
pairs of CuPc molecules in close contact side-by-side is preferentially observed to
lie along the alkane chains (Fig. 11.1a).
In another type of molecular surface consisted of tridodecyl amine (TDA) lamel-
lae, both physical and chemical adsorptions were identified. Physical adsorption
behavior was observed for CuPc molecules that are again exclusively adsorbed at the
alkane part of the tridodecyl amine (TDA) lamellae. At low adsorbate concentration
molecule pairs of CuPc appear again on the TDA template (Fig. 11.1b). However,
different from the case on 1-iodooctadecane, the CuPc molecules in the molecular
pair are separated by the amine groups. The distance between CuPc centers is mea-
sured to be nearly 1 nm. Though the dominating interaction between adsorbates and
molecular surface is van der Waals and therefore can be characterized by physical
adsorption, the formation of molecular pairs separated by the amine groups indicate
the interaction between the CuPc molecule and the amine group plays some role in
136 C. Wang, S. Lei
Fig. 11.1 Isolated single CuPc molecule and molecular pairs observed on top of the alkane part of
1-iodooctadecane (a) and TDA (b). A double-molecule band of CuPc modulated by the TDA layer
has been observed as the increase of CuPc coverage (c). Image condition, a: −772 mV, 761 pA;
b: 990 mV, 320 pA; c: 724 mV, 876 pA
the formation of such CuPc dimers. It was further revealed that with the increase of
adsorbate coverage, a complete overlayer of CuPcs will be modulated by the peri-
odicity of the underneath TDA lamellae (Lei et al., 2004b). Such modulation effect
demonstrates the impact of periodicity of molecular surface (in this specific study,
alkane derivative lamellae) on the adsorption, diffusion and assembling of adsorbed
molecules. One thing worth noticing is the submolecular resolution obtained on
such isolated single molecules and clusters. Normally for STM imaging on the
solid/gas or solid/liquid interface under ambient conditions submolecular resolution
could only be obtained on well ordered crystalline domains. The high resolution
achieved on the isolated molecules on the modulated surface indicates high stability
of these molecules. This stability could probably be attributed to the increased
diffusion barrier for the CuPc molecules on the modulated surface.
The example of chemical adsorption can be found in the hydrogen bond assisted
adsorption of benzoic acid molecules exclusively at the location of amine groups
in the TDA assembly (Lei et al., 2003). TDA forms well ordered lamellae structure
when deposited on the surface of graphite (Fig. 11.2a). High-resolution STM image
reveals that TDA molecules adsorb with their alkyl chains parallel to the basal plane
Fig. 11.2 TDA assembly observed without (a) and with (b) the presence of benzoic acid. The
amine groups appear with dark contrast in the image without benzoic acid, but when benzoic acid
is added into the solution, isolated bright spots appear in the position of the amine groups. Image
condition: a: 1.0 V, 426 pA; b: 1.04 V, 14.6 pA
11 Towards Surface Science Studies Using Scanning Tunneling Microscopy 137
of graphite with a 90◦ angle to the lamella axis. The amine groups show reduced
contrast with respect to the alkane part of the molecule. When benzoic acid was
added, bright spots with diameter of about 0.6 nm appears along the center of the
TDA lamellae (Fig. 11.2b). The diameter of these bright spots is in agreement with
the dimension of single benzene rings. Compared with the observed lamellae in the
uncovered area, these bright spots could be attributed to benzoic acid molecules over
the sites of the amino groups in the TDA lamellae, possibly through the O−H−N
hydrogen bond. It was also observed that the adsorbed benzoic acids on the surface
could be easily removed by tip scanning, indicating the linkage between the benzoic
acid and amine group is weak. Another example of chemical adsorption can be
found in the urea molecules adsorption on the lamella of double-alkyl amino acid
(Hoeppener et al., 2003), in which the unprotected amino acid groups were found to
be the preferential adsorption sites for urea molecules.
11.4 Summarizing Remarks and Future Perspectives
The above discussed examples illustrate the possibilities of adopting molecular

surfaces as an important category of support medium. They are ideally suited as
model systems for mechanistic studies of molecular based devices. The structural
resolution capability of STM can be optimized, in combination with molecular
self-assembling processes, to yield insightful knowledge on the interface between
dissimilar molecular layers. It is conceivable that with the tunneling spectroscopic
capability of STM, one could further pursue the electronic characteristics of the
heterogeneous molecular interfaces. In addition, the vast possibility of organizing
chemical groups in 2D molecular assemblies would also provide much unexplored
novel effects of heterogeneous molecular interfaces.
References
Claypool, C. L., Faglioni, F., Goddard III, W. A., Gray, H. B., Lewis, N. S., Marcus, R. A., 1997,
J. Phys. Chem. B 101, 5978.
Cyr, D. M., Venkataraman, B., Flynn, G. W., Black, A., Whitesides, G. M., 1996, J. Phys. Chem.
100, 13747.
De Feyter, S., De Schryver, F. C., 2003, Chem. Soc. Rev. 32, 139.
De Feyter, S., Miura, A., Yao, S., Chen, Z., Wurthner, F., Jonkheijm, P., Schenning, A. P. H. J.,
Meijer, E. W., De Schryver, F. C., 2005, Nano letts. 5, 77.
Gimzewski, J. K., Joachim, C., 1999, Science 283, 1683.
Hentschke, R., Schurmann, B. L., Rabe, J. P., 1992, J. Chem. Phys. 96, 6213.
Hoeppener, S., Wonnemann, J., Chi, L. F., Erker, G., Fuchs, H., 2003, Chem Phys Chem. 4, 490.
Lei, S. B., Wang, C., Wan, L. J., Bai, C. L., 2003, Langmuir 19, 9759.
Lei, S. B., Wang, C., Yin, S. X., Wang, H. N., Xi, F., Liu, H. W., Xu, B., Wan, L. J., Bai, C. L.,
2001, J. Chem. Phys. B., 105, 10838.
Lei, S. B., Yin, S. X., Wang, C., Wan, L. J., Bai, C. L., 2004a, J. Phys. Chem. B 108, 224.
Lei, S. B., Wang, C., Wan, L. J., Bai, C. L., 2004b, J. Phys. Chem. B 108, 1173.
Rabe, J. P., Buchholz, S. 1991, Science 253, 424.
Xu, B., Yin, S. X., Wang, C., Qui, X. H., Zeng, Q. D., Bai, C. L., 2000, J. Phys. Chem. B
104, 10502.
Chapter 12
Electronic Transport Through Metal
Nanowire Contacts
Y. H. Lin, K. J. Lin, F. R. Chen, J. J. Kai and J. J. Lin∗
12.1 Introduction
Nanoscale structures have recently become the subject of intense theoretical and
experimental investigations because of their importance for both fundamental
researches and potential industrial applications. Among the various nanostructures,
self-assembled quasi-one-dimensional metallic nanowires are of particular interest,
primarily due to their rich and fascinating electrical properties (see, for example,
Lin et al., 2003; Liu et al., 2007). To explore how electrons transport through these
quasi-one-dimensional systems, nanofabrication techniques such as the electron-
beam lithography and the focused ion beam deposition are often employed to this
end. Nevertheless, it is known that fabrication of reliable electronic contacts to
individual nanostructures has been a nontrivial issue, since very often an imperfect
contact inevitably forms and may possess a non-negligible temperature dependent
contact resistance, Rc , which in turn can seriously complicate the measurements.
In the cases involving semiconductors, the contact between a metal electrode and
a semiconductor nanowire usually forms a Schottky barrier, whose properties have
been extensively studied (Ip et al., 2004). A different situation involving heavily
doped semiconductor nanowires has also been discussed (Gu et al., 2001). On the
other hand, the electrical properties of the contact between a metal electrode and a
metal nanowire have not yet been much addressed in the literature. In the case of
metal nanowires, since the as-grown nanowires may readily be covered by a layer
of some insulators (oxidations, contaminations, amorphous coating, etc.) of a few
nm thick, the underlying physics of how electron waves transmit through such a
nanoscale interface is of fundamental interest and urgent industrial concerns.
In this work, we report the temperature behaviour down to liquid-helium tem-
peratures of high-resistance electronic contacts formed at the interfaces between
lithographic-patterned submicron electrodes and single metal nanowires. The con-
tact resistances Rc (T )’s were determined through the electrical measurements on
individual metal nanowires by using the 4-probe, together with the 2-probe, con-
figurations as schematically depicted in Fig. 12.1a–c. In the 4-probe configuration
∗ Institute of Physics and Department of Electrophysics, National Chiao Tung University, Hsinchu
30010, Taiwan e-mail: [email protected]

C Springer 2008
140 Y. H. Lin et al.
Fig. 12.1 (a) Schematic diagram for a single nanowire (thick bar) with four electronic contacts on
it and the equivalent circuit model. Rs denotes the resistance of each segment of the nanowire
(i.e., the sample resistance). Rc1 , Rc2 , Rc3 and Rc4 denote the electronic contact resistances
between the nanowire and the submicron electrodes 1, 2, 3 and 4, respectively. Rel,i denotes
the submicron-electrode resistance of the ith electrode. Four-probe (b) and two-probe (c) mea-
surement configurations, and the corresponding measured resistances. In (c), the approximations
Rc2 ≈ Rc3 ≡ Rc and Rel,2 ≈ Rel,3 ≡ Rel are made (see text)
(Fig. 12.1b), the intrinsic nanowire sample resistance, Rs , can be accurately deter-
mined. In the 2-probe configuration (Fig. 12.1c), the measured resistance is the sum
of Rs plus the two contact resistances and the two submicron-electrode resistances,
i.e., R(T ) = Rs (T ) + 2Rc (T ) + 2Rel (T ), where Rel is the submicron-electrode
resistance. The prefactor 2 is introduced to denote that there are two similar contacts
(submicron electrodes) in series in the 2-probe method. Here we have assumed the
simple case that the two electronic contacts are (almost) symmetric, because they
were fabricated simultaneously under the same conditions on the same nanowire.
(Experimentally, the resistance of each electronic contact can be quantitatively
extracted by further applying the 3-probe measurement configuration in addition to
the 2-probe measurement configuration, as will be described elsewhere (Lin et al.,
2007).) In the case of high-resistance contacts, i.e., R >> Rs and R >> Rel , the Rc
may dominate the measured resistance and the approximation R ≈ 2Rc is valid for
the whole range of experimental temperature. The measured magnitude and temper-
ature behaviour of R thus faithfully reflect the magnitude and temperature behaviour
of Rc . As a result, Rc (T ) can be quantitatively inferred. In the present work, single-
crystalline RuO2 nanowires, which exhibit metallic conductivity comparable with
that of normal metals (Ryden et al., 1970; Lin et al., 2004) and at the same time
12 Electronic Transport Through Metal Nanowire Contacts 141
possess excellent thermal and chemical stability (Trasatti and Lodi, 1981; Trasatti,
1991), were chosen as our sample system to substantiate this idea.
12.2 Experimental Method
Self-assembled single-crystalline RuO2 nanowires were synthesized by the thermal

evaporation method based on the vapour-liquid-solid growth mechanism (Wagner
and Ellis, 1964). The as-grown nanowires were characterized using field-emission
scanning electron microscopy (FE-SEM), transmission electron microscopy (TEM),
x-ray diffraction (XRD), and x-ray energy dispersive spectroscopy (XR-EDS), as
described previously (Liu et al., 2007).
Electrical contacts onto the nanowires were fabricated as described below. A
Si substrate with a ≈ 200-nm thick SiO2 layer grown on top was first photo-
lithographically patterned with Cr/Au (≈ 10/60 nm) “macro-electrodes” using bi-
layer photoresist process to create reverse-slope resist sidewall profiles. Several
droplets of dispersed alcoholic solution containing RuO2 nanowires were dropped
on the substrate. Following the standard electron-beam lithography process, a sin-
gle nanowire was positioned with SEM, and “submicron electrodes” contacting the
nanowire were generated by thermal evaporation of Cr/Au (≈ 10/90 nm). Fine Cu
leads were attached to the macro-electrodes with silver paste, and the substrate was
thermally anchored to the sample holder on a standard 4 He cryostat. Depending on
the different probe configurations employed, the resistances were measured by two
different manners. For the 4-probe method, a Linear Research LR-700 AC resis-
tance bridge was applied. For the 2-probe method, the resistances were measured
by utilizing a Keithley K-6430 source meter as a current source and a K-182 nano-
voltmeter, in which the DC current-reversal method was adopted so that any existing
thermoelectric voltages along the measurement loop were cancelled. We notice that
for all of the R(T ) curves reported in this work, a measurement current of 10 nA was
applied and the current-voltage linearity had been checked and ensured at various
temperatures.
To ensure the metallic nature of our single-crystalline RuO2 nanowires, the resistiv-
ities ρ as a function of temperature from 300 K down to liquid-helium temperatures
for individual as-grown nanowires were first determined with the 4-probe measure-
ment configuration. Several RuO2 nanowires with different diameters ranging from
a few tens to about 150 nm have been studied. Figure 12.2 shows the result for a rep-
resentative nanowire with a diameter of ≈ 55 ± 5 nm. (The diameter of the nanowire
was determined by AFM.) Clearly, the nanowire reveals electrical-transport char-
acteristic of a typical metal, i.e., the resistivity decreases with decreasing temper-
ature. Inspection of Fig. 12.2 indicates that the resistivity ρ(300 K) ≈ 240 μ cm
Fig. 12.2 Resistivity as a function of temperature for a RuO2 nanowire with diameter of
≈ 55 ± 5 nm. The inset shows the SEM image, where the scale bar is 1 μm. Notice that the
length of the nanowire used for the evaluation of the resistivity from the measured resistance is
taken to be the distance between the middles of the two voltage probes
in this nanowire is relatively higher than that (≈ 85 ± 35 μ cm) in bulk single

crystals (Ryden et al., 1970; Lin et al., 2004). On the other hand, the resistivity
ratio ρ(300 K)/ρ0 ≈ 1.6 for this nanowire is considerably lower, as compared with
the corresponding bulk value (≈ 100–1000, depending on the quality of the crystal
(Ryden et al., 1970; Lin et al., 2004)), where ρ0 is the residual resistivity. In fact, we
have found ρ(300 K) ≈ 150–280 μ cm and ρ(300 K)/ρ0 ≈ 1.5–1.7 in a number
of RuO2 nanowires. Such high resistivity values and low resistivity ratio values
imply the presence of a high level of defects (most likely, point defects) in the as-
grown nanowires. This observation is in sharp distinction to the conclusion drawn
from conventional materials characterization techniques, such as XRD and high-
resolution TEM (Liu et al., 2007), which often suggested very high quality atomic
structure. Similar phenomenon had also been found in single-crystalline metallic
NiSi nanowires (Wu et al., 2004). This observation suggests that the electrical-
transport measurements over a wide range of temperature can serve as a powerful
probe for the presence of microscopic defects in nanostructures. For instance, the
electron mean free path, which is highly sensitive to the existence of any level of ran-
domness in the crystal, can be quantitative determined from the ρ(T ) measurements.
On the other hand, with the metallic nature of our individual RuO2 nanowires being
established, we are safe to apply these nanowires to quantitatively explore the elec-
tronic contact resistances Rc (T ) in lithographic-contacting nanodevices, employing
the 2-probe measurement configuration discussed above.
Ideally, the lead resistances and the electronic contact resistances in electrical
measurements must be small to minimize thermal noises. In our case, the elec-
tronic contact resistances formed between the electron-beam lithography patterned
electrodes and the RuO2 nanowires normally fall between several tens and several
hundreds , and are not much dependent on temperature. However, highly resis-
tive electronic contacts with room temperature resistances of order several k or
higher may also be obtained in many fabrications. Figure 12.3 shows typical tem-
perature behaviour of Rc for such a high-resistance contact from 300 K down to
liquid-helium temperatures, as determined from the 2-probe method. In strong con-
trast to the 4-probe result shown in Fig. 12.2, now the measured resistance reveals
semiconducting or insulating behaviour, i.e., the resistance increases rapidly with
decreasing temperature. As discussed, in this case R(T ) ≈ 2Rc (T ), since the two
submicron Cr/Au electrodes and the RuO2 nanowire have resistances much lower
than the measured value. (The total Cr/Au electrode resistance is typically ≈ a few
tens , and the RuO2 nanowire resistance is typically ≈ a few hundreds .) The
large contact resistance (∼ several tens k) may result from a thin, dirty insulating
layer incidentally formed at the interface between the submicron electrodes and the
nanowire. This insulating layer could be the lightly contaminated or oxidized metals
introduced during the electrode evaporation, the amorphous coating resulting from
the complex growth process, the vacancies caused by dramatic surface roughness
near the contact region, or the breaking induced by tensile stress.
In view of the fundamental researches and industrial applications of nanoscale
structures, it is very important to understand the behaviour and the underlying
physics of the electronic contacts formed between interconnects and the metal
nanostructure. The inset to Fig. 12.3 shows a plot of a representative variation
of log R with T −1 for one of our contact resistances. This figure indicates that
the simple thermal activation conduction (the straight solid line) is only respon-
sible near room temperatures. As the temperature reduces from room temperature,
the resistance does not increase as fast as would be expected from the thermally
activated process. At liquid-helium temperatures, the resistance appears roughly
Fig. 12.3 Resistance as a function of temperature for two high-resistance contacts in series,
as determined from the 2-probe measurement configuration on the nanowire device C3 (see
Table 12.1). The inset shows a plot of log R versus T −1 for the same nanowire device. The straight
solid line indicates the thermal activation conduction
constant, i.e., independent of temperature. Such a temperature independent resis-

tance at liquid-helium temperatures can signify a conduction mechanism character-
istic of simple elastic tunnelling. Indeed, quantitative analysis (see below) indicates
that the overall temperature behaviour of the contact resistance can be well inter-
preted in terms of a tunnelling conduction model. Previously, in order to explain the
electrical-transport properties in certain classes of granular metal-dielectric com-
posites, Sheng and co-workers (1978) and Sheng (1980) have proposed a thermally
“fluctuation-induced tunnelling” model, where the thermal effects arising from the
capacitance C of a small junction formed between two large metal grains was con-
sidered.
According to Sheng and co-workers (1978) and Sheng (1980) the temperature
dependent resistance for small applied electric fields across a single small junction
can be expressed as

T1
R(T ) = R0 exp (12.1)
T0 + T
where R0 is parameter which depends only weakly on temperature, and T1 and T0

are characteristic temperatures defined as

8ε0 AV02
T1 = 2 (12.2)
e kB w
and
3/2

16ε0 AV0
T0 = (12.3)
π (2m)1/2 e2 k B w2
where ε0 is the vacuum permittivity, is the Planck’s constant divided by 2π,

and m is the electron mass. In Eq. (12.1), T1 can be regarded as a measure of
the energy required for an electron to cross the potential barrier between the two
conducting regions, and T0 is the temperature below which the fluctuation effects
become insignificant, since, at T << T0 , Eq. (12.1) is temperature independent and
reduces to an expression for the expected simple elastic tunnelling. In the deriva-
tion of Eq. (12.1), the conduction was first modelled (Sheng et al., 1978) as the
tunnelling of electrons through a single potential barrier of width w, height V0 , and
junction area A. (Here A is the size at the point of the two large conducting regions’
closest approach.) If A is small enough, it was found (Sheng et al., 1978; Sheng,
1980) that the potential barrier seen by the electrons could be effectively narrowed
and lowered by the thermal voltage fluctuations (≈ (k B T /C)1/2 , where k B is the
Boltzmann constant) across the insulating gap due to the small effective capacitance
of the junction. Such a potential-barrier modulation effect greatly influences the
tunnelling probability in the low temperature limit, and consequently introduces a
characteristic temperature behaviour to the normally temperature independent tun-
nelling conductivity (i.e., the elastic tunnelling regime).
In the case of granular composites of macroscopic sizes, it was then argued

(Sheng, 1980), via the effective-medium theory, that in a network of independently
fluctuating tunnel junctions with different values of T1 and T0 , the conductivity of
the network could still be well described in terms of a single junction with a repre-
sentative set of T1 and T0 .
Figure 12.4 shows a plot of our experimental results in double logarithmic scale
for four representative NW devices having high contact resistances (as determined
from the 2-probe method). The symbols are the experimental data and the solid
curves are the least-squares fits to Eq. (12.1), with R0 , T1 and T0 as the adjust-
ing parameters. Inspection of Fig. 12.4 clearly indicates that the Eq. (12.1) can
well describe the overall temperature behaviour for a wide range of temperature
between 2 and 300 K. The values of T1 and T0 can then be reliably extracted. Fur-
thermore, by using SEM and/or AFM, we can directly measure the diameter and
the length of our nanowires, as well as the width of the relevant submicron Cr/Au
electrodes overlying the nanowire. Therefore, the junction area A which appeared
in Eqs. (12.2) and (12.3) is independently determined. (In this work, the junction
area A is given by the product of the nanowire diameter and the width of the over-
lying submicron electrode.) With the values of T1 , T0 and A being determined,
the microscopic parameters characterizing the electronic contacts, i.e., the width
w and height V0 of the potential barrier, may then be inferred. Our experimental
values of the relevant parameters are listed in Table 12.1. Notice that, in Table 12.1,
the measured resistance R(300 K) for each nanowire device is at least an order of
magnitude higher than the intrinsic resistance of the nanowire, Rs (300 K), justifying
our approximation R(T ) ≈ 2Rc (T ). Furthermore, the resistance ratio R(T )/Rs (T )
Fig. 12.4 Double logarithmic plot of the resistances versus temperature for four high-resistance
nanowire devices, as determined from the 2-probe measurement configuration. The symbols are
the experimental data and the solid curves are the theoretical fits to Eq. (12.1)
Table 12.1 Values of relevant parameters for four high-resistance nanowire devices, as determined
from the 2-probe measurement configuration. For each device, the nanowire resistance Rs (300 K)
was estimated from the 4-probe measurement configuration, while the junction area A was deter-
mined from the SEM image (see text) .
R (300 K) Rs R0 (k) T1 (K) T0 (K) A (μm2 ) w (nm) V0 (meV)
(k) (300 K)
(k)
C1 241 0.67 112 316 91 0.023 6.6 4.2
C2 33 0.33 13.6 363 104 0.022 6.4 4.5
C3 10.3 0.31 8.08 87 37 0.040 7.0 1.7
C4 3 0.33 2.82 20 8.6 0.028 8.7 1.1
increases rapidly as the temperature decreases below room temperature, making this
approximation greatly satisfied.
It should be noted that, in the fluctuation-induced tunnelling model, since the
two conducting regions remain large in size, the charging energy E c needed to
transfer an electron from one conducting region to the other is completely negli-
gible, i.e., E c << k B T . This situation is very different from that in the case of
the Coulomb blockade which involves fine metal grains or quantum dots, where the
charging energy E c (>> k B T ) rather than the thermal voltage fluctuations plays
the crucial roles in governing the electronic transport properties. In the present
work, the volume of our “long” nanowire is relatively large as compared to the
sizes of the fine metal grains (for example, ∼ 103 nm3 ) used in the Coulomb
blockade studies (Tinkham et al., 2000). The typical volume of our nanowires
is ∼ 100 nm × 100 nm × 3 μm. Thus, our nanowire can be envisioned as a large
conducting region separated by an insulating layer from another large conducting
region (the submicron Cr/Au electrode) with a junction area A (∼ 100 × 500 nm2 ).
This size of A is already small enough to render the aforementioned thermal volt-
age fluctuations important while large enough to make the Coulomb blockade phe-
nomenon irrelevant. In effect, our nanowire devices in the 2-probe configuration
mimic two similar tunnel junctions in series in the context of the fluctuation-induced
tunnelling model.
Previously, the fluctuation-induced tunnelling model has been successfully
applied to explain the temperature behaviour of the resistances in, among others,
carbon polyvinylchloride composites (Sheng et al., 1978; Sichel et al., 1978, 1981),
polymer composites (Paschen et al., 1995), and tin-doped indium oxide thin films
(Ederth et al., 2003). In those “macroscopic” composite samples, a very large
number of tunnel junctions with barely known junction parameters were often
involved. On the contrary, the situation is greatly simplified and straightforward
in our case, because in the 2-probe measurement configuration we deal with only
two electronic contacts characterized by similar junction parameters, as discussed.
Moreover, our junction area A is known. Interestingly, our experimental values of
w and V0 listed in Table 12.1 are on the same orders of magnitude to those obtained
in carbon polyvinylchloride composites (Sheng et al., 1978; Sichel et al., 1978,
1981). This coincidence may be due to the fact that the sizes of our nanowires
are approximately the same to the mean size of the conducting chains found
in those composites.
Finally, it is worth noting that, if in our electronic contacts, the effective junction
area is somewhat reduced from the maximum possible area A defined above, our
values of the barrier width w would be slightly decreased, while values of the barrier
height V0 slightly increased, from those listed in Table 12.1.
12.4 Conclusion
We have measured the temperature dependent resistivities of individual single-

crystalline RuO2 nanowires, using both the 4-probe and the 2-probe methods. In
the 4-probe measurements, the resistance reveals metallic behaviour, as expected.
By employing the 2-probe method, we have quantitatively characterized the elec-
tronic contact resistances formed at the interfaces between the submicron electrodes
and the nanowires. We found that the temperature behaviour of the high-resistance
contacts can be well attributed to the thermally fluctuation-induced tunnelling con-
duction. The junction parameters such as the barrier width and height have been
determined. This work demonstrates that, under certain conditions, the electronic
contacts between interconnects and metal nanodevices can be quantitatively mod-
elled. In further studies, the junction parameters w, V0 , and A may be tailored by
adjusting the diameter and surface condition of the nanowire, the electrode material,
and the width of the electrodes, etc. Finally, we should point out that the electronic
contact resistance is a quantity very sensitive to the interface condition and, in differ-
ent classes of nanowires and electrodes, the temperature behaviour may be governed
by a mechanism different from the fluctuation-induced tunnelling conduction (Lin
et al., 2007).
Acknowledgments The authors are grateful to T. C. Lee, S. P. Chiu, S. J. Hong, Y. L. Liu and J.
J. Huang for their experimental help, and to Ping Sheng for valuable discussion. This work was
supported by the Taiwan National Science Council through Grant Nos. NSC 93-2120-M-009-009
and NSC 94-2120-M-009-010, and by the MOE ATU Program.
References
Ederth, J., Johnsson, P., Niklasson, G. A., Hoel, A., Hultåker, A., Heszler, P., Granqvist, C. G.,
van Doorn, A. R., Jongerius, M. J. and Burgard, D., 2003, Electrical and optical proper-
ties of thin films consisting of tin-doped indium oxide nanoparticles. Physical Review B, 68,
155410.
Gu, G., Burghard, M., Kim, G. T., Düsberg, G. S., Chiu, P. W., Krstic, V., Roth, S. and Han, W. Q.,
2001, Growth and electrical transport of germanium nanowires. Journal of Applied Physics,
90, 5747.
Ip, K., Heo, Y. W., Baik, K. H., Norton, D. P., Pearton, S. J., Kim, S., LaRoche, J. R. and Ren, F.,
2004, Temperature-dependent characteristics of Pt Schottky contacts on n-type ZnO. Applied
Physics Letters, 84, 2835.
Lin, J. F., Bird, J. P., Rotkina, L. and Bennett, P. A., 2003, Classical and quantum transport in
focused-ion-beam-deposited Pt nanointerconnects. Applied Physics Letters, 82, 802.
Lin, J. J., Huang, S. M., Lin, Y. H., Lee, T. C., Liu, H., Zhang, X. X., Chen, R. S. and Huang,
Y. S., 2004, Low temperature electrical transport properties of RuO2 and IrO2 single crystals.
Journal of Physics: Condensed Matter, 16, 8035.
Lin, Y. H. et al, 2007, to be published.
Liu, Y. L., Wu, Z. Y., Lin, K. J., Huang, J. J., Chen, F. R., Kai, J. J., Lin, Y. H., Jian, W. B. and
Lin, J. J., 2007, Growth of single-crystalline RuO2 nanowires with one- and two-nanocontact
electrical characterizations. Applied Physics Letters, 90, 013105.
Paschen, S., Bussac, M. N., Zuppiroli, L., Minder, E. and Hilti, B., 1995, Tunnel junctions in a
polymer composite. Journal of Applied Physics, 78, 3230.
Ryden, W. D., Lawson, A. W. and Sartain, C. C., 1970, Electrical transport properties of IrO2 and
RuO2 . Physical Review B, 1, 1494.
Sheng, P., 1980, Fluctuation-induced tunneling conduction in disordered materials. Physical
Review B, 21, 2180.
Sheng, P., Sichel, E. K. and Gittleman, J. I., 1978, Fluctuation-induced tunneling conduction in
carbon-polyvinylchloride composites. Physical Review Letters, 40, 1197.
Sichel, E. K., Gittleman, J. I. and Sheng, P., 1978, Transport properties of the composite material
carbon-poly(vinyl chloride). Physical Review B, 18, 5712.
Sichel, E. K., Sheng, P., Gittleman, J. I. and Bozowski, S., 1981, Observation of fluctuation modu-
lation of tunnel junctions by applied ac stress in carbon polyvinylchloride composites. Physical
Review B, 24, 6131.
Tinkham, M., Davidovi, D., Ralph, D. C. and Black, C. T., 2000, Tunneling through metallic quan-
tum dots. Journal of Low Temperature Physics, 118, 271.
Trasatti, S., 1991, Physical electrochemistry of ceramic oxides. Electrochimica Acta, 36, 225.
Trasatti, S. and Lodi, G., 1981, Oxygen and chlorine evolution at conductive metallic oxide anodes.
In Electrodes of Conductive Metallic Oxides, Part B, edited by Trasatti, S., (Elsevier), chap. 10.
Wagner, R. S. and Ellis, W. C., 1964, Vapor-liquid-solid mechanism of single crystal growth.
Applied Physics Letters, 4, 89.
Wu, Y., Xiang, J., Yang, C., Lu, W. and Lieber, C. M., 2004, Single-crystal metallic nanowires and
metal/semiconductor nanowire heterostructures. Nature, 430, 61.
Chapter 13
Synthesis and Properties
of Quasi-One-Dimensional Nitride
Nanostructures
Yong-Bing Tang, Dai-Ming Tang, Chang Liu, Hong-Tao Cong∗

and Hui-Ming Cheng∗
13.1 Introduction
One-dimensional (1D) nanomaterials have attracted much attention due to their

novel structures, unique physical and chemical properties and wide potential appli-
cations (Xia et al., 2003). With the progress in synthesis of 1D nanostructures such
as nanotubes (Iijima, 1991), nanowires (Morales and Lieber, 1998) and nanobelts
(Pan et al., 2001), their property investigation and practical application are becom-
ing more and more important. Due to their low values of electron affinity, excel-
lent optical properties, mechanical properties and thermo-physical properties, the
III-V group semiconducting nitrides such as AlN, BN and GaN, are promising
electronic materials for field emitters, ultraviolet and visible light emitters, piezo-
electric devices, etc. The 1D nitride nanostructures are expected to exhibit better
properties than the conventional polycrystalline and single-crytalline nitride bulk
materials because of their nanosize and high aspect ratio. Due to the small or even
negative electron affinity of the nitrides, electrons can be easily extracted from sur-
face to vacuum under electric field (Benjamin et al., 1994; Grabowski et al., 2001).
Furthermore, the advantage of 1D nanostructures is obvious from the field emission
(FE) point of view because high aspect ratio and nanosized tips can generate an
extremely high local electric field and thereby result in a large FE current in a low
applied field. Except for field emission (FE) applications, AlN and BN also possess
many other useful properties such as high thermal conductivity, excellent thermal
and chemical stabilities, and low thermal expansion rates. (Ronning et al., 1998;
Ward et al., 1998; Sugino et al., 2001)
Recently, visible light emissions are observed in 1D nanostructures of many
materials such as ZnS (Liang et al., 2004), MgO (Zhang and Zhang, 2002), In2 O3
(Jeong et al., 2004), but these emissions are different from those of their bulk mate-
rials. Some researchers thought that it is mainly because these 1D nanostructures


C Springer 2008
150 Y.-B. Tang et al.
have so large specific surface area that it is easy to be doped with impurities of some
elements and be introduced into structural defects (Zhang and Zhang, 2002; Jeong
et al., 2004). The III-V group nitrides recently become very attractive in applications
for light emitters, laser diodes, and optoelectronic devices in the ultraviolet and
visible spectral range (Nakamura, 1998), because the wurtzite nitrides can form a
continuous alloy system with adjustable direct band gaps from 6.2 to 1.9 eV. Hence,
ultimate nanoscale optoelectronic devices for a wide range of wavelengths with
many advanced features such as high thermal conductivity and superior stability
may be realized by using these 1D nanostructures. To date, intensive studies have
been devoted to the synthesis, properties and prototypes of 1D III-V group nitrides
(Haber et al., 1997, 1998; Liu et al., 2001, 2005b; Zhang et al., 2001; Yin et al.,
2005a, b; Tondare et al., 2002; Tang et al., 2005; Wu et al., 2003, 2004; Zhao
et al., 2005; Shi et al., 2005). The nitrides series possess wide-range band gaps, are
therefore ideal for the development of solid-state white light emitting devices (Vur-
gaftman et al., 2001). AlN is the only wurtzite compound that has been predicted to
have a negative crystal field splitting at the top of valence band which can lead to
unusual optical properties (Suzuki et al., 1995; Wei and Zunger, 1996; Chen et al.,
1996). An AlN PIN (p-type/intrinsic/n-type) homojunction LED with an emission
of the shortest wavelength of 210 nm has been successfully developed (Tanniyasu
et al., 2006). In a word, IIIA group nitride semiconductors are potential materials for
ultraviolet LEDs and laser diodes, but suffer from difficulties in controlling electri-
cal conduction to achieve exciton-related light-emitting devices as well as replacing
gas light sources with solid-state light sources.
With the rapid development of microelectronics, electronic packaging materials
have become vital to the success of advanced designs. AlN and BN are expected
to be excellent reinforcements in composites for electronic packaging or structural
materials because of their good mechanical properties, high thermal conductivity,
low thermal expansion (CTE) of AlN (∼3.3 × 10−6/K) (Geiger and Jackson, 1989)
is close to that of Si and GaAs, and low density. (Bradshaw and Spicer, 1999; Huang
and Li, 1994) Moreover, with high chemical and thermal stability, AlN and BN
do not react with Al, thereby avoiding the formation of brittle reaction products
on the reinforcement/matrix interface. In fact, AlN micro- or nano-particles have
been applied as reinforcement for metal and polyamide matrix composites because
of the above characteristics. (Huang and Li, 1994; Lai and Chung, 1994; Inoue
et al., 1993; Chedru et al., 2001; Vicens et al., 2002; Zhang et al., 2003; Chen and
Gonsalves, 1997; Zhang et al., 2002; Wu et al., 2003) AlN nanowires, compared
with conventional powder-form AlN, are predicted to exhibit better mechanical and
thermal properties (Haber et al., 1998; Zhang et al., 2001) and better reinforcing
performance due to their high aspect ratio and near-perfect crystal structure (Haber
et al., 1998; Zhang et al., 2001; Liu et al., 2001, 2005). Combining Al with 1D AlN
nanostructures may create novel nanocomposites with high strength and thermal
conductivity, as well as low density and CTE. Although nitride nanowires or nan-
otubes were synthesized in the past years, (Haber et al., 1998; Zhang et al., 2001; Liu
et al., 2001, 2005b; Yin et al., 2005a, b; Tondare et al., 2002; Tang et al., 2005; Wu
et al., 2003, 2004; Zhao et al., 2005; Shi et al., 2005) their reinforcement behaviors
in composites have not been investigated yet.
13 Synthesis and Properties of Quasi-One-Dimensional Nitride Nanostructures 151
It is known that many unique properties of materials depend on their structural

characteristics, which has stimulated the on-going interest in synthesizing various
nanostructures with different morphologies. Considerable effort has been made on
the synthesis of 1D nitride nanostructures as well. 1D nitride nanostructures such as
nanotubes, nanowires, and nanorods are successfully synthesized by several routes,
including chloride-assisted growth (Haber et al., 1998; Liu et al., 2005), carbon
nanotube-confined reaction (Zhang et al., 2001, Liu et al., 2001), arc-discharge
(Tondare et al., 2002; Tang et al., 2005c), direct nitridation (Wu et al., 2003a, b; Zhao
et al., 2005), and vapor transport and condensation (VTC) (Yin et al., 2005). Despite
of the recent progress in preparing 1D nitride nanostructures such as nanotubes (Yin
et al., 2005b; Tondare et al., 2002), nanowires (Haber et al., 1998; Zhang et al., 2001;
Liu et al., 2001, 2005), nanobelts (Tang et al., 2006), and comb-like morphologies
(Yin et al., 2005a) by different methods, the controllable synthesis or regulation of
these nanostructures is still a remained challenge, and the growth mechanism of 1D
nitrides is not yet clear.
In this study, high purity nitride nanostructures, including BN nanotubes and
nanofibers, AlN nanorods, nanotips, nanobelts and nanofibers are synthesized
(Tang et al., 2005, 2006a, 2007; Liang and Che, 1993). Especially, based on the
understanding of the growth mechanism, single-crystalline AlN mushroom-like
nanorod array, Eiffel-tower-shape nanotip array, honeycomb-like network of verti-
cally aligned nanoplatelets and flowerlike Si-doped AlN nanoneedle array have been
intentionally synthesized. Moreover, the field emission properties, optical proper-
ties, mechanical properties and thermo-physical properties of these nanostructures
and their composites are investigated.
13.2 Experimental
Most of the 1D nitride nanostructures were synthesized using a chemical vapor

transport and condensation (CVTC) method. The synthesis was carried out in a
conventional electric resistance furnace with a horizontal alumina tube. In a typical
process, 1.5 g Al (99.5%) nanoparticles (∼80 nm in size) was used as raw material
and placed in a ceramic boat which was set at the center of the alumina tube. A
Si wafer was put at the downstream end of the ceramic boat with a separation of
1.0 cm from raw reactants in order to collect the grown product. Argon gas was kept
flowing into the alumina tube at a rate of 200 sccm to minimize any side reactions
before the furnace was heated to the reaction temperature. Some metals or oxides
were selectively used to be raw materials or catalyst. The Al source was maintained
at certain temperatures and NH3 was introduced at a rate of 80 sccm for 1hour.
Then the furnace was cooled down to room temperature under flowing NH3 , a layer
of light-gray product was found on the Si wafer, the ceramic boat and the inner
wall of downstream-end tube. The product was collected from the Si wafer for later
characterization.
A vacuum apparatus that can produce a plasma electric arc within a big reactor
was employed for the synthesis of AlN nanorod arrays. The stainless steel reactor
chamber has a diameter of 600 mm and height of 400 mm. Commercial Al (100 ×
100 × 30 mm3 ) with a purity of 99.98% was employed as both the raw material and
the anode of the vacuum arc discharge apparatus, and it was placed under the tip of
the tungsten cathode (the diameter was 10 mm) with a distance about 5–10 mm.
The phase purity and crystal structure of the products were detected by X-ray
diffraction (XRD, Regaku D/max-2400). A scanning electron microscopy (SEM,
Philips XL30) was used to observe the morphology of the as-prepared samples.
Further detailed structural characterization and elemental composition analysis were
performed on a high resolution transmission electron microscopy (HRTEM, JEOL
JSM-2010) equipped with an energy dispersive spectroscopy (EDS, Oxford). Pho-
toluminescence (PL) spectrum of the AlN nanostructures was measured using a
He-Cd laser (325 nm) as the excitation source at room temperature.

13.3.1 Synthesis of 1D Nitrides
13.3.1.1 Platelet BN nanowires
The platelet BN nanowires were synthesized by using amorphous boron powder and
boron oxide powder as boron source, NH3 as nitrogen source, nickelocene as cata-
lyst precursor, and FeS as a growth promoter (Tang et al., 2006a). The raw chemicals
(except for nickelocene) were evenly mixed and put into an h-BN boat covered with
a graphite substrate, and then placed at the centre of the furnace, while nickelocene
was placed upstream of the reactor. After fluxing with Ar for half an hour, NH3 and
Ar mixture gas at a ratio of 1:8 was introduced. The temperature was increased to
1350◦ within 35 minutes and kept unchanged for 1 hour. Then, the furnace temper-
ature was allowed to lower down to room temperature gradually. The as-prepared
products are white cotton-like substance deposited on the graphite substrate. By
tuning the growth time, gas flow rate, and the deposition position, the platelet BN
nanowires can be selectively obtained (Fig. 13.1). We can see that the nanowires are
not uniform in diameter along their axis direction, and they are actually composed of
numerous series-wound cup-like sections (Fig. 13.1a, b). The lengths of the “cups”
range from tens of nanometers to 500 nm. The cone angles of the “cups” are similar
within one nanowire, while they may differ from each other for different nanowires.
HRTEM observation (Fig. 13.1c) shows that the cup-like section is single-crystalline
and their (0002) planes are perpendicular to the axis direction with an interplanar
distance of 0.34 nm. The chemical composition of the nanowires was analyzed by
using electron energy loss spectroscopy (EELS). A typical EELS spectrum is shown
in Fig. 13.1d. Clear K edges of boron and nitrogen are shown at around 188 and
401 eV, respectively. No noticeable C-K edges (∼284eV) can be detected.
13.3.1.2 AlN Nanorod Array Controlled by Temperature Gradient
A vacuum apparatus that can produce a plasma electric arc was employed for the
synthesis of AlN nanorod arrays (Fig. 13.2, Tang et al., 2005c). Commercial Al
Fig. 13.1 SEM (a, b), TEM (c) images, and an electron energy loss spectrum (d) of the as-prepared
platelet BN nanowires
(100 × 100 × 30 mm3 ) with a purity of 99.98% was employed as both the raw
material and the anode of the vacuum arc discharge apparatus, and it was placed
under the tip of a tungsten cathode (the diameter was 10 mm) with a distance about
5–10 mm. The discharge current was kept at 150 A throughout the reaction. After
the arc discharge proceeded for 30 minutes, a layer of white products was obtained
around the anode Al.
Fig. 13.2 The SEM images of AlN nanorod array: (a) tilted top-view, and (b) high magnification.
High-resolution TEM image of AlN nanorod (c) and the corresponding ED pattern (d)
The essence of 1D nanostructure formation is about crystallization which

involves two fundamental steps: nucleation and growth (13.1). The nucleation
of a typical vapor-liquid-solid (VLS) process are mainly induced and dictated
by catalyst liquid droplets. Without using catalyst, the growth mechanism of the
single-crystalline mushroom-like AlN nanorod arrays should be a special VLS
process. Figure 13.3A schematically illustrates three major steps involved in the
growth process of the AlN nanorod. Based on the Al-N binary phase diagram
(Fig. 13.3B) (Tang et al., 2005b), Al and N form AlN at very low nitrogen partial
pressure when the temperature is raised above the eutectic point (660.45◦). The
reaction can be expressed as: N(g) + Al(l) → AlN(s). Furthermore, the higher the
reacting temperature, the quicker the reaction. Due to the high current densities and
plasma temperatures involved, target material melts and atomic flux protuberated
from the molten surface reacts with ionized nitrogen to form AlN nuclei by a
homogeneous nucleation process (step I) (Haber et al., 1998). This mechanism is
supported by the SEM observation (Fig. 13.3C) of the product obtained on the
initial stage of synthesis process, which clearly records the nucleation of AlN
nanorods. Subsequently (step II), the location of liquid isotherm of AlN is gradually
raised to form an extending temperature gradient field as the tungsten cathode was
moved upward, and the AlN nuclei served as active sites for Al cluster adsorption
Fig. 13.3 The growth mechanism of the AlN nanorod arrays: (A) schematic illustration three
major steps in the growth process, (B) the Al-N binary phase diagram, (C) the SEM observation
in step I, (D) the SEM observation in step II, (E) the SEM observation in step III, the scale bar in
(C)–(E) is 200 nm
would preferentially grow along the temperature gradient which is also [001] of
the nanorod crystal. In this process, the temperature gradient plays an important
role to induce the growth of AlN nanorod. A SEM image of the product obtained
on the middle stage of synthesis process is shown in Fig. 13.3D, which obviously
proves the growth process of AlN nanorod. As the cathode was kept at a certain
height, the speed of the growth along the temperature gradient would slow down
and form the droplet of [Al(l)+AlN(s)] on the top of AlN nanorod which would
react subsequently with the nitrogen to form a polyhedral AlN ball (Fig. 13.3E).
The reason for the formation of the polyhedrical end of the nanorod is likely due
to reducing the surface area to decrease the surface energy according to the Wulff’s
theorem. In a word, unlike the typical catalyst induced VLS growth mechanism,
temperature gradient is one of the key factors in the nucleation and growth process.
13.3.1.3 Oxide-assisted Growth of AlN Nanotip Array Controlled

by the Amount of Reactants
AlN nanotip array has also been synthesized by a CVTC method (Tang et al.,
2005a). Some typical SEM images of the as-synthesized product are shown in
Fig. 13.4a, b and c. It is very interesting that the morphology of the product shows
an Eiffel-tower-shape nanostructure, as shown in Fig. 13.4a. Most of the deposited
AlN nanotips are oriented vertically with respect to the Si substrate, while some
tips with different directions are occasionally observed (Fig. 13.4b). Independent of
their size, AlN nanotips are always composed of two parts, a submicron-sized base
Fig. 13.4 SEM images of AlN nanotip array: (a) tilted top view, (b) side view and (c) high magni-
fication, and (d) EDS spectrum
and a sharp tip. As shown in Fig. 13.4c, the base is a polygonal pyramid with a few
hundred nanometers in size, and the sharp tip is several to tens of nanometers in
diameter and ∼1.0μm in length. The EDS spectrum (Fig. 13.4d) indicates that these
nanostructures only consist of Al and N elements.
A representative transmission electron microscopy (TEM) image of an individual
nanotip is shown in Fig. 13.5a, which was taken along one of its side surface per-
pendicular to the electron beam. Selected area electron diffraction (SAED) patterns
were recorded during TEM observations. SAED pattern (Fig. 13.5b) taken from the
region marked by a circle in Fig. 13.5a can be indexed as that of single-crystalline
wurtzite AlN along the [010] zone axis. Figure 13.5c presents the corresponding
HRTEM image which was taken near the edge and along the length of this tip. The
spacing between any two adjacent lattice fringes (0.25 nm) corresponds to that of the
(002) lattice planes of the hexagonal AlN. The axis of the tip is also perpendicular to
the lattice plane (002), that is to say, the axis direction of AlN nanotips is generally
along [001]. Both the HRTEM and SAED results demonstrate that the as-grown
nanotips are single-crystalline growing along [001] direction of AlN hexagonal
wurtzite structure.
Experiments were carried out to study the role of each reactant played in growth
and to explore the growth mechanism. To verify the formation mechanism of the
Eiffel-tower-shape AlN nanotips, and to see whether the starting material without
Fe2 O3 is suitable for synthesizing this nanostructure, only Al nanoparticle was used
as Al source for the reaction under similar experimental conditions; but no tips sim-
ilar to those obtained with Fe2 O3 were formed. We also used the mixture of Al and
Al2 O3 nanoparticles as raw reactant under the same condition and did not observe
any nanostructures as well, only leading to the deposition of amorphous particles
of AlN (several microns in size) on Si substrate. It can therefore be concluded that
Fig. 13.5 TEM image (a), SAED pattern (b) and the corresponding HRTEM image (c) of a single
AlN nanotip
Fe2 O3 plays a key role in this synthesis process. It is speculated that the starting
material will undergo the following reactions:
2Al(l) + 2NH3 (g) → 2AlN(s) + 3H2 (g) , G = −401.72 kJ/mol (13.1)
where G is the change of Gibbs free energy of the reaction at 1000 K. This reaction
is highly exothermic, which may result in substantial local heating, therefore pro-
moting the following reactions (Zhang et al., 2001; Liu et al., 2001). Subsequently,
Al2 O vapors are obtained from the chemical reaction:
6Al(l) + Fe2 O3 (s) → 3Al2 O(g) + 2Fe(s), G = −106.69 kJ/mol (13.2)
In the previous synthesis methods (Zhang et al., 2001; Liu et al., 2001), the Al2 O
vapor is difficult to obtain at low temperature because the Al2 O3 is hard to reduce
by Al or carbon due to its very low Gibbs free energy (for example, G(Al2 O3 ) is
−1777 kJ/mol at 1000 K) (Liang et al., 1993). In this case, we used Fe2 O3 instead
of Al2 O3 as oxidizer and it can significantly facilitate reaction (13.2) to proceed.
The Al2 O vapor will reacts with NH3 to generate AlN species:
Al2 O(g) + 2NH3 (g) → 2AlN(s) + H2 O(g) + 2H2 (g),

G = −518.22 kJ/mol (13.3)
Reaction (13.3) is a vapor-vapor reaction between Al2 O and NH3 . The formed
AlN species can be readily transported to the deposition zone by the carrier gas and
grow into AlN crystals.
It is well known that the nucleation of a typical vapor-liquid-solid (VLS) process
are mainly induced and dictated by catalyst liquid droplets. In our experiment, the
growth of the nanotips should be governed by a vapor-solid epitaxy process. From
the morphology of the tips (Fig. 13.4), it is believed that the nanotips are formed by
two steps, as schematically illustrated in Fig. 13.6a. Step I: growth of the submicron-
sized base initially on the Si substrate and step II: growth of nanoneedle on the top
of the base. In the initial stage, the abundant reactants provide sufficient AlN vapor,
therefore, the growth of AlN is so rapid that relatively big bases are formed. With
the deposition going on, the concentration of AlN atoms reduces gradually due to
limited supply from the source, so the quantity of AlN atoms at the VS interface
becomes depleted, which leads to a gradual reduction of the area of VS interface
along the growth direction, and brings on the growth of AlN nanoneedles. The SEM
observation (Fig. 13.6b) of the product obtained at the middle stage of the synthesis
process (50 minutes) clearly records the growth process of this nanostructure.
13.3.1.4 AlN Honeycomb-like Network Controlled by the Surface Conditions

of Substrate
AlN nanoplatelets were synthesized by heating the mixture of Al and AlCl3 powders
in the flow of N2 (Tang et al., 2006b). Briefly, a Si wafer (2 × 3 cm) was used as
Fig. 13.6 The schematic illustration and SEM observation of controlled growth of AlN nan-
otip array
the substrate, which was etched using 2% hydrofluoric acid for about 5 minutes and
ultrasonically cleaned by acetone and de-ionized water several times to remove sili-
con oxide on its surface. The Si substrate was placed in a ceramic boat and deposed
at downstream of the gas flow with a distance of ∼15 cm from the reactants. The Ar
gas was substituted by N2 (50 sccm) as the furnace was heated to 1200 ◦C at a rate
of 30 ◦ C/min. The reaction was maintained for 1 hour at a pressure of 3 Torr. After
the furnace was naturally cooled down to room temperature under the flow of Ar,
a layer of white product was observed on the surface of the Si substrate. However,
when the non-etched Si wafer was used as the substrate and other conditions were
kept the same, we did not get similar results.
The AlN nanoplatelets synthesized on Si substrate are shown in Fig. 13.7. The
nanoplatelets vertically grow on the Si substrate and connect with each other to form
a honeycomb-like network. The thickness of the single-crystalline nanoplatelets
varies from 10 to 100 nm, and their height varies from several hundred nanometers
to several microns (Fig. 13.8).
Fig. 13.7 The schematic illustration and SEM observations of controlled growth of AlN
nanoplatelets
13.3.1.5 Catalyst-seeded Synthesis of Flowerlike Si-doped

AlN Nanoneedle Array
Si wafers (1 × 2 cm), on which cobalt (Co) catalyst particles (1–3 μm in diame-

ter, 99.9%) were evenly dispersed (Fig. 13.9a], were employed as substrate for the
growth of Si-doped AlN nanostructures (Tang et al., 2006c). In a typical process,
1 g Pluronic P-123 triblock copolymer ordered from Aldrich was dissolved in 30 ml
ethanol, and 0.10 g Co particles were added to this solution and then stirred for 30
minutes to form an emulsion. The catalyzed Si substrates were prepared by coating
the emulsion solution containing Co particles on Si wafers. Before being loaded
into the furnace, the coated Si substrates were annealed at 500 ◦C in hydrogen atmo-
sphere for 20 minutes to remove the polymer. AlCl3 (99.9%) powders were placed
in an alumina boat and the catalyzed Si substrate was put downward on the top of
the source boat at the center of a horizontal tube furnace. After that, the furnace
was purged with argon gas which was then substituted by NH3 /H2 (60 ml/min, 2:1
in volume) as the furnace was heated to 900 ◦C. About 20 ml liquid SiCl4 (99.9%)
was gradually fed into the tube by an injector. A white-color thin film formed on the
substrate after about 30 minutes reaction.
A typical SEM image of the AlN nanostructures synthesized at 900 ◦C is shown
in Fig. 13.9b, displaying the sample is an AlN nanoneedle array with a flower-like
morphology, which were obtained in MoS2 , GaP, and MgO, etc. (Liu et al., 2005a;
Feng et al., 2006; He et al., 2006). As an example, Fig. 13.9c presents a complete
AlN micro-flower, from which we can see that numerous AlN nanoneedles grow
radially from a seeded Co catalyst particle. These nanoneedles are not absolutely
straight, and they are several micrometers long and gradually taper along the growth
Fig. 13.8 The SEM images of AlN nanoplatelets: (a) tilted top-view and (b) high magnification.
XRD diffraction spectrum (c) and EDS spectrum (d) of AlN nanoplatelets. TEM image (e), SAED
pattern (f) and the magnified HRTEM image (g) of a single AlN nanoplatelet
direction (Fig. 13.9d). The base diameters of the nanoneedles are in the range of
50–150 nm, and the tips are 5–30 nm in diameter. By varying the concentration of
the seeding emulsion, the area density of the nanoneedle array on the substrate is
able to be adjusted.
Figure 13.10a is a TEM image of an AlN nanoneedle. The selected area
electron diffraction (SAED) pattern (Fig. 13.10b) reveals that the nanoneedle
Fig. 13.9 (a) The Si substrate covered with Co catalyst particles, and (b)–(d) SEM images of
flower-like AlN nanoneedle array synthesized at 900 ◦ C
is single-crystalline and grows along the [001] direction. Figure 13.10c is the
magnified HRTEM image taken from the tip of this nanoneedle. The distance
between adjacent lattice planes perpendicular to the growth direction is 0.50 nm,
corresponding to the spacing of (001) planes of hexagonal AlN. Figure 13.11 dis-
plays a TEM image of an AlN nanoneedle, and its corresponding Al, N, and Si
elemental mappings. It reveals that Si element is uniformly distributed in the AlN
nanoneedle.
Fig. 13.10 TEM image (a), SAED pattern (b), the magnified HRTEM image (c)
Fig. 13.11 TEM image of an AlN nanoneedle (a), and its corresponding Al (b), N (c), and Si (d)
elemental mappings
The catalytic Co particles are believed to be responsible for the formation of

flower-like AlN nanoneedles since no similar nanostructures were formed on the
Si substrate without covering Co particles under similar experimental conditions.
The growth of the Si-doped AlN nanoneedles should be a base-model VLS process
which is similar to the previous reports where Ni films were used as catalyst for
synthesizing 1D AlN nanostructures (Liu et al., 2005b; Xu et al., 2004). At the
reaction temperature, Al, N, and Si vapors are generated by thermal decomposition
of AlCl3 , NH3 , and SiCl4 , and they deposit on the surface of the Co particles to
form eutectic liquid droplets. Solid AlN nuclei will precipitate from alloy droplets
that become supersaturated with Al and N supplied from the gas phase. Meanwhile,
Si doping is achieved by the process of Si substituting Al atom in AlN crystal.
Continuous dissolving of Al, N and Si atoms into the liquid droplets maintains the
growth of the Si-doped AlN nanostructures. The growth rate of the AlN nanostruc-
tures in the axial direction far exceeds that in the radial direction, resulting in a
needle-like morphology, that is, AlN nanoneedles. In this study, when the synthesis
temperature was higher than 1000 ◦C, larger AlN nanocones could be obtained as
shown in previous reports (Liu et al., 2005b; Zhao et al., 2005), which probably
originates from the fact that the radial growth becomes more obvious at a higher
temperature. Nevertheless, the detailed growth mechanism still needs further study.
13.3.1.6 AlN Nanobelt Array
The synthesis of AlN nanaobelt arrays is a CVTC process. Al and Fe2 O3 nanopar-
ticles (∼50 nm in size) were fully mixed in a weight ratio of 1:1. When the
temperature of the furnace reached 900 ◦C, the Ar gas was substituted by 80 sccm
Ar/NH3 (3:1 in volume) for 1 hour. The Si substrate was covered with a layer of
fluffy thin film after the reaction.
Figure 13.12a shows the typical SEM image of an as-grown sample deposited on
the Si substrate. It can be seen that high-purity vertically aligned 1D nanostructures
distribute relatively uniformly on the surface of the Si substrate. The length of this
1D structure is in the range of 1–3 μm. Some nanostructures are bent along their
length without being broken, which indicates that they are flexible and probably
belt-like in nature. TEM observations further confirm the belt-like morphology of
these 1D nanostructures. The uniform contrast and electron beam transparent char-
acteristics demonstrate that the formed structures are nanobelts (Fig. 13.12b). The
ripple-like contrast in the TEM image is due to the bending of the nanobelts. The
width of the belts is in the range of 20–150 nm and the thickness is in the range
of 10–30 nm. The ratio of width to thickness of the belts ranges from 2 to 5, as
estimated by TEM observations.
Figure 13.12c presents a TEM image of an individual nanobelt, which was taken
along one of its side surfaces perpendicular to the electron beam and used for SAED
and HRTEM observations. Its corresponding SAED is displayed in Fig. 13.12d.
From the SAED pattern, the crystal structure of this belt is identified to be hexagonal
wurtzite AlN phase. It also indicates that the belt is single crystalline and enclosed
by ±(100) and ±(010) crystallographic facets. The HRTEM image taken near the
edge of this nanobelt is shown in Fig. 13.12e. The adjacent lattice spacing in longitu-
dinal direction is about 0.25 nm, corresponding to that of the (002) planes of wurtzite
AlN, which further confirms that the nanobelt is wurtzite AlN phase. Meanwhile, it
shows that the nanobelt grows along [001] direction, which is the frequent growth
orientation for hexagonal wurtzite-structure materials. In addition, a ∼2 nm thick
amorphous coating is clearly seen along the surface of the belt.
13.3.1.7 AlN Nanofibers
The mixture of Al and Fe2 O3 nanoparticles (less than 100 nm in size) with a weight
ratio of 5:2 was used as raw material and was placed in a ceramic boat which was set
at the center of the alumina tube. The furnace was rapidly heated to and maintained
at 800 ◦C for 1.5 hours under a flow of NH3 /Ar (2:1 in volume) at a rate of 150 sccm.
After the furnace was cooled down to room temperature, a layer of white wool-like
product was found on the surface of the substrate and it was collected for later
characterization (Tang et al., 2005b).
The morphology of the as-synthesized products was observed by scanning elec-
tron microscopy (SEM). The typical SEM images of the products are shown in
Fig. 13.13. The Si substrate is covered by a large number of randomly distributed
wire-like nanostructures (Fig. 13.13a), and most of these 1D nanostructures have
uniform diameter over their entire length. Figure 13.13b shows that some nanofibers
are bent to a 180 angle without being broken, which demonstrates that the nanofibers
are very flexible. To further verify the geometrical morphology of the AlN nanos-
tructures, a high magnification SEM image of the nanofibers is shown in Fig. 13.13c,
Fig. 13.12 (a) The typical SEM image of the AlN nanobelt array on the Si substrate. (b) Low-
magnification TEM image of the AlN nanobelts. (c) a high-magnification TEM image, (d) SAED
pattern, and (e) the corresponding HRTEM image of a single AlN nanobelt
which displays prism morphology. The nanofibers are several tens of micrometers
in length and 10–200 nm in width.
Figure 13.14a shows the representative TEM image of the end of a nanofiber. It
can be seen that this nanofiber has a rectangular cross-section shape and the ratio
of width to thickness is in the range of 1–2, which is different from the previous
reported nanobelts (Tang et al., 2006). No particles were observed at the tip of the
Fig. 13.13 SEM images of AlN nanofibers
nanofibers during the TEM observations. The typical EDS spectrum (Fig. 13.14b)
shows that the nanofibers are composed of Al and N elements (the Cu signal comes
from the copper grid) with an approximate atomic ratio of 1:1. The further detailed
microstructures of the rectangular cross-section AlN nanofibers were studied by
high resolution TEM (HRTEM) and selected area electron diffraction (SAED). TEM
image of an individual nanofiber is shown in Fig. 13.14c, which was taken oriented
with one of its side surfaces perpendicular to the electron beam. The corresponding
SAED pattern (Fig. 13.14d) was recorded along the [010] zone axis the hexagonal
wurtzite AlN. This indicates that the nanofiber is single-crystalline and it is enclosed
by (100) and (010) crystal planes. Fig. 13.14e is the HRTEM image which was taken
near the edge along the length of this nanofiber. The adjacent lattice distances in
directions of parallel and perpendicular to the nanofiber axis are 0.27 and 0.50 nm,
respectively, corresponding to the d-spacing of (100) and (001) crystal planes of
hexagonal AlN. Meanwhile, the axis of the nanofiber is perpendicular to the lattice
plane (001), which suggests the growth direction of the nanofiber is along [001]
direction, which is the frequent growth orientation for hexagonal wurtzite structure
materials. A 1–2 nm thick amorphous coating is clearly seen along the surface of
the nanofiber. This thin layer is determined to be aluminum oxide (AlOx ) from the
EDS spectrum. Amorphous oxide coating is often found in AlN 1D nanostructures
(Zhang et al., 2001; Liu et al., 2001). Both the HRTEM and SAED observations
reveal that the nanofibers are single-crystalline wurtzite AlN which grow along
[001] direction.
Fig. 13.14 (a) TEM image of the end of an AlN nanofiber. (b) The typical EELS spectrum of AlN
nanofibers. (c) TEM image, (d) SAED pattern, and (e) the corresponding HRTEM image of an
AlN nanofiber
13.3.2 Properties of 1D AlN
13.3.2.1 Field Emission of 1D AlN Arrays
The FE measurements for four kinds of 1D AlN arrays (Fig. 13.15) were carried
out inside a ball-type chamber which was pumped down to 5.0 × 10−9 torr by an
ultrahigh vacuum system. The anode was a cylinder-shaped platinum probe (1 mm
in diameter) and the cathode was AlN nanotip array, with a distance of 100 μm.
High voltage was supplied by a power source (Keithley 248) and the current under
increasing applied voltage which varied with a step of 10 V was recorded by a sensi-
Fig. 13.15 SEM images of AlN samples for FE measurements: (1) nanorod array, (2) nanotip
array, (3) nanoplatelets and (4) Si-doped nanoneedle array
tive electrometer (Keithley 6514) with an accuracy of 10−11 A. Figure 13.16a shows
the measured FE current as a function of the applied voltage and the corresponding
Fowler-Nordheim (F-N) plot (the inset) of 1D AlN nanotip array. It can be seen that
the emission starts at 320 V, and the anode current reaches 81 nA as the voltage is
increased to 470 V. Here, we define the turn-on field (E to ) and the threshold field
(E thr ) as the electric fields required to produce a current density of 10 μA/cm2 and
10 mA/cm2 , respectively.
The FE current-voltage characteristics were analyzed by using the Fowler- Nord-
heim (FN) equation (Taniyasu et al., 2004):
I = (1.54 × 10–10 β 2 V 2 A/d 2 φ) exp(−6.83 × 103 φ 3/2 d /βV )
where I is current, V is applied voltage, φ is work function of the emitting material,

β is field enhancement factor, A and d are the area of emission and the distance
between the anode and the cathode, respectively. The inset in Fig. 13.16a shows the
FN plot of ln(I /V 2 ) vs I /V . The plot has an approximately linear relation within
the measurement range, which confirms that the current indeed results from field
emission. The field enhancement factor β can be calculated from the slope of the
FN plot if the work function of the emitter is known.
Figure 13.16b shows the result of FE current versus time for a period over
240 min at a pressure of 5.0 × 10−9 torr under the fixed applied voltage of 1060 V
for an initial current 77.89 μA. The average current and the standard deviation
were calculated to be 77.53 and 0.57 μA, respectively. Since the emission area
was 0.78 mm2 , the E to and the E thr for the AlN nanotip array were 4.7 V/μm
and 10.6 V/μm, respectively. Taniyasu et al. (2004) obtained a field of 23 V/μm
at 9.5 μA/cm2 for Si-doped AlN film, which is about five times higher than our
result. The E to of AlN nanotip array is comparable to the reported values of carbon
Fig. 13.16 The typical FE curves of 1D AlN array: (a) FE current as a function of the applied
voltage, inset is the corresponding Fowler-Nordheim plot; (b) field emission current stability
nanotubes (Bonard et al., 1998; Jeong et al., 2001). The reason for the low turn-
on electric field and high FE current density are probably due to single-crystalline
structure and nanosize of the emitter. Many researchers ever reported that in order to
avoid a field screening effect, aligned nanostructures with low density was favored
(Jeong et al., 2001). The density of our AlN nanotip array is relatively low, approxi-
mately 106 –107 tips/cm2 . Therefore, low value of E to and high value of FE current
are obtained.
The average current and the standard deviation were calculated to be 77.53 and
0.57 μA, respectively, for 4 hours FE stability test. The ratio of them is as low as
0.74% (Table 13.1), which proves the high current stability of Eiffel-tower-like AlN
Table 13.1 The field emission properties of four 1D AlN arrays obtained
Samples 1D arrays Eto (V/μm) Ethr (V/μm) B Fluctuation
1 Mushroom rod 8.8 19.2 565 <2%/1hour
2 Eiffel tip 4.7 10.6 1175–1889 0.74%/4hour
3 Honeycomb 3.2–5.0 7.8–12.1 1015–1785 2.7%/5hour
4 Si-doped needle 1.8 4.6 3271 <5%/5hour
Bonard et al. 1998 AlN film 34 84 —— 3%/2hour
Taniyasu et al. 2004 AlN film 11 23 —— 5.5%/1hour
Kasu et al. 2001 Carbon nantube 1.5–4.5 3.9–7.8 3600 >15%
nanotip. To the best of our knowledge, such a low fluctuation has not ever been
reported for AlN material. Yin et al. (2005) reported a fluctuation of 7% for hier-
archical AlN comb-like nanostructures. A fluctuation of 5.5% for heavily Si-doped
AlN has been reported by Taniyasu et al. (2004). When an electric field is applied to
an emitter, heat is generated at its sharp end because of high current density (Tondare
et al., 2002). Therefore, the FE device will exhibit poor stability if the emitter has
a poor thermal stability. The submicron-sized bases of our nanostructures have so
large connection area that they can quickly transfer the heat from the tip to the Si
substrate, so the tip can be effectively protected from being destroyed due to the
superheat. Furthermore, the observed high stability of FE current is also attributed
to the little change of the emission site, due to the strong bonding between the bases
of AlN nanotips and the Si substrate.
The E to value of the Si-doped AlN nanoneedle array in this work is lower than
that previously reported (Table 13.1). The low E to , high FE current, and high β of
the Si-doped AlN nanoneedle array are mainly attributed to their sharp tips, high
aspect ratio, and especially Si doping. The high aspect ratio and small radii of cur-
vature of the AlN nanoneedles can generate a high local electric field at their tips,
which decreases the FE potential barrier and so increases the FE current. On the
other hand, according to Taniyasu et al. (2004), the Si dopant will form an impurity
band just below the conduction band of AlN, therefore, the electrons for FE can
easily tunnel through this shallow donor level to vacuum under the electric field.
In addition, the Si doping remarkably lowers the resistance of AlN and increases
the carrier concentration, which reduces the voltage drop along the nanoneedles and
increases the effective field at their tips.
13.3.2.2 Photoluminescence (PL) of AlN Nanobelt Array
Figure 13.17 depicts the PL spectrum from the AlN nanobelt array (Tang et al.,
2007a). It consists of a weak violet emission centered at ∼ 410 nm (3.02 eV) and
Fig. 13.17 Room-temperature PL spectra of the as-prepared AlN nanobelts, AlN nanobelts after
NH3 treatment and pure AlN particles
a strong green emission centered at ∼ 573 nm (2.16 eV) with two shoulders of 523
and 619 nm. Apart from a slight difference of intensity, the main features of the
PL spectrum are similar when the laser beam is focused on different regions of
as-synthesized sample. Like the PL spectra from AlN nanowires and AlN nanocones
reported by Zhao et al. (2005) and Liu et al. (2005), respectively, the direct band gap
emission of the AlN nanobelt array can not be seen due to detection limitation. The
observed weak violet emission of this nanostructure is similar to that of nanocrys-
talline AlN (420 nm) (Cao et al., 2000), in which such emission is related to the
transition from shallow donor level to the excited state of the deep level. Lan et al.
(1999) also reported emission at 405 nm (3.06 eV) from AlN nanopowders, and they
attributed it to deep-level nitrogen vacancies and surface-luminescence.
The green luminescence band (2.16 eV) in this study is somewhat similar to the
broad emission spectrum with peaks at 2.9, 2.3, and 2.2 eV from epitaxial AlN
reported by Yim et al. (1973), and these peaks involve recombination at deep levels
of unknown nature. The origin of this green emission in our AlN nanobelts is still
not fully clear, but we believe that it is mainly originated from oxygen-related and
structural defects associated with the peculiar nanostructures such as AlN nanowires
and nanocrystalline AlN layer (Zhao et al., 2005; Siwiec et al., 1998). In this case,
the nanobelts with small size and high specific surface area should also be easily
doped with oxygen, which was detected by the TEM and EELS. The oxygen impuri-
ties will produce defects including oxygen point defects (ON + ), nitrogen vacancies
(VN ), and VAl 3− − 3 × ON+ complexes in this nanobelt (Mattila and Nieminen,
1996; Youngman and Harris, 1990). Thus the green emission may be resulted from
the radioactive recombination of a photo-generated hole with an electron occupying
the nitrogen vacancies and/or from the transition between the shallow level of VN
and the deep level of VAl 3− − 3 × ON + complexes (Cao et al., 2000; Mattila and
Nieminen, 1996). In addition, structural defects such as stacking faults and lattice
dislocations may exist in our nanobelts according to the ripple-like contrast on the
surface of some nanobelts in the TEM images because of strain resulting from the
bending, which possibly induce the green emission because wurtzite AlN similar to
wurtzite ZnO is a polar material.
Violet and green light emissions can be obtained from this AlN nanostructure,
suggesting its potential applications in light emission display devices. And this
nanostructure can be expected to serve as light sensors and field-emission devices,
etc. which requires not only high purity but also good alignment.
13.3.2.3 Mechanical Properties and Thermal Expansion of an AlN

Nanowires/Al Composite
The AlN-NWs were synthesized on quartz substrates by a vapor deposition

method (Tang et al., 2007b). Figure 13.18a presents the typical scanning electron
microscopy (SEM) image of the nanowires. The quartz substrate was covered by
a large quantity of high-purity nanowires, and each of them has uniform diameter
over its entire length. The nanowires have diameters in the range of 10–50 nm and
lengths over several micrometers, and their aspect ratios are generally over 100. The
Fig. 13.18 (a) SEM image and (b) XRD pattern of the synthesized AlN nanowires
X-ray diffraction (XRD), shown in Fig. 13.18b, reveals the nanowires to exhibit the
AlN wurtzite structure with the lattice constants of a = 3.114 Å and c = 4.979 Å
(JCPDS: No. 25–1133).
The Al nanoparticles were produced by an active H2 plasma evaporation tech-
nique. Figure 13.19 shows the schematic diagram of fabrication process of the
AlN-NWs/Al composite. The three AlN-NWs/Al composites with AlN-NWs load-
ing of 5, 10 and 15 vol.-% were prepared by ultrasonic mixing, and each composite
powder was hot pressed in pure argon atmosphere under 50 MPa pressure at 700 ◦C
to form a disk. For comparison, the Al matrix with the same size was also fabricated
under the same conditions utilizing Al nanoparticles. The relative densities of all
the AlN-NWs/Al composites and non-reinforced Al matrix were measured and
calculated to be higher than 99%.
Figure 13.20 shows the representative SEM image of the polished AlN-NWs/Al
composite containing 15 vol.-% AlN-NWs after etched by 1% hydrofluoric acid.
AlN-NWs can be seen to be embedded in the continuous Al matrix exposing both
their cross-sectional and the longitudinal views, indicating a non-directional align-
ment of the nanowires in the compacted specimen. There is no agglomeration of
AlN nanowires in composite specimens due to the uniform dispersion of AlN-
NWs in Al nanoparticles during mixing because of the minute difference of den-
sity between the AlN (∼3.2 g/cm3 ) and Al (∼2.7 g/cm3 ). The composite material
appears to be macroscopically homogeneous and free of porosity owing to the high
Fig. 13.19 The schematic diagram of fabrication process of the AlN-NWs/Al composite
pressure employed during fabrication to make the molten Al to fill in the mould
from edge to center completely.
It is known that interfacial characteristics are of great importance to the perfor-
mance and properties of composites (Dhingra and Fishman, 1986). Having a good
interface, composites can effectively transfer the external load from the matrix to
the reinforcement. And the interface mainly depends on the wetting between the
matrix and reinforcement. The adhesion between the reinforcement and the matrix
must be adequate, so that the composite achieves good shear and off-axis properties
Fig. 13.20 The representative SEM image of the polished 15 vol.-% AlN-NWs composite after
etched by 1% hydrofluoric acid
normally required by reinforced metals (Piggott, 1980). On the contrary, micro-

voids will form on the interface if the wetting is inadequate, which will result in the
decrease of interfacial bonding strength. Additionally, harmful chemical reactions
between the reinforcement and the matrix must be limited to prevent the reinforce-
ment from being weakened (Dhingra and Fishman, 1986). So, it is necessary to do
further investigation on the interface of AlN-NWs/Al composites.
Figure 13.21a and b are the TEM images of the composite containing 15 vol.-%
AlN-NWs. AlN-NWs can be clearly observed in the matrix, indicating that the
nanowires are not broken by the high temperature and pressure consolidation. An
enlarged HRTEM image of the interfacial region is shown in Fig. 13.21c. It reveals
that the interface between AlN nanowire and Al matrix is clean and bonded well. No
porosity or interfacial reaction product is observed at the nanowire/matrix boundary
by the HRTEM observations. We assume that the improvement of mechanical prop-
erties should be attributed to the strongly bonded interfaces, which can efficiently
transfer stress from the matrix to the reinforcement.
Mechanical properties including tensile strength, Young’s modulus and hardness
of the composites are improved with AlN-NWs volume fraction changing from
5 to 15 vol.-%, and the maximum tensile strength, Young’s modulus and hardness
of the composite are about 5, 3 and 2 times higher than those of Al matrix. Mean-
while, AlN-NWs effectively decrease the coefficient of thermal expansion (CTE) of
Al matrix, and the CTE of the AlN-NWs/Al composite with 15 vol.-% AlN-NWs
is about one half of that of the Al matrix. The results obtained suggest that the
Fig. 13.21 (a), (b) TEM images of the composite containing 15 vol.-% AlN-NWs, and (c) an
enlarged HRTEM image of the interfacial region
AlN nanowire is an ideal reinforcement for optimizing the mechanical and thermal
properties of metal matrix composites.
13.4 Conclusion and Outlook
Various quasi-one-dimensional (1D) nitride nanostructures including BN nanotubes

and nanofibers, AlN nanorods, AlN nanotips, AlN nanobelts, and AlN nanofibers
are synthesized. Especially, the single-crystalline AlN mushroom-like nanorod
array, Eiffel-tower-shape nanotip array, honeycomb-like network of vertically
aligned nanoplatelets and flowerlike Si-doped AlN nanoneedle array are selectively
synthesized by an improved chemical vapor transport and condensation method
and a mobile nitrogen arc discharge method, respectively. Field emission (FE)
measurements for AlN aligned 1D nanostructures show that they have very low
turn-on voltages, high emission currents and small FE current fluctuations. It is
revealed the morphology of the nanostructures can be tailored to optimize their
FE properties and make them promising field emitters. Moreover, a strong green
light emission is obtained from AlN nanobelts. Mechanical properties and thermal
expansion of a composite of AlN nanowires and nanocrystalline Al implies that
the AlN nanowire is also an ideal reinforcement for improving the mechanical and
thermal properties of metal matrix composites.
Acknowledgments The authors thank Drs. F. Li, Z.G. Chen, B. Wu, and J.F. Wu, Mr. B.J. Xia and
K.Y. Hu for their kind help. The project was supported by National Natural Science Foundation of
China (Grant No. 50371083)
References
Benjamin, M. C., Wang, C., Davis, R. F., and Nemanich, R. J. 1994, Observation of a negative
electron-affinity for heteroepitaxial AlN on alpha(6H)-SiC(0001), Applied Physics Letters, 64,
pp. 3288–3290.
Bonard, J. M., Salvetat, J. P., Stockli, T., de Heer, W. A., Forro, L. and Chatelain, A. 1998, Field
emission from single-wall carbon nanotube films. Applied Physics Letters, 73, pp. 918–920.
Bradshaw, S. M. and Spicer, J. L. 1999, Combustion synthesis of aluminum nitride particles and
whiskers. Journal of the American Chemical Society, 82, pp. 2293–2300.
Cao, Y. G., Chen, X. L., Lan, Y. C., Li, J. Y., Xu, Y. P., Xu, T., Liu, Q. L. and Liang, J. K. 2000,
Blue emission and Raman scattering spectrum from AlN nanocrystalline powders. Journal of
Crystal Growth, 213, pp. 198–202.
Chedru, M., Chermant, J. L. and Vicens, J. 2001, Thermal properties and Young’s modulus of
Al-AlN composites. Journal of Materials Science Letters, 20, pp. 893–895.
Chen, G. D., Smith, M., Lin, J. Y., Jiang, H. X., Wei, S. H., Khan, M. A. and Sun, C. J. 1996,
Fundamental optical transitions in GaN, Applied Physics Letters, 68, pp. 2784–2786.
Chen, X. and Gonsalves, K. E. 1997, Synthesis and properties of an aluminum nitride/polyimide
nanocomposite prepared by a nonaqueous suspension process. Journal of Materials Research,
12, pp. 1274–1286.
Dhingra, A. K. and Fishman, S. G. 1986, Interfaces in Metal-Matrix Composites (New Orleans,
Metallurgical Society Inc.), p. 211.
Geiger A. L. and Jackson, M. 1989, Low-expansion MMCs boost avionics. Advanced Materials &
Processes, 136, pp. 23–30.
Grabowski, S. P., Schneider, M., Nienhaus, H., Monch, W., Dimitrov, R., Ambacher, O. and Stutz-
mann, M. 2001, Electron affinity of AlxGa1-xN(0001) surfaces. Applied Physics Letters, 78,
pp. 2503–2505.
Haber, J. A., Gibbons, P. C. and Buhro, W. E. 1997, Morphological control of nanocrystalline
aluminum nitride: Aluminum chloride-assisted nanowhisker growth, Journal of the American
Chemical Society,119, pp. 5455–5456.
Haber, J. A., Gibbons, P. C. and Buhro, W. E. 1998, Morphologically selective synthesis of
nanocrystalline aluminum nitirde. Chemistry of Materials, 10, pp. 4062–4071.
He, J. H., Yang, R., Chue, Y. L., Chou, L. J., Chen, L. J. and Wang, Z. L. 2006, Aligned AlN
nanorods with multi-tipped surfaces - Growth, field-emission, and cathodoluminescence prop-
erties. Advanced Materials, 18, pp. 650–654.
Huang, J. L. and Li, C. H. 1994, Microstructure and mechanical properties of aluminum nitride
aluminum composite. Journal of Materials Research, 9, pp. 3153–3159.
Iijima, S. 1991, Helical microtubules of graphitic carbon. Nature, 354, pp. 56–58.
Inoue, A., Nosaki, K., Kim, B. G., Yamaguchi, T. and Masumoto, T. 1993, Mechanical strength of
ultra-fine Al-AlN composites produced by a combined method of plasma-alloy reaction, spray
deposition and hot pressing. Journal of Materials Science, 28, pp. 4398–4404.
Jeong, J. S., Lee, J. Y., Lee, C. J., An, S. J. and Yi, G. C. 2004, Synthesis and characterization of
high-quality In2 O3 nanobelts via catalyst-free growth using a simple physical vapor deposition
at low temperature. Chemical Physics Letters, 384, pp. 246–250.
Jeong, S. H., Hwang, H. Y., Lee, K. H. and Jeong, Y. 2001, Template-based carbon nanotubes and
their application to a field emitter. Applied Physics Letters, 78, pp. 2052–2054.
Kasu, M. and Kobayashi, N. 2001, Spontaneous ridge-structure formation and large field emission
of heavily Si-doped AlN. Applied Physics Letters, 78, pp.1835–1837.
Lai S. W. and Chung, D. D. 1994, Superior high-temperature resistance of aluminum nitride
particle-reinforced aluminum compared to silicon-carbide or alumina particle-reinforced alu-
minum. Journal of Materials Science, 29, pp. 6181–6198.
Lan, Y. C., Chen, X. L., Cao, Y. G, Xu, Y. P., Xun, L. D., Xu, T. and Liang, J. K. 1999,
Low-temperature synthesis and photoluminescence of AlN. Journal of Crystal Growth, 207,
pp. 247–250.
Li, Y. B., Bando, Y. and Golberg, D. 2003, MoS2 nanoflowers and their field-emission properties.
Liang, C., Shimizu, Y., Sasaki, T., Umehara, H. and Koshizaki, N. J. 2004, Au-mediated growth of
wurtzite ZnS nanobelts, nanosheets, and nanorods via thermal evaporation. Journal of Physics
Chemical B, 108, pp. 9728–9733.
Liang, Y. J. and Che M. X. 1993, Handbook for Thermodynamic Data of Inorganic Compounds
(China, Press of Northeast University).
Liu, B. D., Bando, Y., Tang, C. C., Golberg, D., Xie, R. G. and Sekiguchi T. 2005a,
Synthesis and optical study of crystalline GaP nanoflowers, Applied Physics Letters, 86,
p. 083107.
Liu, C., Hu, Z., Wu, Q., Wang, X. Z., Chen, Y., Sang, H., Zhu, J. M., Deng, S. Z. and Xu, N. S.
2005b, Vapor-solid growth and characterization of aluminum nitride nanocones, Journal of the
American Chemical Society, 127, pp. 1318–1322.
Liu, J., Zhang, X., Zhang, Y., He, R. and Zhu J. 2001, Novel synthesis of AlN nanowires with
controlled diameters. Journal of Materials Research, 16, pp. 3133–3138.
Mattila, T. and Nieminen, R. M. 1996, Ab initio study of oxygen point defects in GaAs, GaN, and
AlN. Physical Review B, 54, pp. 16676–16682.
Morales, A. M. and Lieber, C. M. 1998, A laser ablation method for the synthesis of crystalline
semiconductor nanowires. Science, 279, pp. 208–211.
Nakamura, S. 1998, The roles of structural imperfections in InGaN-Based blue light-emitting
diodes and laser diodes. Science, 281, pp. 956–961.
Pan, Z. W., Dai, Z. R. and Wang, Z. L. 2001, Nanobelts of semiconducting oxides. Science, 291,
pp. 1947–1949.
Piggott, M. R. 1980, Load-Bearing Fibre Composites (New York: Pergamon Press Inc.), Ch. 10,
p. 286.
Ronning, C., Banks, A. D., McCarson, B. L., Schlesser, R., Sitar, Z., Davis, R. F., Ward, B. L. and
Nemanich, R. J. 1998, Structural and electronic properties of boron nitride thin films containing
silicon. Journal of Applied Physics, 84, pp. 5046–5051.
Shi, S. C., Chen, C. F., Chattopadhyay, S., Lan, Z. H., Chen, K. H. and Chen, L. C. 2005,
Growth of single-crystalline wurtzite aluminum nitride nanotips with a self-selective apex
angle. Advanced Functional Materials, 15, pp. 781–786.
Siwiec, J., Sokolowska, A., Olszyna, A., Dwilinski, R., Kaminska, M. and Hrabowska, J. K. 1998,
Photoluminescence properties of nanocrystalline, wide band gap nitrides (C3 N4 , BN, AlN,
GaN), Nanostructural Materials, 10, pp. 625–634.
Sugino, T., Hori, T., Kimura, C. and Yamamoto, T. 2001, Field emission from GaN surfaces rough-
ened by hydrogen plasma treatment. Applied Physics Letters, 78, pp. 3229–3231.
Suzuki, M., Uenoyama, T. and Yanase, A. 1995, First-principles calculations of effective- mass
parameters of AlN and GaN, Physical Review B, 52, pp. 8132–8139.
Tang, Y. B., Cong, H. T., Chen, Z. G. and Cheng, H. M. 2005a, An array of Eiffel-tower-shape AlN
nanotips and its field emission properties. Applied Physics Letters, 86, p. 233104.
Tang, Y. B., Cong, H. T., Wang, Z. M. and Cheng H. M. 2005b, Synthesis of rectangular cross-
section nanofibers AlN by chemical vapor deposition. Chemical Physics Letters, 416, pp.
171–175.
Tang, Y. B., Cong, H. T., Zhao, Z. G. and Cheng, H. M., 2005c, Field emission from AlN nanorod
array. Applied Physics Letters, 86, p. 153104.
Tang, D. M., Liu, C., Cong, H. T. and Cheng, H. M. 2006a, Platelet boron nitride nanowires. Nano,
1, pp. 65–71.
Tang, Y. B., Cong, H. T. and Cheng, H. M. 2006b, Field emission from honeycomb-like network
of vertically aligned AlN nanoplatelets, Applied Physics Letters, 89, p. 093113.
Tang, Y. B., Cong, H. T., Wang, Z. M. and Cheng, H. M. 2006c, Catalyst-seeded synthesis and field
emission properties of flowerlike Si-doped AlN nanoneedle array. Applied Physics Letters, 89,
pp. 253112.
Tang, Y. B., Cong, H. T., Li, F. and Cheng, H. M. 2007a, Synthesis and photoluminescent property
of AlN nanobelt array on Si substrate. Diamond and Related Materials, 16, pp. 537–541.
Tang, Y. B, Liu, Y., Sun, C. H. and Cong, H. T. 2007b, AlN nanowires for Al-based composites
with high strength and low thermal expansion. Journal of Materials Research (accepted).
Taniyasu, Y., Kasu, M. and Makimoto, T. 2004, Field emission properties of heavily Si-doped AlN
in triode-type display structure. Applied Physics Letters, 84, pp. 2115–2117.
Tanniyasu, Y., Kasu, M. and Makimoto, T. 2006, An aluminium nitride light-emitting diode with
a wavelength of 210 nanometres. Nature, 441, pp. 325–328.
Tondare, V. N., Balasubramanian, C., Shende, S. V., Joag, D. S., Godbole, V. P., Bhoraskar, S.
V. and Bhadbhade, M. 2002, Field emission from open ended aluminum nitride nanotubes.
Vicens, J., Chedru, M., Cubero, H. and Chermant, J. L. 2002, Effects of AlN additions and heat
treatments on the compression behavior of Al-AlN composites. Journal of Materials Science
Letters, 21, pp. 1505–1508.
Vurgaftman, I., Meyer, J. R. and Ram-Mohan, L. R. 2001, Band parameters for III-V compound
semiconductors and their alloys. Journal of Applied Physics, 89, pp. 5815–5875.
Ward, B. L., Nam, O. H., Hartman, J. D., English, S. L., McCarson, B. L., Schlesser, R., Sitar, Z.,
Davis, R. F. and Nemanish, R. J. 1998, Electron emission characteristics of GaN pyramid arrays
grown via organometallic vapor phase epitaxy. Journal of Applied Physics, 84, pp. 5238–5242.
Wei, S. H. and Zunger, A. 1996, Valence band splittings and band offsets of AlN, GaN, and InN.
Wu, Q., Hu, Z., Wang, X., Lu, Y., Chen, X., Xu, H. and Chen Y. 2003a, Synthesis and character-
ization of faceted hexagonal aluminum nitride nanotubes. Journal of the American Chemical
Society, 125, pp. 10176–10177.
Wu, Q., Hu, Z., Wang, X. Z. and Chen, Y. 2003b, Synthesis and optical characterization of alu-
minum nitride. Journal of Physical Chemistry B, 107, pp. 9726–9729.
Wu, Q., Hu, Z., Wang, X., Hu, Y., Tian, Y. and Chen, Y. 2004, A simple route to aligned AlN
nanowires. Diamond and Related Materials, 13, pp. 38–41.
Xia, Y., Yang, P., Sun, Y., Wu, Y., Mayers, B., Gates, B., Yin, Y., Kim, F. and Yan, H. 2003, One-
dimensional nanostructures: synthesis, characterization, and applications. Advanced Materials,
15, pp. 353–389.
Xu, C. X., Sun, X. W. and Chen, B. J. 2004, Field emission from gallium-doped zinc oxide
nanofiber array. Applied Physics Letters, 84, pp. 1540–1542.
Yim, W. M., Stofko, E. J., Zanzucchi, P. J., Pankove, J. L., Ettenberg, M. and Gilbert, S. L. 1973,
Epitaxially grown AlN and its optical band gap. Journal of Applied Physics, 44, pp. 292–296.
Yin, L. W., Bando, Y., Zhu, Y. C., Li, M. S., Li, Y. B. and Golberg, D. 2005a, Growth and field
emission of hierarchical single-crystalline wurtzite AIN nanoarchitectures. Advanced Materi-
als, 17, pp. 110–114.
Yin, L. W., Bando, Y., Zhu, Y. C., Li, M. S., Tang, C. C. and Golberg, D. 2005b, Single-crystalline
AlN nanotubes with carbon-layer coatings on the outer and inner surfaces via a multiwalled-
carbon-nanotube-template-induced route. Advanced Materials, 17, pp. 213–217.
Youngman, R. A. and Harris, J. H. 1990, Luminescence studies of oxygen-related defects in alu-
minum nitride. Journal of American Ceramic Society, 73, pp. 3238–3246.
Zhang, J. and Zhang, L. 2002, Intensive green light emission from MgO nanobelts, Chemical
Physics Letters, 363, pp. 293–297.
Zhang, Q., Chen, G., Wu, G., Xiu, Z. and Luan, B. 2003, Property characteristics of a AlNp/Al
composite fabricated by squeeze casting technology. Materials Letters, 57, pp. 1453–1458.
Zhang, Q., Wu, G., Sun D. and Luan, B. 2002, Study on the thermal expansion and thermal cycling
of AlNp/Al composites. Journal of Materials Science & Technology, 57, pp. 63–65.
Zhang, Y. J., Liu, J., He, R. R., Zhang, Q., Zhang, X. Z. and Zhu, J. 2001, Synthesis of aluminum
nitride nanowires from carbon nanotubes. Chemistry of Materials, 13, pp. 3899–3905.
Zhao, Q., Zhang, H., Xu, X., Wang, Z., Xu, J., Yu, D., Li, G. and Su, F. 2005, Optical properties of
highly ordered AlN nanowire arrays grown on sapphire substrate, Applied Physics Letters, 86,
p. 193101.
Chapter 14
Electron Energy-Loss Spectroscopy
for Nannomaterials
C. H. Chen∗ and M. W. Chu
Electron energy-loss spectroscopy (EELS) using high-energy incident electrons has

been known to be a powerful tool for studies of electronic structure in solids. The
main energy-loss mechanisms in thin solid samples are due to single-particle exci-
tations such as interband and core-level transitions, and collective excitations such
as bulk plasmons and surface plasmons. In EELS, the intensity of inelastically
scattered incident electrons could be measured as a function of energy-losses and
scattering angles (reciprocal space) under the condition of parallel beam illumina-
tion. Energy-loss spectra obtained as a function of scattering angles can yield the
dispersion relations, i.e., the excitation energy vs. wave vector, of the corresponding
excitations. The angular (wave vector) dependence of the EELS spectra plays an
important role for us to fully understand the elementary excitations in solids.
High resolution spatially-resolved EELS spectra can be obtained from different
areas of interest in a sample when a convergent beam electron nanoprobe is utilized.
In this case, the spatial resolution is determined primarily by the electron probe
size. This is readily accomplished when EELS is performed in conjunction with a
modern scanning transmission electron microscope (STEM) which can deliver an
electron probe ∼ 0.2 nm or smaller. The combination of EELS with a STEM is a
very powerful tool for the studies of local electronic structures on an atomic-scale
for complex materials. This is particularly suitable for studies of nanomaterials in
which inhomogeneity of sizes, shapes and structures often can not be avoided. The
combination of STEM/EELS can circumvent this difficulty and allows us to study
each nano-object individually. It should be noted, however, that the use of nano-
probes under convergent beam conditions for spatially resolved EELS inevitably
collected electrons inelastically scattered through a wide angle. The measured scat-
tering cross-section is, therefore, angle-integrated in this case.
In most solids, the character of collective plasmon excitations normally can be
easily distinguished from single particle excitations such as interband and core-level
transitions. This is particularly true for metals well described by the nearly free
electron-gas model in which the energies of these two types of excitations are well
separated. Aluminum which exhibits a bulk plasmon peak at 15 eV and an inter-
band transition at 1.5 eV is an excellent example in this category. The situation,
∗ Center for Condensed Matter Sciences, National Taiwan University, Taipei, Taiwan

C Springer 2008
180 C. H. Chen, M. W. Chu
Fig. 14.1 Scattering

kinematics for EELS
however, becomes more difficult for transition and nobles metals where clear dis-
tinction between collective and single particle effects can not be made unambigu-
ously. This is mainly because the oscillator strengths for single particle transitions
involving d electrons are distributed over a rather wide energy range. The intimate
mixture of collective and single particle excitations presents new challenges to the
interpretation of the EELS spectra in this class of metals, especially for studies of
nanomaterials. We’ll first review some basic physics involved in EELS measure-
ments and this is followed by exemplifications of EELS studies of thin aluminum
films and gold nanoparticles.
A simple scattering kinematics is shown in Fig. 14.1. A parallel beam of incident
high energy electron scattered by an inelastic event resulting in a loss of energy
ω will also undergo a corresponding momentum transfer q. For typical incident
electron energies in a TEM (E0 = 200 keV) and energy transfers (typically less than
2 keV), the momentum transfer q can be approximated by,

q = K 0 θ 2 + θ E2 , where θ E = ω/2E 0
The momentum transfer q is essentially perpendicular to incident beam direction

under normal scattering conditions for bulk excitations since θ ∈ θ E . The differ-
ential inelastic scatterings cross-section of a material characterized by a complex
dielectric function ε (q, ω) is given by (Pines, 1963; Raether, 1965),

d 2σ 1 −1
≈ 2 Im , where ε = ε1 + i ε2
dωdΩ q ε (q, ω)
and

−1 ε2
Im = 2
ε (q, ω) ε1 + ε22
is the so-called energy-loss function. Well-defined maximum of the energy-loss

function occurs at a particular frequency and wave-vector when the real part of the
dielectric function is near zero and the corresponding imaginary part is small. In the
long wavelength limit when q ∼ 0, the dielectric function for a nearly free electron
gas, such as aluminum, can be well described by the Drude model,
14 Electron Energy-Loss Spectroscopy for Nannomaterials 181

ω2P 4πne2
ε(ω) = 1 − , where ω P =
ω (ω + i γ ) m
is the plasmon frequency and γ is the damping factor. The energy-loss function is
then sharply peaked at the plasma frequency where the real part of the dielectric
function crosses the frequency axis with a positive slope. The predominant inelastic
scattering cross-section is therefore due to the excitations of bulk plasmons. At finite
q, the random phase approximation for a nearly free electron
gas yields a plasmon
energy varying quadratically in q, ω = ω P + α m q 2 where α = 3E F /5ω P
is the plasmon dispersion coefficient and E F is the Fermi energy.
Let’s now look at some experimental EELS results for aluminum. We show in
Fig. 14.2 a couple of EELS spectra obtained at two different momentum trans-
fers. At q = 0, we observe a strong plasmon resonance at 15 eV consistent
with earlier experimental results and a secondary peak at 30 eV due to plural
scatterings (Batson et al., 1976). Unlike inelastic x-ray scattering, multiple scat-
tering is a common occurrence in EELS due to the much stronger interactions
with solids unless extremely thin samples (∼ a few nanometers in thickness)
are used. Removal of multiple scattering contributions in the EELS spectra is
often necessary in order to compare with theoretical calculations. In addition to
the bulk plasmon excitations, it is noted that a surface plasmon peak located at
7.0 eV is also present. Surface plasmons are charge density fluctuations localized
at the sample surfaces. The condition for the existence of surface plasmon is
that ε1 must be negative and the resonance occurs when ε1 = −1. For a clean
aluminum surface, the surface plasmon energy is then determined to take place
at 10.6 eV. Nevertheless, the presence of a thin natural oxide layer on the sur-
face is expected to push the surface plasmon energy down to 7.0 eV as observed
(Batson et al., 1976). The dispersive nature of bulk plasmon is demonstrated
clearly in the spectra taken with q = 1.18 A−1 . The plasmon peak has now
moved upward to an energy ∼ 18.5 eV and becomes much broader as expected.
The weaker and non-dispersive peak at 15 eV is the spurious peak due to mul-
tiple scatterings involving a plasmon and phonons. It is noted that the inten-
sity of surface plasmon peak has decreased drastically as the scattering angle
increases and becomes unobservable at this momentum transfer. For aluminum,
Fig. 14.2 Electron

energy-loss spectra of
aluminum with momentum
transfer q = 0 and 1.18 A−1 .
The momentum momentum
transfer is parallel to [111]
Fig. 14.3 Frequency dependent dielectric function of Au
the strongest interband transition occurs at 1.5 eV (Chen and Silcox, 1977), far
below the plasmon excitation, and its effect on the collective plasmon peak is
negligible.
The situation for gold metal is totally different. This difference is clearly demon-
strated in the frequency dependent dielectric function of Au (Schlueter, 1972;
Cooper and Ehrenreich, 1965) shown in Fig. 14.3. It certainly cannot be approx-
imated by the simple Drude model discussed above for aluminum. From the dielec-
tric function, bulk plasmon is expected to occur at 5.0 eV where the ε1 becomes zero
with a positive slope. However, this bulk plasmon resonance is strongly damped due
to the high values of ε2 , which is caused by interband transitions in the propinquity.
Below 5 eV, ε1 is negative exhibiting the highly reflective metallic characteristics
of Au and the presence of surface plasmon peak at 2.4 eV in the visible spectral
regime is well known. There exist several broad maxima in ε2 with energies higher
than 10 eV signifying the widely distributed oscillator strengths. It is also noted that,
in this deep-UV spectral regime, ε1 has becomes positive with poorly defined broad
resonances occurring in tandem with the broad maxima in ε2 . The strongest peaks
in EELS of Au are actually observed in this spectral regime, consistent with the
energy-loss function calculated from the dielectric function. It is in this deep-UV
spectral regime that we observe the unexpected excitations of surface plasmon-
like resonances in Au nanoparticles using EELS with a 0.2 nm electron probe in
near-field geometry. The 0.2 nm electron probe is generated in a FEI field-emission
STEM, Tecnai F20, operated at 200 kV.
Figure 14.4 shows a set of corresponding EELS spectra taken by placing the
0.2 nm electron probe in various locations inside or around the nanopartilces as
indicated. Spectra #1 and 5 were obtained from the center of the nanoparticles and
are predominated by the characteristic bulk excitations at 17, 25.5 and 33.5 eV. The
free electron bulk plasmon energy for the s and s plus d electrons in Au is calculated
to be ∼ 9 and 33 eV, respectively. We note that there is no any indication of the 9 eV
14 Electron Energy-Loss Spectroscopy for Nannomaterials 183
Fig. 14.4 Electron energy-loss spectra obtained from gold nanoparticles. Spectra #1 and 5 predom-
inated by bulk excitations were taken from the center of the nanoparticles and spectra #2, 3, 4 and 6
reflecting the surface excitations were obtained by placing the 0.2 nm electron probe at a distance
1 nm away from the particle surface. The diameter of the rod-like and spherical nanoparticle is 10
and 13 nm, respectively
peak in spectra #1 and 5 that can be associated with the bulk plasmon excitation due
to s electrons. The physical origin of the 17 eV peak is unclear, although interband
transition could be responsible. The 33.5 eV peak may be related to the bulk plasmon
excitations of s plus d electrons, but the significant oscillator strengths of interband
transition are still present at this energy regime and the effective number of valence
electrons participating the 33.5 eV excitation are still far below the total 11 electrons
in the s and d bands.
Spectra #2, 3, 4, and 6 were taken with the probe placed outside the nanoparticles
and ∼ 1 nm away from the surface. In this near-field geometry, only surface exci-
tations are possible due to the evanescent wave-fields of excitations confined at the
surface and no bulk excitations are expected. The most noticeable feature in these
spectra is the appearance of the excessive intensities ∼ 10 eV, absent in the bulk
spectra of #1 and 5. Furthermore, all the bulk peaks at 17, 25.5 and 33.5 eV shown in
the spectra #1 and 5 appear to have shifted to lower energies by about 2 eV, but slight
variations of the peak positions in the spectra due to different geometrical configu-
rations around the probe positions are discernible. The appearance of these surface
plasmon-like features above 10 eV is not expected in view of the positive dielectric
function ε1 shown in Fig. 14.3, which indicates that Au is no longer metallic and
behaves like an insulator in this spectral regime. Conventional surface plasmons, as
mentioned earlier, are known to exist only in the metallic spectral regime where ε1
is negative. It should be noted that the surface plasmon-like excitations above 10 eV
are closely linked to the presence of interband transition maxima in the imaginary
part of the dielectric function shown in Fig. 14.3. A positive value of ε1 alone is
not sufficient to warrant the appearance of these surface excitations. For instance,
although ε1 becomes positive above the bulk plasmon energy at 15 eV in aluminum,
no surface excitations similar to what we just described for Au nanoparticles are
expected since there are no interband transitions of any significant oscillator strength
in this spectral regime, which will link to resonance peaks in ε2 .
The intricate mixing of single particle characteristics in a collective Surface
excitation calls for a new thinking of elementary excitations in solids. A new
physical picture regarding the presence of surface plasmon-like excitations in the
non-metallic spectral regime seems necessary to grasp the full significance of these
occurrences observed in electron energy loss spectroscopy.
In this manuscript we show an application of EELS/STEM for Au nanoparticles.
The ability to obtain local electronic structure with a 0.2 nm spatial resolution makes
this technique ideal for studies for individual nanoparticle in a, otherwise, inhomo-
geneous nanomaterials. One can also explore different excitation modes governed
by the local geometry of the nanoparticle by placing the electron nanoprobe at the
particular points of interest. Moreover, coupled excitation modes between nanopar-
ticles can also be studied. It is expected some coupled modes can only be excited by
EELS and not by any optical method. EELS can also be used to measure bandgaps
of semiconductors and insulators. Possible local bandgap fluctuations in nanomate-
rials, which would be an impossible task for most experimental measurements, can
be readily explored by the EELS/STEM technique.
References
Batson, P. E., Chen, C. H. and Silcox, J., 1976, Phys. Rev. Lett. 37, p. 937.
Chen, C. H. and Silcox, J., 1977, Phys. Rev. B16, p. 4246.
Cooper, B. R. and Ehrenreich, H., 1965, Phys. Rev. 138, p. A494.
Pines, D., 1963, Elementary Excitations in Solids (New York-Amsterdam, W. A. Benjamin Inc.)
Raether, H., 1965, Mod. Phys. 38, p. 85.
Schlueter, M., 1972, Z. Physik 250, p. 87.
Chapter 15
Fabrication of Photovoltaic Devices
by Layer-by-Layer Polyelectrolyte
Deposition Method
Wai Kin Chan∗ , Ka Yan Man, Kai Wing Cheng and Chui Wan Tse
15.1 Introduction
Layer-by-layer (LbL) electrostatic self-assembly is a versatile method for prepara-

tion of polymer multilayer thin films (Decher and Hong, 1991). This method has
been employed in the preparation of different functional polymer films for a variety
of applications. In this paper, we report the fabrication of different photovoltaic
devices using the layer-by-layer polyelectrolyte deposition method. Compared to
other film forming techniques in polymers (e.g. spin-coating), this method can yield
multilayer thin films with highly reproducible thickness (resolution up to few nm)
easily. The control of device thickness is a very important issue in the fabrica-
tion of photovoltaic cells because film thickness will affect the device performance
strongly. On one hand, a thick film is desirable because it can maximize the number
of photons absorbed by the film. On the other hand, increasing film thickness also
results in the increase in electrical resistance. Optimization of device performance
requires a careful control in device thickness. LbL polymer film deposition is easily
to perform, and the waste of materials can be kept at minimum. When a photo-
sensitizing polymer is co-deposited with a charge transport polymer, the polymer
molecules essentially form an interpenetrating network (Schlenoff et al., 1998), in
which charges can be transported in these network once the excitons are separated.
In this work, the photosensitizing material is based on a ruthenium bisterpyridine

complex containing conjugated polymer (polymer 1) (Ng and Chan, 1997), which
serves as the polycation. Ruthenium polypyridine complexes have been known to
be efficient photosensitizers. The polymer exhibits a strong absorption at 500 nm,
which is due to the metal-to-ligand charge transfer transition. The excited states
are of triplet in nature, which may facilitate the exciton dissociation process. The
anionic charge transport polymer is sulfonated polyaniline (SPAN) (Yue et al.,
1991). The structures of these polymers are shown in Fig. 15.1.
∗ Department of Chemistry, The University of Hong Kong, Pokfulam Road, Hong Kong

C Springer 2008
186 W. K. Chan et al.
Fig. 15.1 Polymers used in device fabrication
Multilayer polymer films were fabricated by immersing the indium-tin-oxide

ITO glass substrates in the two polymer solutions in sequence. Since polymer 1
is only soluble in DMF and SPAN is only soluble in water, the substrates have to
be immersed in another transition solvent (ethanol) before transferring the substrate
from an aqueous to DMF solution in order to prevent precipitation of polymers. The
schematic diagram of the film deposition process is shown in Fig. 15.2.
The polymer film deposition conditions were studied in detailed by varying
the solution pH, salt concentration, and post-deposition treatment (Decher and
Schlenoff, 2003). Multilayer thin films with 20 bilayers were prepared under dif-
ferent conditions. After deposition, all the films were dried under vacuum at room
temperature for 1 day. The gradual increase in polymer film thickness was monitored
by different methods including optical absorption, quartz crystal microbalance, and
spectroscopic ellipsometry.
Table 15.1 summarizes the effect of varying the deposition conditions to the film
properties. It can be seen that the rms roughness of the films are not affected by the
Fig. 15.2 Schematic diagram showing the polymer film deposition process
15 Photovoltaic Devices by Layer-by-Layer Polyelectrolyte Deposition Method 187
Table 15.1 Effect of deposition conditions

pH of SPAN Electrolyte used Thickness (nm) rms roughness
solution (nm)
3 – 58 2.1
3 KI 48 2.2
3 KPF6 29 2.4
10 – 42 2.5
10 KI 38 2.2
10 KPF6 9 2.2
pH and presence of electrolyte in the solution. However, the thicknesses of the mul-
tilayer films obtained are strongly dependent on the presence of salt in the polymer
solutions. In general, those polymer films prepared in salt-free solutions are thicker
than those deposited in solutions with electrolytes, and KPF6 yielded polymer films
with smallest thickness. This may be related to the conformation change in polymer
molecules in different salt solution.
Photovoltaic cells with device structure ITO/(polymer 1/SPAN)n /Al were fab-
ricated and their photocurrent responses under illumination of simulated solar light
were studied. When polymer 1 and SPAN was co-deposited into a multilayer thin
film, they did not form distinct and stratified layers. Instead, the polymer molecules
in a particular layer may disperse up to 3 or 4 layers away from its nominal loca-
tion within the thin film. The morphology of the film resembles that of a bulk
heterojunction photovoltaic cell. The current-voltage characteristics of the device
ITO/(polymer 1/SPAN)20 /Al under illumination with simulated AM1.5 solar light
are shown in Fig. 15.3. Compared to the dark current, the photocurrent response
increased by 3 orders of magnitude upon light irradiation. The short circuit cur-
rent density Jsc , open circuit voltage Voc , fill factor FF, and power conversion
efficiency ηp of the device were measured to be 32 μA/cm2 , 0.68 V, 0.21, and
4.4 × 10−3 %, respectively. Table 15.2 summarizes the photovoltaic performance of
the devices fabricated from polymer 1/SPAN multilayer films with different number
of bilayers. It can be seen that there is no significant difference in fill factors among
the devices with different number of bilayers. The open-circuit voltages decrease
Fig. 15.3 Current-voltage

characteristics of the device
ITO/(polymer 1/SPAN)20 /Al
under illumination with
simulated AM1.5 solar light
188 W. K. Chan et al.
Table 15.2 Summary of the photovoltaic device performance

Number of Short circuit Open circuit volt- Fill factor FF Power conversion
bilayers n current density age Voc (V) efficiency
Jsc (μA/cm2 ) ηp (10−3 %)
20 32.9 0.68 0.21 4.4
40 11.4 0.54 0.22 1.4
60 7.5 0.54 0.24 0.95
40 0.72 0.36 0.21 0.56
slightly with increasing the number of bilayers, which might be due to the exciton
recombination. For the device with 20 bilayers, it exhibits the best performance.
Compared to the device with thicker active layer, increasing the multilayer thickness
results in enhancement in optical absorption, the electrical resistance of the film also
increased.
In order to study the role of photosensitizers in photovoltaic process, we mea-
sured the incident photon to electron conversion efficiency (IPCE) of the devices at
different wavelength. IPCE is the number of electron generated by the device per
photon absorbed, which is defined as:
1240 × Jλ (A/cm 2 )
I PC E (%) = × 100% (15.1)
Pin (W/cm 2 ) × λ(nm)
where Jl is the current density, Pin is the incident light power, and l is the wavelength
of the incident light. From the IPCE vs. l plot, we are able to study which functional
group in the polymer is responsible for the sensitization process. This provides
important information in modifying the polymer structure for device optimization.
Figure 15.4 shows the plot of IPCE vs. wavelength and the absorption spectrum of
the multilayer film for the device ITO/(polymer 1/SPAN)20 /Al. The peak observed
at ca. 500 nm clearly demonstrates that the generation of photocurrent is mainly
contributed by the ruthenium complex in the polymer.
Fig. 15.4 Plot of IPCE as the

function of incident light
wavelength and the
absorption spectrum
for the device
ITO/(polymer 1/SPAN)20 /Al
15 Photovoltaic Devices by Layer-by-Layer Polyelectrolyte Deposition Method 189
15.3 Conclusions
A series of photovoltaic cells were fabricated by the layer-by-layer deposition pro-

cess using a ruthenium containing polymer and sulfonated polyaniline. This is a
versatile approach in preparing polymer thin film devices because the film thickness
can be accurately controlled. It was found that the film properties are dependent on
the fabrication conditions, and the generation of photocurrent was mainly due to
the ruthenium terpyridine complexes. These devices showed photocurrent response
upon irradiation of simulated solar light, and the power conversion efficiencies are
in the order of 10−3 %. The low efficiency measured may be due to the low optical
absorbance and exciton recombination in the films. In addition, the effect of the
presence of electrolytes in the films is not well understood. More detailed work will
be required in order to understand the fundamental mechanisms in the photosensiti-
zation, exciton separation, and charge transport processes in the film.
Acknowledgments The work is substantially supported by the Research Council of The Hong
Kong Special Administrative Region, China (Project Nos. HKU 7009/03P, 7008/04P, and RGC
Central Allocation HKU2/05C). Financial supports from the Committee on Research and Confer-
ence Grant (University of Hong Kong) are also acknowledged.
References
Decher, G. and Hong. J. D., 1991, Buildup of ultrathin multilayer films by a self-assembly process.
1. Consecutive adsorption of anionic and cationic bipolar amphiphiles on charged surfaces.
Makromolekulare Chemie-Macromolecular Symposia 46, pp. 321–327.
Decher, G. and Schlenoff, J. B., 2003, Multilayer Thin Films: Sequential Assembly of Nanocom-
posite Materials (Weinheim: Wiley-VCH).
Ng, W. Y. and Chan, W. K., 1997, Synthesis and photoconducting properties of poly(p-
phenylenevinylene) containing a bis(terpyridine) ruthenium(II) complex. Advanced Materials
9, pp. 716–719.
Schlenoff, J. B., Ly, H., Li, M., 1998, Charge and mass balance in polyelectrolyte multilayers.
Journal of The American Chemical Society 120, pp. 7626–7634.
Yue, J., Wang, Z. H. et al., 1991, Effect of Sulfonic Acid Group on Polyaniline Backbone. Journal
of The American Chemical Society 113, pp. 2665–2671.
Chapter 16
Optical Properties of Arrays of Iodine Molecular
Chains Formed Inside the Channels of AlPO4 -5
Zeolite Crystals
J. T. Ye and Z. K. Tang∗
16.1 Ordered Nanostructures Synthesized Using

Zeolite Frameworks
Condensing guest molecular species in the nano pores of zeolite from the vapor
phase form array of nanostructures which replicated the periodicity of the zeolite
frameworks. By selecting a proper combination of host framework and guest mate-
rial, one can realize various nano structures with different ordered spatial assembling
which exhibit a variety of interesting characteristics (Tang et al., 1998; Koide et al.,
2001; Kurbitz et al., 2001; Li et al., 2001; Toshima et al., 2001). Examples include
semiconductor nano clusters and wires (Bogomolov, 1973, 1976; Goto et al., 1993;
Poborchii et al., 1996; Nozue et al., 1996; Poborchii, 2001), organic zeolite lasers
(Vietze et al., 1998), and zeolite-based second harmonic generators (Herance et al.,
2004).
16.2 Fabrication of Iodine Molecular Chains Inside

AlPO4 -5 Zeolite
Iodine is the heaviest common halogen (atomic number: 53, atomic mass: 127)
exists as a solid at room temperature in sublimation equilibrium with its vapor.
Like other halogens in their solid phase, the iodine crystal consists of weakly bound
diatomic molecules, I2 . Iodine sublimates even at low temperature and generates
enough vapor pressure facilitating its insertion and diffusion into the zeolite chan-
nels which act as a natural template to guide the growth of iodine nanostructures.
Figure 16.1a shows the method of condensing iodine molecules to the AlPO4 -5
(IUPAC code: AFI) zeolite framework schematically. Pure AFI crystals are obtained
by calcination the tripropylamine (CH3 CH2 CH2 )3 N organic templates at 580 ◦C in
an O2 atmosphere (Qiu et al., 1989; Tang et al., 1997; Sun et al., 1999). To prepare
∗ Department of Physics and Institute of Nano-Science and Technology, Hong Kong University of
Science and Technology, Clear Water Bay, Kowloon, Hong Kong, China

C Springer 2008
192 J. T. Ye, Z. K. Tang
Fig. 16.1 (a) Schematic presentation of formation of iodine chains from iodine molecules. (b) The
diffusion process of occluding iodine molecules into the channels of AFI zeolite crystals. (I), Iodine
molecules colour the opened ends of the zeolite crystal at room temperature. (II) Iodine molecules
diffuse along the channel and color the whole crystal at around 100 ◦ C. (III) High density iodine
structures are formed inside the channel after heating the zeolite crystal in the iodine vapor at
400 ◦ C for several hours
the iodine structures in the nano pores of the AFI crystals (iodine@AFI), calcinated
AFI zeolite is dehydrated at 580 ◦C for 2 hours under a vacuum, and then mixed with
solid iodine source (BDH 99%) in a Pyrex tube inside a moisture-free (< 1 ppm)
glove box. The tube is then sealed under a vacuum of around 1 × 10−1 m bar.
As shown in the Fig. 16.1b, even at room temperature, iodine molecules enter the
channels of AlPO4 -5 zeolite crystals as evidenced by the red-brown colour at the
opened ends of the zeolite crystal. The opened ends are either the two natural facets
of the crystal or the cracks formed during the calcination process. By increasing
temperature to around 100 ◦C, iodine molecules diffused along the channels and
distributed to the whole crystal as shown in stage II. While in stage III, high den-
sity of iodine structures were occluded inside the channels of the AFI crystal after
heating the crystal at around 400 ◦C for several hours. The iodine@AFI crystal is
optically isotropic when the iodine density is relatively low inside the channel, while
optical anisotropy is observed when the density becomes higher (Ye et al., 2006).
As shown in stage III, the iodine@AFI crystal shows a uniform dark purplish color,
indicating a higher iodine density.
16.3 Optical Characterization of Iodine Chains
In this section, optical measurements of the ordered one-dimensional array of stable

iodine molecular chains formed by iodine molecules adsorbed inside the channels
of AFI zeolite are discussed. Polarized Raman scattering showed that these chains
were almost perfectly aligned to the c-axis of the AFI crystal. Two chain species,
In− and (I2 )n , were distinguished by measuring polarized optical absorption and
resonant Raman scattering.
16 Properties of Arrays of Iodine Molecular Chains 193
16.3.1 Raman Scattering From iodine@AFI

with Different Densities
The different loadings of the iodine molecules inside AFI zeolite were characterized
by means of Raman scattering excited using the 514.5 nm line of an Ar/Kr ion laser.
The photon energy of this laser line corresponds to a transition from the lowest
vibrational levels of the singlet electronic ground state to high vibrational levels of
a triplet excited state. In the low diffusion stage, the Raman spectrum of the iodine
species inside the zeolite channels exhibited a typical spectral feature of diatomic
iodine molecule in the vapor phase as shown in the upper panel of Fig. 16.2. This
Raman spectrum is characterized by the principal Raman signal at 212 cm−1 with
Δν = 1, and the other features at 419, 625, 829, 1032 cm−1 , . . . correspond to vibra-
tional energy levels with Δν = 2, 3, . . ., respectively. The principle vibrational
frequency of vapour phase iodine molecules is at 214 cm−1 . The slight downshift of
2 cm−1 can be explained by the increased interaction between the iodine species and
the zeolite frameworks (Wirnsberger et al., 1997, 1999; Hertzsch et al., 2002). The
Raman spectrum of low density iodine indicates that the iodine species incorporated
in the channels are molecular vapor. The structures formed inside the zeolite crystals
are very sensitive to the diffusion temperature. With increase of the diffusing tem-
perature, higher density of iodine molecules can be loaded into the zeolite channels.
At high loading density, two new Raman features at 110 and 175 cm−1 , are observed
at the low frequency side of the molecular principle vibration, indicating new iodine
structures are formed inside the AFI channels. From stage I to III, the molecular
vibration frequency shifted from 212 to 206 cm−1 , due to the increasing interaction
between the iodine molecules at high loading density. As a result of greater interac-
tion, the vibrational frequencies of Raman modes emerged in stage II are also down-
shifted from 175 to 168 cm−1 with the increase of iodine density from stage II to III.
Fig. 16.2 The upper penal shows the Raman spectrum of an iodine@AFI crystal with a low
iodine loading. The lower panel shows the evolution of the Raman spectra when the iodine density
increases from stage I to III as shown in Fig. 16.1
16.4 Polarized Raman Spectroscopy Characterization
Polarized Raman spectra of the iodine@AFI single crystals were measured on a

micro-Raman system (Jobin Yvon T64000) in a backscattering configuration using
the 514.5 nm excitation line from an Ar/Kr-ion laser. As shown in Fig. 16.3, for
both VH (polarization of the excitation laser light and the detection of Raman
scattering are perpendicular) and VV (polarization of the excitation laser light and
the detection of Raman scattering are parallel) configurations, highly polarized
Raman modes at 110 and 168 cm−1 were observed in the stage III samples. In
the VH configuration, the intensity of these two modes increases from θ = 0◦ to
θ = 45◦ and decreases from θ = 45◦ to θ = 90◦ with a maximum at θ = 45◦ .
In the VV configuration, the intensity decreases dramatically from θ = 0◦ to
θ = 90◦ with a maximum at θ = 0◦ . Obvious anisotropy of the iodine@AFI sys-
tem in the polarized Raman scattering indicates that these two anisotropic modes
are associated with one-dimensional structures formed inside the channels. The
vibration mode at 110 cm−1 can be attributed to linear I3− chains, which were
previously observed at a similar frequency in iodine solution (Kiefer and Bern-
stein, 1972), defects in CsCl (Martin, 1976), as well as liquid iodine (Magana
and Lannin, 1985). The formation of ionic chains can originate from the charge
transfer between iodine chains and crystal defects or residual carbonaceous clus-
ters inside the channels. The Raman mode at 168 cm−1 is originated from long
(I2 )n chains that have been observed at 175–185 cm−1 in liquid iodine as well
as amorphous iodine (Shanabrook et al., 1981; Magana and Lannin, 1985). It is
worth noting that the vibration frequency of long (I2 )n chains inside the AFI chan-
nels is substantially lower than that observed in liquid iodine. The red shift in the
frequency is due to longer and more stable chain structure formed inside the AFI
Fig. 16.3 Polarized Raman spectra of iodine@AFI at VH and VV configurations are shown in the
top and bottom panels, respectively. The inset shows the angle dependence of Raman intensities in
the VH and VV configurations. The relative intensities of IV H /IV V 0◦ and IV V /IV V 0◦ are fitted as
functions of cos2 θ · sin2 θ and cos4 θ, respectively
channels. The spectral width of the Raman band at 168 cm−1 is broad, suggesting
that this mode is composed of Raman scattering from (I2 )n chains of many different
lengths. It is also possible to contain contributions from species like In− with n>3
vibrating at frequencies in a similar range of the Raman spectrum (Mulazzi et al.,
1981). The series of intra-molecular vibrations observed at 209 cm−1 is attributed
to vapor-like molecular species. In iodine vapor, these modes are usually observed
at 214 cm−1 . The slight 5 cm−1 downshift can be explained by a stronger inter-
action between iodine molecules inside the zeolite channels (Wirnsberger et al.,
1997, 1999; Hertzsch et al., 2002). The vapor-like molecular iodine is expected
in a random orientation as its Raman intensity at 209 cm−1 shows no polariza-
tion dependence. Since the 6.8 Å Van der Waals radius of molecular iodine is
slightly smaller than the 7.3 Å channel diameter of AFI, it is geometrically allow-
able for iodine molecules to rotate inside the channels. However, species like In−
and (I2 )n with chains composed of more than three iodine atoms can only align
to the channel direction since a three-iodine linear chain is already wider than the
channel.
Micro polarized absorption is measured on a single crystal of iodine@AFI
which is mechanically polished into a 10 μm thick foil. As shown in Fig. 16.4a,
the optical absorption spectra have strong polarization dependence with two
absorption bands at 2.3 and 2.7 eV, respectively, exhibiting high absorption for
the light polarized parallel to the channel direction and high transparency for the
light polarized perpendicular to the channel direction. Another absorption band,
which is not sensitive to polarization, is observed at 2.38 eV. This absorption band
can be attributed to the molecular iodine inside the channels, as its absorption
energy coincides well with the B ← X transition (2.39 eV) of vapor phase iodine
molecules (Mulliken, 1971). It is intuitive to attribute the polarized absorption
bands to the chain structures. These iodine chains are isolated by the channel
walls and they are almost perfectly aligned along the channel direction. This
configuration of one-dimensional arrays has an inherent strong depolarization
effect (Ajiki and Ando, 1994). The contribution of the anisotropic bands is
significant when the polarization angle is relatively small, while it is negligible
for the spectrum at θ = 90◦ . In Fig 16.4a, the spectra at θ < 75◦ are sub-
tracted from the spectrum at θ = 90◦ to separate the polarized and unpolarized
bands since the spectrum at θ = 90◦ shows only absorption from molecular
iodine.
16.5 Resonant Raman Scattering and Micro Optical Absorption
Resonant Raman scattering is utilized to distinguish the origin of the anisotropic

absorption bands. As shown in the left panel of Fig.16.4b, Raman scattering is
excited using the laser lines with wavelength ranging from 457.9 to 647.1 nm. The
Raman bands of the iodine chains at 110 and 168 cm−1 exhibit obvious resonant
behaviour. For Raman signals of the I3− and (I2 )n chains, the resonant maxima are
observed at the laser excitations of 2.66 and 2.34 eV, respectively. As shown by
Fig. 16.4 (a) Polarized optical absorption spectra of an iodine@AFI crystal. Spectra labeled θ = 0◦
and θ = 90◦ stand for measurement when the polarization of incident light are parallel and per-
pendicular to the crystal axis, respectively. (b) Normalized resonant Raman spectra of iodine@AFI
with excitation wavelengths from 457.9 to 647.1 nm. Resonant maxima of 110 and 168 cm−1 bands
are observed at 465.8 and 530.9 nm, respectively
the intensity mapping at the bottom of Fig.16.4b, the profile of normalized Raman
intensity of these two bands versus the excitation energies indicates resonant Raman
intensities of I3− and (I2 )n chains coincide with the optical absorption band at
2.3 and 2.7 eV, respectively. The coincidence of the resonant laser energy with the
energy of the absorption band suggests that the optical absorption at 2.7 eV can be
assigned to the I3− chains and 2.3 eV to the (I2 )n chains. The anisotropic absorption
band at around 2.3 eV (540 nm) is related to the B ← X (518 nm) transition in vapor
phase iodine that shows an evident red shift (22 nm) due to the formation of (I2 )n
chains from individual I2 molecules. The absorption band of I3− corresponding to
1
the 1 → transition is at around 2.6 eV (Mulazzi et al., 1981), which is close
to the anisotropic absorption band observed at 2.7 eV.
In summery, a uniform array of one-dimensional molecular iodine chains is
formed inside the AFI zeolite frameworks using vapor phase diffusion method.
The structures of iodine inside the AFI channels are highly dependent on the den-
sity of iodine inside the channels characterized by different Raman scattering for
iodine densities at different diffusion stages. Polarization dependence of intensity of
Raman modes of iodine species indicates that the chains are aligned perfectly along
the c-axis of AFI crystal. Polarized optical absorption as well as resonant Raman
scattering reveal two chain species inside the channels of AFI zeolite: In− and (I2 )n .
Acknowledgments This research was supported by Hong Kong CERGE Grants of 602807, RGC
DAG 05/06.sc33, RPC 06/07.sc06, and HKUST President direct allocation F0204-A.
References
Ajiki, H. and T. Ando, 1994, Physica B 201, p. 349.

Bogomolov, V. N. 1973, Fiz. Tverd. Tela Leningrad 15, p. 1312.
Bogomolov, V. N. 1976, Pis’ma Zh. Eksp. Teor. Fiz. 23, p. 528.
Goto, T., Y. Nozue, et al., 1993, Mater. Sci. Eng. B 19, p. 48.
Herance, J., D. Das, et al., 2004, Chem. Phys. Lett. 395, p. 186.
Hertzsch, T., F. Budde, et al., 2002, Angew. Chem. Int. Ed. 41, p. 2282.
Kiefer, W. and H. J. Bernstein, 1972, Chem. Phys. Lett. 16, p. 5.
Koide, T., H. Miyauchi, et al., 2001, Phys. Rev. Lett. 87, p. 257201.
Kurbitz, S., J. Porstendorfer, et al., 2001, Appl. Phys. B 73, p. 333.
Li, Z. M., Z. K. Tang, et al., 2001, Phys. Rev. Lett. 87, p. 127401.
Magana, R. J. and J. S. Lannin, 1985, Phy. Rev. B 32, p. 3819.
Martin, T. P., 1976, Phys. Rev. B 13, p. 3618.
Mulazzi, E., I. Pollini, et al., 1981, Phy. Rev. B 24, p. 3555.
Mulliken, R. S., 1971, J. Chem. Phys. 55, p. 288.
Nozue, Y., T. Kodaira, et al., 1996, Magnetic Properites of Alkali Metal Cluster in Zeolite Cages.
Mater. Sci. Eng. A 217–218, p. 123–128.
Poborchii, V. V., 2001, Raman microprobe polarization measurement as a tool for studying the
structure and orientation of molecules and clusters incorporated into cubic zeolites: S8 and S12
rings in zeolite A. J. Chem. Phys. 14, p. 2707.
Poborchii, V. V., M. S. Ivannova, et al., 1996, Raman and absorption spectra of the zeolites A and X
containing selenium and tellurium in the nanopores. Mater. Sci. Eng. A 217–218, pp. 129–134.
Qiu, S., W. Pang, et al., 1989, Zeolites 9, p. 440.
Shanabrook, B. V., J. S. Lannin, et al., 1981, Phys. Rev. Lett. 46, p. 130.
Sun, H. D., Z. K. Tang, et al., 1999, Solid State Commun. 109, p. 365.
Tang, Z. K., M. M. T. Loy, et al., 1997, Appl. Phys. Lett. 70, p. 34.
Tang, Z. K., H. D. Sun, et al., 1998, Appl. Phys. Lett. 73, p. 2287.
Toshima, N., Y. Shiraishi, et al., 2001, J. Mol. Catal. 177, p. 139.
Vietze, U., O. Krasb, et al., 1998, Zeolite-Dye Microlaser. Phys. Rev. Lett. 81, p. 4628.
Wirnsberger, G., H. P. Fritzer, et al., 1997, J. Mol. Struct. 410, p. 123.
Wirnsberger, G., H. P. Fritzer, et al., 1999, J. Mol. Struct. 481, p. 699.
Ye, J. T., Z. K. Tang, et al., 2006, Applied Physics Letters 88, pp. 073114.1–073114.3.
Part IV
Molecular Electronics
Chapter 17
Quantum Manipulation at Molecule Scale
J. G. Hou∗
17.1 Introduction
The properties of matters in the single-atom and single-molecule regime and the
interactions among them are quantum mechanical in nature. Detecting and manip-
ulating the quantum states on the molecular scale are important and essential for
future applications in designing and constructing novel artificial structures and
functional nanodevices in the “bottom-up” scheme (Feynman, 1960) because they
have numbers of advantages as basic building blocks of matter and serve as good
candidates for conceptually new device components when traditional microelec-
tronics reaches the physical limit (Joachim et al., 2000). To realize novel func-
tional nanodevices, controlled manipulation of quantum properties of molecules and
nanoparticles are desired. As a local probe, scanning tunneling microscope (STM)
has shown powerful ability in physical sciences and allows us to directly detect
and manipulate single atoms, molecules and nanoparticles as well as their quantum
states at the very basic level of matters since its invention by Binnig et al. (1982a,b).
Although the concept of molecular electronics has been proposed for more than
30 years (Aviram and Ratner, 1974), to manipulate single molecules mechanically
and quantum mechanically is still not easy. The basic issue in molecular electron-
ics is to realize electronic devices employing only one or very few molecules, that
requires controllable formation of transport junctions in which a single molecule
can be precisely put between two electrodes and connect with both (Joachim and
Ratner, 2005). It is extremely difficult for modern lithography to make electrode
junctions as narrow as few nanometers, and more difficult to introduce a molecule
into such a junction and connect well with both electrodes. While, an STM in which
tip and substrate are two electrodes can provide such ability facilely owing to its
atomic lateral resolution and vertical precision, and thus becomes a powerful tool
for the study of quantum manipulation at the molecule scale. Controlled doping can
alter the electronic properties of single molecules which can be readily investigated
by STM. The facility and flexibility of STM also allows implementation of a diver-
∗ Hefei National Laboratory for Physical Sciences at Microscale, University of Science and Tech-
nology of China, Hefei, Anhui 230026, P. R. China. E-mail: [email protected]

C Springer 2008
202 J. G. Hou
sity of single-molecule manipulations using the STM tip to create novel artificial
nanostructures mechanically (Stroscio and Eiger, 1991; Eigler and Schweizer, 1990;
Neu et al., 1995; Jung et al., 1996) or chemically (Hla and Rieder, 2003; Ho, 2002;
Lorente et al., 2005; Hou and Zhao, 2006; Hou et al., 2004). In this paper we report
the quantum manipulation and the realization of some nanodevices at the molecule
scale with STM.
17.2 Manipulating the Electronic Properties of Single Molecules

by Chemical Doping
Fullerene molecules have attracted much attention in recent years due to their unique
properties in both bulk and single-molecule states. They are good candidate for
realizing nanometer molecular devices due to their nanometer structures and dis-
tinct electronic properties. Chemical doping of fullerenes will greatly change their
electronic properties which can be investigated with STM. According to the princi-
ples of electron tunneling, in the simplest case, the tunneling current is proportional
to a convolution of the DOS of the STM tip and the DOS of the sample over an
energy range eV, where V is the bias voltage applied (Chen, 1993). Therefore, if
assuming tip DOS is constant, the dI /dV spectroscopy reflects the LDOS of the
sample near E F (Lu et al., 2003b). The dI /dV technique can be used for investigat-
ing complex electronic structures of individual molecules especially those can not
be detected by common STM measurements. We showed that the energy-resolved
metal-cage hybrid states of a single √endohedrally
√ doped metallofullerene Dy@C82
isomer I adsorbed on a Si(111) − 3 × 3 − Ag surface can be spatially mapped
via STM dI /dV technique (Wang et al., 2003b), supporting a complex picture con-
sisting of the orbital hybridization and charge transfer for the interaction between
the cage and the metal atom.
In most cases, the electronic properties of single molecules adsorbed on metal
or semiconductor surfaces are strongly influenced by the substrate electrons. By
introducing another tunneling barrier, usually a thiol monolayer or a thin oxide film,
to isolate single molecules and forming a double-barrier tunneling junction (DBTJ)
configuration, the STM can take advantage of some unique and intrinsic properties
of the molecules to realize electronic devices with specific functions. We found
that the azafullerene C59 N molecule, a kind of chemically doped C60 , can be used
as a rectifier in a STM-based DBTJ via the single electron tunneling (SET) effect
(Zhao et al., 2005b). A fullerene molecule in a DBTJ show SET effect analogous
to semiconductor quantum dot with zero-current Coulomb Blockade (CB) around
the Fermi surface (Li et al., 2006). For fullerene C60 with a full-occupied high-
est occupied molecular orbital (HOMO) and lowest unoccupied molecular orbital
(LUMO), we observed a symmetrically CB behavior in the I-V curve where the
current onsets at both negative and positive sample bias voltage are approximately
equal. While, for azafullerene C59 N with a half-occupied HOMO, pronounced I-V
asymmetry was observed and the rectifying effect exists in all measurements we
17 Quantum Manipulation at Molecule Scale 203
Fig. 17.1 (a) The diagram of the experimental system. (b) The STM image of an isolated C59N
adsorbed on thiol SAM at 5 K. Image size: 8 nm × 8 nm, scanning parameters: 1.86 V/0.1 nA.
Inset shows line profile along the line AB. (c) A set of I-V curves for individual C59N molecules
measured at 5 K and at various setting parameters. The inset shows the I-V curves for the SAM
substrate and individual C60 . (From Zhao et al., 2005b, Copyright 2000 by the American Institute
of Physics)
made. The positive onset sample bias voltage was always about 0.5–0.7 V, while the
negative onset voltage was about 1.6–1.8 V (Fig. 17.1). By theoretical analysis we
found that the half-occupied HOMO and the asymmetric shift of the Fermi level
when the molecule was charged (the HOMO-LUMO gaps of C59 N− and C59 N+ are
distinctly different) were responsible for the molecular rectification.
17.3 Controlling the Transport Properties of Single Molecules

and Nanoparticles by Direct Manipulation
Direct manipulation including vertical manipulation and lateral manipulation of sin-

gle adsorbates can be used to construct novel tunneling structures such as an NDR
device involving two fullerene molecules vertically coupled. A single fullerene
molecule can be accommodated into an STM-based DBTJ system with thiol SAM
and manifest as an NDR device (Zeng et al., 2000) Atomic scale NDR effect could
arise in the tunneling structures with an STM tip and a localized surface site when
the LDOS of both the tip and the surface site have adequately narrow features.
204 J. G. Hou
Single C60 molecules are great candidates for realize NDR devices in principle
because neutral C60 molecule has sharp LDOS features owing to its degenerated
and well-separated energy states. In a DBTJ system, C60 molecules adsorbed on
a thiol SAM remain their native properties such as LDOS in the main. A tip with
distinct and narrow LDOS features was prepared by controllably attaching a single
C60 molecule onto its apex. By using such a tip we observed obvious NDR effect
on individual C60 molecules on SAM (Zeng et al., 2000) (Fig. 17.2).
Nanoparticles are also good candidates for nanoscale electronics devices. The
I-V transport properties of a nanparticle in a STM-based DBTJ system will show CB
and Coulomb staircases due to its discrete energy levels and the charging effect (Hou
et al., 2001; Wang et al., 2001, 2003a, 2004). Metal or semiconductor nanoparticles
with different sizes and properties can be readily controlled by chemical methods
(Lu et al., 2003a) and the chemical properties of the nanoparticles will greatly affect
the quantum transport properties in an STM. For example, size dependent multi-
peak spectral features in the differential conductance curve are observed for the
crystalline Pd nanoparticles but not for the amorphous particles (Hou et al., 2003)
(Fig. 17.3a–c). Theoretical analysis shows that these spectral features are related
to the quantized electronic states in the crystalline Pd particle. The suppression of
the quantum confinement effect in the amorphous particle arises from the reduction
Fig. 17.2 I-V curves obtained from four kinds of tunneling structures, as shown in the insets. (a),
(b), (c), and (d) correspond to bare Pt-Ir tip over thiols, bare Pt-Ir tip over C60 , C60 -modified tip
over thiols, and C60 -modified tip over C60 , respectively. Clear NDR effect was found in (d). (From
Zeng et al., 2000, Copyright 2000 by the American Institute of Physics)
Fig. 17.3 (a) The numerical differential conductance spectra of two 2 nm amorphous Pd and crys-
talline Pd (c-Pd) particles with various set point of the tunneling currents. (b) A STM image show-
ing an amorphous Pd particle A and a c-Pd particle B; (c) dI /dV spectra acquired on particle A and
particle B, respectively. (d) I-V curves measured on two coupled c-Pd particles with the structure
depicted as inset. The zero current gaps of curves were centered at zero bias for comparison. (From
Hou et al., 2003, Copyright 2001 by the American Physical Society)
of the degeneracy of the eigenstates and the level broadening due to the reduced
lifetime of the electronic states.
Due to the narrow features of the discrete energy levels, similar to fullerene
molecules, nanoparticles can also be used to realize NDR devices (Hou et al., 2003)
by directly manipulating a crystalline Pd nanoparticle to the STM tip (Fig. 17.3d).
17.4 Manipulating the Magnetic Properties of Adsorbed Ions

by Single-molecule Chemistry
Scanning tunneling microscope can be employed to perform single-molecule chem-

istry by injecting energetic tunneling electrons into individual adsorbed molecules
and hence selectively exciting a specific chemical bond or a vibrational mode
of the molecule. These kinds of electron induced manipulations provide more
powerful abilities in single-molecule chemical modification and allow formation
of novel artificial molecular structures and hence tuning of the properties of the
molecules.
Ligand molecules like phthalocyanines and porphrins serve as good carriers for
magnetic metal ions. Using an STM, we show that the Kondo effect of the magnetic
moment of a single adsorbed ion can be controlled by altering its chemical envi-
ronment (Zhao et al., 2005a). The Kondo effect arises from the coupling between
localized spins and conduction electrons. At sufficiently low temperatures, the cou-
pling can lead to changes in the transport properties through scattering or resonance
effects (Kondo, 1964; Kouwenhoven and Glazman, 2001) which are signatures of
the presence of magnetism.
For a cobalt phthalocyanine(CoPc) molecule, a magnetic Co2+ ion is embedded
in its center, but the magnetism totally quenched when the molecule is adsorbed
onto the Au surface due to the electronic interactions between the Co ion and the
206 J. G. Hou
substrate. The nonmagnetic behavior can be verified both from experimental dI /dV
spectroscopy and from theoretical calculations. However, after cutting away eight
hydrogen atoms from the molecule with high-energy tunneling electrons emitted
from a STM tip, both the electronic and geometric structures of the molecule were
changed remarkably. The feasibility of such a manipulation was previous demon-
strated in the dehydrogenation experiments of single benzene molecules on copper
surfaces (Komeda et al., 2004; Lauhon and Ho, 2000) in which a benzene bond-
ing geometry change from flat-lying to up-right configuration was observed and
the breaking of C-H bonds were proposed to be induced by the formation of the
transient negative ion due to the trapping of the injected electrons at the unoc-
cupied π ∗ resonant states of the benzene molecules. In our experiment, when we
apply a positive high voltage (>3.3 V) over the edge of a lobe, the I-t curve shows
two sudden drops indicating the sequential dissociation of the two end H atoms
from the benzene ring in the lobe (Fig. 17.4). After the dehydrogenation process,
the lobe disappeared in topographic image, indicating a strong chemical bonding
of the highly-reactive benzene ring to the Au surface. After pruning off all eight
hydrogen atoms from the periphery of the CoPc molecule, all the four lobes disap-
peared and the molecule arched away from the surface. In this fully-dehydrogenated
CoPc (d-CoPc) molecule, we found a strong resonance at EF arised in the dI /dV
spectroscopy, however, the original dz 2 -orbital induced resonant state located at
–150 mV in a pristine CoPc molecule now completely disappeared (Fig. 17.5).
This zero-bias resonance was attributed to the Kondo effect associated with Fano
resonance by careful theoretical analysis. First-principles calculations and simu-
lations showed the cobalt atom in a free CoPc has unpaired d electrons and the
Fig. 17.4 (A) individual CoPc molecules adsorbed on Au(111) surface. (B) Diagram of the dehy-
drogenation induced by the tunneling electrons. (C) Current traces during two different voltage
pulses on the brink of one lobe. Inset shows the molecular structure of CoPc and the two hydrogen
atoms to be pruned off in each one lobe. (D to H) STM images (25Å by 25 Å, V = 1.2 V, I =
0.4 nA) of a single CoPc molecule during each step of the dehydrogenation process, from (D) an
intact CoPc to (H) a final dehydrogenated CoPc (d-CoPc). The color scale represents apparent
heights, ranging from 0 Å (low) to 2.7 Å (high)
Fig. 17.5 (A) dI /dV spectra measured directly over center Co atoms of CoPc and d-CoPc. All
spectra were taken with the same set point of V = 600 mV and I = 0.4 nA. (B and D) Side views of
the optimized structures for CoPc/Au(111) and d-CoPc/Au(111) adsorption systems, respectively.
(C and E) The PDOS of the Co atom in a CoPc and a d-CoPc molecule on an Au(111) surface,
respectively. Insets are the corresponding simulated STM images
magnetic moment is 1.09 Bohr magnetons (μB). But in the CoPc/Au(111) adsorp-
tion system, the magnet moment of the Co atom is completely quenched by the
molecule-substrate interaction. In the d-CoPc/Au(111) artificial molecular struc-
ture, the magnet moment of the Co atom is recovered to be 1.03 μB, very close
to the value of a free CoPc molecule. This recovered localized spin in the Co
atom as well as the chemical bonding of the d-CoPc ligand lead to the forma-
tion of Kondo ground state at low temperatures and result in the observed Kondo
effect.
These experiments of electron-excited single-molecule manipulation demon-
strate the feasibility of an STM in manipulating not only the electronic properties
but also the magnetic properties of a single molecule that promises potential appli-
cations in future molecular electronics and spintronics on the atomic level.
17.5 Summary
In conclusion, we try to manipulate the electronic structures and the transport prop-
erties of single molecules and nanoparticles by controlled doping, direct manipula-
tion, and single-molecule chemistry. By introducing single molecules or nanoparti-
cles into the tunnel junctions in which the tip of STM and supporting surface are two
electrodes, quantum transport properties were studied and their electronic structures
were determined from the results of STM combined with the theoretical modeling.
Novel effects associated with the discrete energy levels and the charging energy of
ultra small tunneling junction, like NDR effect, rectifying and Kondo effect at the
molecular scale are reported.
Acknowledgments This work is partially supported by the Ministry of Science and Technology of
China (grant nos. G1999075305, G2001CB3095, and G2003AA302660) and the National Natural
Science Foundation of China.
208 J. G. Hou
References
Aviram A., Ratner M. A., 1974, Chem. Phys. Lett. 29, p. 277.
Binnig G., Rohrer H., Gerber C., Weibel E., 1982a, Phys. Rev. Lett. 49, p. 57.
Binnig G., Rohrer H., Gerber C., Weibel E., 1982b, Appl. Phys. Lett. 40, p. 178.
Chen C. J., 1993, Introduction to Scanning Tunneling Microscopy (Oxford University Press,
New York).
Eigler D. M., Schweizer E. K., 1990, Nature 344, p. 524.
Feynman R. P., 1960, Engineering and Science, (California Institute of Technology, Pasadena),
p. 22.
Hla S. W., Rieder K. H., 2003, Annu. Rev. Phys. Chem. 54, p. 307.
Ho W., 2002, J. Chem. Phys. 117, p. 11033.
Hou J. G., Zhao A. D., 2006, NANO 1, p. 15.
Hou J. G., Wang B., Yang J., Wang K., Lu W., Li Z., Wang H., Chen D. M., Zhu Q., 2003, Phys.
Rev. Lett. 90, p. 246803.
Hou J. G., Wang B., Yang J., Wang X. R., Wang H., Zhu Q., Xiao X., 2001, Phys. Rev. Lett. 86,
p. 5321.
Hou J. G., Zhao A. D., Huang T., Lu S., 2004, in: H. S. Nalwa, (Ed.) Encyclopedia of Nanoscience
and Nanotechnology, Vol. 1, (American Scientific Publishers, New York), p. 409.
Joachim C., Ratner M. A., 2005, Proc. Natl. Acad, Sci. USA 102, p. 8801.
Joachim C., Gimzewski J. K., Aviram A., 2000, Nature 408, p. 541.
Jung T. A., Schlittler R. R., Gimzewski J. K., Tang H., Joachim C., 1996, Science 271, p. 181.
Komeda T., Kim Y., Fujita Y., Sainoo Y., Kawai M., 2004, J. Chem. Phys. 120, p. 5347.
Kondo J., 1964, Prog. Theor. Phys. 32, p. 37.
Kouwenhoven L., Glazman L., 2001, Physics World 14(1), p. 33.
Lauhon L. J., Ho W., 2000, J. Phys. Chem. A 104, p. 2463.
Li B., Zeng C., Zhao J., Yang J., Hou J. G., Zhu Q., 2006, J. Chem. Phys. 124, p. 064709.
Lorente N., Rurali R., 2005, Tang H., J. Phys.: Condens. Matter 17, p. S1049.
Lu W., Wang B., Wang K., Wang X., Hou J. G., 2003a, Langmuir 19, p. 5887.
Lu X., Grobis M., Khoo K. H., Louie S. G., Crommie M. F., 2003b, Phys. Rev. Lett. 90, p. 096802.
Neu B., Meyer G., Rieder K. H., 1995, Mod. Phys. Lett. B 9, p. 963.
Stroscio J. A., Eiger D. M., 1991, Science 254, p. 1319.
Wang B., Wang H., Li H., Zeng C., Hou J. G., Xiao X., 2001, Phys. Rev. B 63, p. 035403.
Wang B., Wang K., Lu W., Wang H., Li Z., Yang J., Hou J. G., 2003a, Appl. Phys. Lett. 82, p. 3767.
Wang K., Zhao J., Yang S., Chen L., Li Q., Wang B., Yang S., Yang J., Hou J. G., Zhu Q., 2003b,
Phys. Rev. Lett. 91, p. 188504.
Wang B., Wang K., Lu W., Yang J., Hou J. G., 2004, Phys. Rev. B 70, p. 205411.
Zeng C., Wang H., Wang B., Yang J., Hou J. G., 2000, Appl. Phys. Lett. 77, p. 3595.
Zhao A., Li Q., Chen L., Xiang H., Wang W., Pan S., Wang B., Xiao X., Yang J., Hou J. G., Zhu Q,
2005a, Science 309, p. 1542.
Zhao J., Zeng C., Cheng X., Wang K., Wang G., Yang J., Hou J. G., Zhu Q., 2005b, Phys. Rev.
Lett. 95, p. 045502.
Chapter 18
Silicon-Based Nano Devices for Detection
of DNA Molecules
M. C. Lin, C. J. Chu, C. J. Liu, Y. P. Wu, C. S. Wu, and C. D. Chen∗ ,

I. C. Cheng, L. C. Tsai, H. Y. Lin, Ya-Na Wu, Dar-Bin Shieh, H. C. Wu
and Y. T. Chen
18.1 Introduction
Silicon-On-Insulator (SOI) based nanowire field-effect-transistors were used to

detect the molecular charges on the wire surface. The sensors hold great potential
for various biomedical applications (Li et al., 2004) due to their intrinsic advantages
in many aspects including high sensitivity with minimal requirement of the target
molecules, capability for multiplex parallel processing, and requiring no tedious
sample labeling process. The variations of the molecular charges can be detected by
measuring the changes of the wire conductance.
In this work, we show that when incorporated with surface APTES molecules, the
SOI-based silicon nanowires can act as a detector for DNA-hybridization. The high
charge sensitivity of the nanowires allowed real time detection of hybridization of
ssDNA with lengths as short as 15 mer, and with high degree of specificity. Details
about device and microfluidic channel fabrication processes, and procedures for sur-
face modification, DNA immobilization are provided in the following sections.
18.1.1 Fabrication of SOI-Based Silicon Nanowire Devices
The p-type top Si layer of the four-inch SOI wafer was used to make the nanowire
device. Micro-sized leads and contact pads were first patterned by photolithograph
and inductively coupled plasma (ICP) etching. These structures provided a bridge
between micro-meter scaled leads and nano-scaled nanowires, and allowed electri-
cal transport measurements of the nanowires. Shown in Fig. 18.1 are photoresist
patterns defining micrometer sized leads. The leads were then formed after an ICP
etching process. In addition to these leads, cross structures were defined at the same
step, and they severed as electron beam exposure alignment keys. Boron ion implan-
tation was carried out to reduce the resistance of the leads.
The regions where nanowires are to be defined were protected by the photore-
sist during the ion implantation process. The ion beam with 11 KeV energy was
∗ Institute of Physics, Academia Sinica, Nankang 11529, Taipei, Taiwan

C Springer 2008
210 M. C. Lin et al.
Fig. 18.1 Optical micrographs of the SOI-based silicon nanowire devices. (a) The resist mask
for ion-implantation. The cross-patterns are alignment keys for e-beam lithography processes.
(b) Silicon nanowires after e-beam lithography process and ICP etching. Each chip consists eight
3μm-long silicon nanowires with widths vary from 70 to 800 nm
employed and the implantation depth in to the top Si-layer was estimated to be
about 20 nm, and the concentration of dopant was controlled to about 2 × 1014/cm2 .
Subsequently, the nanowires (shown in Fig. 18.1b) were defined by electron beam
lithography, and formed by ICP etching process. To ensure a good isolation between
the devices and the molecular solutions that are to be introduced during the sensing
measurements, both Si nanowires and Si leaders were covered by a ∼10 nm-thick
thermally grown silicondioxide, which was formed in 900 ◦C dry oxygen ambient
for 20 min. The SiO2 layer could not only provide electrical isolation with solution,
but also act as a buffer layer for later surface chemical modification. As the last step
of the device fabrication, the surface oxide layer on the outside contact pads was
removed by buffer oxide etch process, and 50 nm-thick Au buffered by 10 nm-thick
Ni serving as Ohmic contact metals were thermally deposited using the same
resist mask.
18.1.2 Characterizations of Unmodified Silicon Nanowires
The fabricated silicon nanowires were characterized by DC electric measurement,

and showed field effect transistor behaviors. A bias voltage Vd was applied on
the drain electrode while the source electrode was grounded, and a gate voltage
Vg was applied on the back-gate electrode fabricated on the backside of the bulk
Si chip. Shown in Fig. 18.2a is the measured silicon nanowire, which is 200 nm
wide and 3μm-long. Figure 18.2b shows drain-source current Id as a function
of back-gate voltage Vg measured at various drain voltage Vd . The Current Id
increased with decreasing gate-voltage Vg , indicating that the silicon nanowire is
of p-type. Figure 18.2c shows a comparison between silicon nanowires of different
widths. It is found that the drain currents at a given biasing condition were not
proportioned to the wire width; this may be attributed to the inevitable presence
18 Silicon-Based Nano Devices for Detection of DNA Molecules 211
Fig. 18.2 (a) Schematic of the device structure with an SEM image of the measured silicon
nanowire which is 200 nm in width and 3 μm in length. (b) The Id − Vg characteristics measured
with the drain voltage Vd ramped from −1 to 1 V with a step of 0.1 V. The device shows p-type FET
behavior. (c) Id − Vg characteristics for wires with different widths (70, 100, 200, 400, 800 nm),
represented by different colors. (d) Gate voltage dependence of the transconductance. The charge
sensitive is at the maximum at the gate voltage corresponding to the maximum transconductance
of random distribution of impurities/defects. Figure 18.3d shows derivative of the

curves in Fig. 18.3c in respect to the gate voltage, i.e. dId /dVg , which is named
as transconductance. The carrier mobility is related to the transconductance as
μ = (L 2 /Vd C)(dId /dVg ), where L is the wire length and C is the wire capaci-
tance (Vashaee et al., 2006) to the backgate electrode. This equation, together with
estimated capacitance values and the data shown in Fig. 18.3d, yields a mobility
value of about 40 cm2 /V s. This is small compared to the bulk silicon values, but is
comparable with that of SOI-based silicon films (Ohata et al., 2006).
18.1.3 Aptes Modification Procedures
In this process, chips with Si-nanowires were first oxidized with 2% cholic acid
(C24 H40 O5 , Fluka, BioChemika, purity ≥ 99%) in alcohol for 12 h to generate−OH
on the SiO2 surface. As shown in Fig. 18.3, the chips were then silanized with 2%
APTES (from Sigma-Aldrich, purity 99%) in acetone at 80 ◦ C for 2 h to form a
single layer of APTES with the amino functional group (−NH2 ) on the surface
(Fixe et al., 2006; Sakata and Miyahara, 2005; Xiang et al., 2004). The devices
were subsequently cleaned with distilled H2 O, blow-dry with nitrogen to remove
Fig. 18.3 A schematic drawing of the APTES surface modulation procedures
unbound APTES and dried at 110 ◦ C for 1 h. Self-assembly of APTES layer was
confirmed with a fluorescent microscope using fluorescein-5-isothiocyanate (FITC)
as dye which reacts specifically with NH2 to form covalent bonds.
18.1.4 Immobilization of Capture DNA Molecules
In this process, 15-mer poly-T ssDNA molecules acting as the capture DNA for sub-
sequent hybridization test was attached to the surface of a APTES-modified 70 nm-
wide n-type silicon nanowire. The amino functionalized substrates were modified
with cross-linker glutaraldehyde (25% in 0.1M phosphate buffer, pH 7.4) at room
temperature for 1 h to let one terminal of glutaraldehyde forming an amido bonds
-CONH- with APTES and left the other terminal available for further immobiliza-
tion with single strand DNA (ss-DNA). After cross-linking, 1 μM 15-mer poly-T
ssDNA probe solution was added and stayed for 2 h to allow for linkage between
DNA probe at the 5 end and the terminal aldehyde group on the wire surface. The
unbound DNA molecules were then washed away by large amount of phosphate
buffered solution (PBS).
18.1.5 Fabrication of Microfluidic Channels
In order to detect molecule reactions, such as DNA hybridization, the silicon

nanowire sensors were placed inside a microfluidic channel. The channels were
made of polydimethylsiloxane (PDMS) using SU8 thick photoresist as mold
a
b
200 m
c d
20 μm
150 m
adapter
e f
PDMS
PDMS
SOI
Fig. 18.4 (a) Photo-mask of the microfluidic channel. (b) SEM image and (c) Optic microscope
image of the SU8 mold. (d) SEM image of the finished PDMS structure which is a reverse of the
mold shown in (b). The images (b) and (d) are taken at the cross-marks in (a). (e) PDMS slab
with microfluidic channel on the bottom and tube adapter on the top. (f) Schematic of the fluid
system setup
(Ryu, 2004). The fabrication procedures are as follows: The SU8 mold was made
on silicon chips by standard photolithography process. SU8-2007 resist was spun
on silicon chip at a speed of 3000 rpm for 30 s, which gave a thickness of about
8 μm after a 30 min curing at 65 ◦ C, followed by a second 4-h curing at 95 ◦C. After
exposure with the photo-mask shown in Fig. 18.4a, and post-exposure baked at
95 ◦ C for 30 min, the resist was subjected to a develop process. This produced SU8
molds shown in Fig. 18.4(b) and 4c. Separately, PDMS base was mixed with curing
agent with a ratio of 10:1, and was poured into a stainless steel container with the
SU8 mold placed at the bottom. The resultant PDMS fluidic channel is shown in
Fig. 18.4(d), which was 8 μm high and 10 μm wide. As shown in Fig. 18.4e, the
PDMS slab was incorporated with two 800 μm-OD Teflon fluid tube adapts on
the top, which connected the tube to the microfluidic channels. Figure 18.4f is a
schematic drawing of the fluid circuit.
18.1.6 Detection of DNA Hybridization
The 70 nm-wide silicon nanowire with capture ssDNA on the surface was used as a
sensor for hybridization of complementary ssDNA. Real-time resistance response to
the DNA hybridization is shown in Fig. 18.5. The resistance value was taken using a
lock-in amplifier with an AC excitation of 80 mV and 0.777 Hz. Initially, the silicon
Fig. 18.5 Measured nanowire resistance in response to the injection of ssDNA molecules. Prior
to the measurement the wire was immobilized with poly-T ssDNA. Poly-C DNA (20 pM) and
buffer solution were added at t = 100 and 2000 s, respectively, as controls. At 2700 s ploy-A DNA
(20 pM) was added, and an abrupt increase in resistance was observed
nanowire device was immersed in a buffer solution of pH 8. At t = 100 s, 20 pM

15-mer poly-C ssDNA molecules was introduced, which should not interact with
the capture poly-T ssDNA, and the wire showed no obvious change in the resistance
as expected. The unpaired poly-C ssDNA was then removed by buffer solution at
t = 2000 s. Then the target DNA, that is 15-mer poly-A ssDNA molecules with a
concentration of 20 pM, was introduced at t = 2700 s and the wire showed a rapid
increase in the resistance. Since DNA molecules are negatively charged in pH 8
buffer solution, hybridization of two ssDNA simply implies an increased negative
charge on the wire surface, which then results in depletion of conduction electrons
in the n-type silicon nanowires. This is manifested in an increased wire resistance
as shown in our experiment. This result demonstrates detection of hybridization
between very short DNA molecules using silicon nanowires.
18.2 Conclusions
In this study, we developed techniques for fabrication of nanometer-wide SOI-based

silicon wires and PDMS micro fluid channels. Furthermore, conditions for APTES
modification of the silicon nanowire surface as well as immobilization of capture
DNA molecules were explored. With these technologies, we investigated the poten-
tial of using silicon nanowires as a detector for DNA hybridization. Our experiment
result showed that detection of hybridization between complementary ssDNA with
lengths as short as 15 mer is possible with a 70nm-wide SOI-based silicon nanowire.
References
Fixe, F., Dufva, M., Telleman, P. and Christensen, C. B. V., 2004, Functionalization of poly(methyl
methacrylate) (PMMA) as a substrate for DNA microarrays. Nucleic Acids Research, 32, p. e9.
Li, Z., Chen, Y., Li, X., Kamins, T. I., Nauka, K. and Williams, R. S., 2004, Sequence-Specific
Label-Free DNA Sensors Based on Silicon Nanowires. Nano Letters, 4, pp. 245–246.
Ohata, A., Cristoloveanu, S. and Cassé, M., 2006, Mobility comparison between front and back
channels in ultrathin silicon-on-insulator metal-oxide- semiconductor field-effect transistors
by the front-gate split capacitance-voltage method. Applied Physics Letters, 89, pp. 032104–
032106.
Ryu, K. S., Wang, X., Shaikh, K. and Liu, C., 2004, A Method for Precision Patterning of Silicone
Elastomer and Its Applications. Journal of Microelectromechanical Systems, 13, pp. 568–575.
Sakata, T. and Miyahara, Y., 2005, Potentiometric Detection of Single Nucleotide Polymorphism
by Using a Genetic Field-effect transistor. ChemBioChem, 6, pp. 703–710.
Vashaee, D., Shakouri, A., Goldberger, J., Kuykendall, T., Pauzauskie, P. and Yang, P., 2006, Elec-
trostatics of nanowire transistors with triangular cross sections. Journal of Applied Physics, 99,
pp. 054310–054314.
Xiang, J., Zhu, P., Masuda, Y. and Koumoto, K., 2004, Fabrication of Self-Assembled Mono-
layers (SAMs) and Inorganic Micropattern on Flexible Polymer Substrate. Langmuir, 20,
pp. 3278–3283.
Chapter 19
From Simple Molecules to Molecular Functional
Materials and Nanoscience
Vivian Wing-Wah Yam and Keith Man-Chung Wong
19.1 Introduction
The rational design and synthesis of molecular and nano-scale materials have
attracted growing attention in recent years owing to their enormous and unpre-
dictable potentials in molecular devices and machineries, and nanoscopic sciences.
Apart from the design of the discrete molecules and an understanding of the
structure-property relationships based on the molecules themselves, an exploration
of the work beyond the molecular level towards the supramolecular and nanoscopic
scale would represent challenging areas of research.
By combining the advantages of high synthetic versatility and easy processibility
in the molecular properties, together with an understanding of the structure-property
relationship established, molecular functional materials could be readily prepared
by the approach of “function by design”. Moreover, it would also be interesting
to study how the properties in the bulk might be related to those at the nano and
molecular level and whether some insights could be provided through a structure-
property relationship study. All these would be beneficial to the future design of
functional materials in the area of nanoscience and materials.
In this report, the design and synthesis of a number of transition metal com-
plex molecules including Cu(I), Ag(I), Au(I), Au(III), Re(I), Pt(II) and Ru(II) have
been described. All the complexes have been found to exhibit strong lumines-
cence. Their fundamental photophysical properties including electronic absorption
and luminescence behaviors have been studied and their spectroscopic origins have
been elucidated. Through rational design, judicious functionalization and assem-
bly strategies, some of these metal complexes could find potential applications
as various molecular functional materials, and as templates for the preparation of
nano-sized materials.
Centre for Carbon-Rich Molecular and Nano-Scale Metal-Based Materials Research, and Depart-
ment of Chemistry, The University of Hong Kong, Pokfulam Road, Hong Kong, P. R. China

C Springer 2008
218 V. W. Yam, K. M. Wong
19.2 Transition Metal Chalcogenides and Pnictogenide
19.2.1 Tetranuclear Copper(I) and Silver(I) Chalcogenides

and Pnictogenide
Despite a large number of transition metal chalcogenide cluster and organometallic

aggregates of d 10 metal centers containing short metal-metal interactions are
known (Grohmann et al., 1995), most of them are insoluble and the photophys-
ical and photochemical studies of the related species are relatively less explored.
We have successfully synthesized a series of tetranuclear copper(I) and silver(I)
complexes with μ4 -bridging chalcogenides, [M4 (P∧ P)4 (μ4 -E)]2+ {M = Cu, E =
S, Se; M = Ag, E = S, Se, Te; P∧ P = bis(diphenylphosphino)methane
(dppm), bis[bis(4-methylphenyl)-phosphino]methane (dtpm)}, from the reaction
of [M2 (P∧ P)2 (MeCN)2 ]2+ with Na2 E or Li2 E (Scheme 19.1) (Yam et al., 1993,
1996b,c, 2001b). All the complexes have been structurally characterized by X-ray
crystallography, with the four metal centers, in each case, arranged in a distorted
rectangular geometry bridged by a μ4 -chalcogenide ligand. Short Cu · · · Cu and
Ag · · · Ag contacts in the range of 2.869–3.271 and 3.038–3.357 Å, respectively,
are observed. A related isostuctural tetranuclear copper(I) pnictogenide com-
plex, [Cu4 (dppm)4 (μ4 -PPh)]2+ , has also been prepared by reacting PhPH2 with
[Cu2 (dppm)2(MeCN)2 ]2+ in the presence of Na(acac) (Deveson et al., 1997). These
complexes were found to exhibit intense long-lived luminescence at 527–718 nm
with excitation wavelength at ≥ 350 nm in the solid state. The transition associated
with this luminescence is assigned to originate from the ligand-to-metal charge
transfer (LMCT) [E2̃ /PPh2̃ → M4 ] triplet excited state, with some mixing of a
2+
P P 2+ E P
P
Na2E or Li2E M M
MeCN M M NCMe P
P
M M P
P P P
P P
2+
PhPH2
/ Na(acac)
P P P
M M
P P
M M
P P
P P
Scheme 19.1 Synthetic scheme for tetranuclear copper(I) and silver(I). chalcogenide and phos-
phinidene complexes
19 From Simple Molecules to Molecular Functional Materials and Nanoscience 219
E=S E=S
Normalized Emission Intensity

E = Se E = Se
E = PPh E = Te
500 600 700 800 400 500 600 700 800

Wavelength / nm Wavelength / nm
Fig. 19.1 Solid-state emission spectra of tetranuclear copper(I) and silver(I) chalcogenide and
phosphinidene complexes
metal-centered (MC) state. In general, the solid-state emission energies were found
to follow the order: E = S > Se > Te > PPh and M = Ag > Cu (Fig. 19.1). This
energy trend is in line with the ionization energies of the chalcogens or pnictogen
and supports an excited-state assignment of large LMCT [E2̃ /PPh2̃ → M4 ] charac-
ter. Depending on the metal centers and the chalcogens or pcictogen in these metal
clusters, different luminescence colors were observed (Table 19.1).
19.2.2 High Nuclearity Gold(I) Sulfido Clusters
Despite the fact that numerous copper(I) and silver(I) chalcogenide clusters
(Canales et al., 1996; Corrigan and Fenske, 1997; Yam and Lo, 1997) are
known, examples of chalcogenide clusters of gold(I), which belongs to the
same group as copper(I) and silver(I), are limited (Yam et al., 1999). A series
of high-nuclearity gold(I) sulfido complexes with bridging diphosphine ligands
have been synthesized and isolated by us, with the general formulae of [Au12
(μ-P∧ P)6 (μ3 -S)4 ]4+ , [Au10 (μ-PNP)4 (μ3 -S)4 ]2+ and [Au6 (μ-PNP)3 (μ3 -S)2 ]2+
[P∧ P = bis(diphenylphosphino)-methane (dppm); PNP = bis(diphenylphosphanyl)-
n-propylamine (Ph2 PN(n Pr)PPh2 )] (Yam et al., 1996a, 2000a, 2001a). Depend-
Table 19.1 Emission color of tetranuclear copper(I) and

silver(I) chalcogenide and pnictogenide complexes
M E Luminescence color
Cu S Orange-yellow
Cu Se orange
Cu PPh Red
Ag S Green
Ag Se greenish-yellow
Ag Te orange-yellow
Scheme 19.2 Synthetic scheme for high nuclearity gold(I) sulfido clusters
ing on the precursor complexes, [Au2 (μ-P∧ P)Cl2 ] or [Au2 (μ-PNP)Cl2 ], with
different bridging diphosphine ligands, dodecanuclear [Au12 (μ-P∧ P)6 (μ3 -S)4 ]4+ ,
could be obtained from the reaction of [Au2 (μ-P∧ P)Cl2 ] and H2 S while
decanuclear [Au10 (μ-PNP)4 (μ3 -S)4 ]2+ and hexanuclear [Au6 (μ-PNP)3 (μ3 -S)2 ]2+
were prepared and isolated from [Au2 (μ-PNP)Cl2 ] and H2 S (Scheme 19.2).
Intense orange luminescence was observed in [Au12 (μ-P∧ P)6 (μ3 -S)4 ]4+ and
[Au6 (μ-PNP)3 (μ3 -S)2 ]2+ while [Au10 (μ-PNP)4 (μ3 -S)4 ]2+ exhibited dual lumi-
nescence in green and orange-red upon excitation at λ ≥ 350 nm both in the
solid state and in solutions. The orange and orange-red luminescence have been
attributed to originate from the excited state derived from the ligand-to-metal
charge transfer character, mixed with metal-centered (ds/sp) states modified by
Au · · · Au interactions (LMMCT) [S → Au · · · Au], while the high-energy green
luminescence is derived from the metal-perturbed intraligand (IL) transition.
19.3 Luminescent Transition Metal Alkynyls
The chemistry of metal alkynyls has attracted enormous attention, in particular,

with the emerging interest in their potential applications in the field of materials
science. The linear geometry of the alkynyl unit and its π-unsaturated nature have
made metal alkynyls attractive building blocks, not only through σ-bonding but also
via π-bonding interactions, for applications as molecular wires and organometallic
oligomeric and polymeric materials which may possess unique properties. Despite
the growing interests and studies in metal alkynyls, relatively less attention was
focused on the luminescence behavior of this class of compounds. By incorporation
of linear alkynyl groups into luminescent transition metal complexes, the alkynyl
moiety with good σ-donor, π-donor and π-acceptor abilities is envisaged to play
an important role in the tuning and in the perturbation of the emission behaviors,
including emission energy, intensity and lifetime, either through its indirect involve-
ment as an auxiliary ligand or through its direct involvement in the emission origin.
19.3.1 Tetranuclear Copper(I) Alkynyl
A novel tetranuclear copper(I) ethynyl complex, [Cu4 (μ-dppm)4 (μ4 − η1 , η2 −

C C)]2+ , has been isolated from the reaction of [Cu2 (μ-dppm)2 (MeCN)2 ]2+
with trimethylsilylacetylene and n-butyllithium in THF (Scheme 19.3) (Yam et al.,
1993c). X-Ray crystallographic studies revealed that the complex exhibits a struc-
ture similar to that of the related tetranuclear copper(I) and silver(I) chalcogenides
and pnictogenide, in which the four copper(I) centers are arranged in a distorted
rectangular array with the four dppm ligands bridging each of the four Cu−Cu
edges in a saddle-like arrangement. The most remarkable feature is the presence
of an ethynyl ligand bridging the four copper(I) centers in both η1 and η2 bonding
modes.
The electronic absorption spectrum shows a low-energy absorption at ca. 374 nm,
assignable to a LMCT [(C C)2− → Cu4 ] transition, given the good σ-donating
ability of the (C C)2− unit. Upon excitation at λ ≥ 350 nm at room temperature,
intense green (509 nm) and greenish yellow (562 nm) emissions are observed, in
the solid state and in solution (φlum = 0.22 in acetone), respectively. In view of
the strong σ-donating properties of the ethynyl and the relatively long Cu · · · Cu
distances [3.245(2)–3.264(2) Å], the origin of the emission has been assigned to be
predominantly 3 LMCT [(C C)2− → Cu4 ] in nature.
19.3.2 Trinuclear copper(I) alkynyls
The triangulo-Cu3 complexes, [Cu3 (P∧ P)3 (μ3 − η1 − C CR)2 ]+ and [Cu3 (P∧ P)3
(μ3 − η1 − C CR)]2+ , have been prepared by the reaction of the binuclear
complex [Cu2 (P∧ P)2 (MeCN)2 ]2+ with RC CH in the presence of KOH or
Scheme 19.3 Synthetic scheme for a tetranuclear copper(I) ethynyl complex

Scheme 19.4 Molecular structures of [Cu3 (μ − P∧ P)3 {μ3 = η1 -(C ≡ C)n R}]2+ and
[Cu3 (μ − P∧ P)3 {μ3 = η1 -(C ≡ C)n R}2 ]+
n BuLi in the appropriate ratio under anaerobic and anhydrous conditions

(P∧ P = Ph2 PCH2 Ph2 (dppm); Ph2 PN(R )PPh2 , R = Ph, n Pr, C6 H4 -CH3 - p)
(Scheme 19.4) (Lo et al., 2004; Yam et al., 1993b, 1997). The interesting pho-
tophysical and photochemical properties of a number of luminescent polynu-
clear copper(I) alkynyl complexes have recently been studied. In all cases,
the excited states of the complexes have been proposed to bear a substantial
3 LMCT [(C C)n R → Cu] character, and probably mixed with a metal-centered
d-s triplet-state. The emission energies were found to be dependent on the R
substituent group with same alkynyl chain length and lower emission energies
were observed for the complexes with the stronger electron-donating substituents
on the alkynyl ligand. For instance, the emission energies follow the trend:
[Cu3 (μ-dppm)3 {μ3 − η1 -(C C)2 C6 H4 -OCH3 - p}2 ]+ < [Cu3 (μ-dppm)3 {μ3 −
η1 -(C C)2 C6 H4 -CH3 - p}2 ]+ < [Cu3 (μ-dppm)3 {μ3 − η1 -(C C)2 C6 H5 }2 ]+
(Fig. 19.2), which is in line with the assignment of 3 LMCT[(C C)2 R → Cu]
character. Moreover, their emission energies could also be tuned by the variation of
the alkynyl chain length. The greater is the number of alkynyl units that the complex
1811 cm–1
1915 cm–1
–1
R = Ph
2058 cm n=1
n=2
n=2
R = C6H4OCH3-p
R = C6H4CH3-p
R = Ph
500 550 600 650 700 750 400 500 600 700
Fig. 19.2 Normalized solid-state emission spectra of [Cu3 (μ-dppm)3 {μ3 = η1 -(C ≡ C)n R}2 ]+
possesses, the lower is the emission energy observed (Fig. 19.2). This energy trend
is also supportive of the assignment of a mixed metal cluster-centered d–s/LMCT
state for the emission origin since the π-electron donating ability decreases on going
from the diynyl moiety to the monoynyl counterparts.
19.3.3 Rhenium(I) Diimine Alkynyls
The photoluminescence properties of tricarbonylrhenium(I) diimine complexes

have also been extensively studied and a metal-to-ligand charge-transfer (MLCT)
[dπ → π∗ (diimine)] excited state were assigned for their emission origin (Wrighton
and Morse, 1974; Fredericks et al., 1979). In view of the well documented MLCT
excited state chemistry of rhenium(I) diimine complexes, a project of the synthesis
and photophysical studies of the related luminescent alkynylrhenium (I) diimine
complexes was launched.
Luminescent mono- and dinuclear rhenium(I) alkynyl complexes, [Re(CO)3
(N∧ N)−(C C)n R]] and [(N∧ N)(CO)3 Re−(C C − C C)n −Re(CO)3 (N∧ N)]
(n = 1; N−N = bpy, t Bu2 bpy, Me2 bpy, phen) were first synthesized in our
laboratory (Scheme 19.5) (Yam et al., 1995, 1996d, 1998, 2000b; Yam, 2001).
Intense orange phosphorescence was observed upon excitation at λ ≥ 350 nm in the
solid state and in fluid solution of these rhenium(I) alkynyls. Such luminescence is
ascribed to be originated from a 3 MLCT [dπ(Re) → π∗ (N∧ N)] excited state, with
some mixing of a [π(C C) → π∗ (N∧ N)] ligand-to-ligand charge transfer (LLCT)
character. The emission energies have been found to be dependent on the nature
of both the diimine and alkynyl ligands. Higher emission energy was observed
for the complexes containing better electron-donating substituents on the diimine
ligand for the complexes with the same alkynyl ligand (Yam, 2001), which is in
agreement with the 3 MLCT assignment. Upon extending the carbon chain in these
complexes, the emission energy was found to shift to the blue (Yam et al., 2000b)
and this is an unusual phenomenon in contrast to the common concept of tuning the
emission energy to the red by extending the number of acetylenic units in organic
polyynes.
Scheme 19.5 Molecular structures of luminescent mono- and dinuclear rhenium(I) alkynyl
complexes
19.3.4 Platinum(II) Terpyridyl Alkynyls
Platinum(II) polypyridyl complexes have been widely studied for their lumines-
cence behavior due to their intriguing photophysical and spectroscopic properties
resulting from their square planar coordination geometry, which permits easy access
to important frontier orbitals of the metal center, d 8 − d 8 metal–metal interactions
as well as π − π interactions (Miskowski and Houlding, 1989; Chan et al., 1994;
Marsh et al., 1995; McMillin and Moore, 2002). As a continuation of our great inter-
ests in luminescent transition metal alkynyls, a series of luminescent alkynylplat-
inum(II) terpyridyl complexes, [Pt(N∧ N∧ N)(C C−R)]+ (N∧ N∧ N =terpyridyl
ligand; R = alkyl, aryl or C CH) (Yam et al., 2001c, 2002b), have been synthe-
sized from the substitution reaction of the labile acetonitrile group on the precursor
complex [Pt(N∧ N∧ N)(MeCN)]2+ by the alkynyl group (Scheme 19.6).
The platinum(II) terpyridyl complexes with various substituted phenylethynyl
groups, [Pt(trpy)(C CC6 H4 -R- p)]+ (R = H, Cl, NO2 , CH3 , OCH3 , NH2 , NMe2 ,
N(C H2 C H2 OC H 3 )2 ) (Wong et al., 2004), in acetonitrile solution absorb strongly
at 286–350 nm and less intensely at 412–546 nm. The low-energy absorption
bands are assigned to a [dπ(Pt) → π∗ (N∧ N∧ N)] metal-to-ligand charge-transfer
(MLCT) transition, with mixing of some [π(C C) → π∗ (N∧ N∧ N)] ligand-to-
ligand charge-transfer (LLCT) contribution. In general, the stronger the electron-
donating ability of the alkynyl ligand is, the lower is the energy of the lowest
energy absorption. Therefore, such absorption energy follows the trend with
R = NO2 > Cl > H > CH3 > OCH3 > NH2 > N(CH3 )2 > N(CH2 CH2 OCH3 )2 ,
which is in line with the assignment of a transition of MLCT/LLCT admixture.
Unlike the chloro-counterpart, [Pt(trpy)Cl]+ , which is non-emissive in the solution
state at ambient temperature, most of the platinum(II) terpyridyl alkynyl complexes
were found to exhibit luminescence at 560–665 nm in acetonitrile solution at 298 K.
An origin of a [dπ(Pt) → π∗ (N∧ N∧ N)]3 MLCT excited state, with some mixing of
a [π(C CR) → π ∗ (N∧ N∧ N)]3 LLCT state, is assigned to this emission. Similar to
the electronic absorption studies, the emission energies of the complexes in solution
were found to depend on the nature of the substituents on the phenyl ring of the
alkynyl ligands, i.e. the stronger the electron-donating ability of the alkynyl ligand
is, the lower is the emission energy. However, for the complexes with stronger
electron-donating substituents, such as methoxy and amino groups, on the phenyl
ring of the alkynyl ligand, no emission was observed in acetonitrile solution at
298 K. The emission is believed to be quenched by the methoxy or amino group via
Scheme 19.6 Synthetic scheme for luminescent alkynylplatinum(II) terpyridyl complexes

Scheme 19.7 Molecular structures of platinum(II) terpyridyl alkynyl complexes with basic amino
functionalities and with crown ether pendants
photo-induced electron transfer or by the presence of the low-lying non-emissive

3 LLCT state.
In view of the sensitive dependence of the absorption and emission energy

on the R substituents of the alkynyl ligands, functionalization of the alkynyl
substituents could afford the generation of various molecular functional materials.
A series of platinum(II) terpyridyl alkynyl complexes that have been derivatized
with basic amino functionalities, [Pt(trpy)(C C-C6 H4 -NR2 - p)]X (X = OTf− ,
or Cl− ; R = H, CH3 , CH2 CH2 OCH3 ), and with benzo-15-crown-5 and N-
phenylaza-15-crown-5 pendants, [Pt(trpy)(C C-benzo-15-crown-5)]OTf and
[Pt(trpy)(C C-N-phenylaza-15-crown-5)]OTf, have been synthesized and char-
acterized (Scheme 19.7) (Wong et al., 2005a; Tang et al., 2005).
Their photophysical responses at various acid or metal cation concentrations
were studied. The complexes with basic amino functionalities exhibit dramatic
and reversible color change and luminescence switching upon addition of acid
(Fig. 19.3), as a result of the protonation of the amino group which decreases the
electron-donating ability of the alkynyl ligand and results in a blue shift of the
1.5
1.0
0.8
Absorbance
1.0
0.6
0.4
0.5
0.2
0.0
0.0
400 500 600 700 500 600 700 800
Fig. 19.3 UV-vis absorption (left) and emission (right) spectral changes of [Pt(trpy)
(C ≡ CC6 H 4 -N(CH2 CH2 OCH3 )2 - p)](OTf) in MeCN (0.05 M n Bu4 NPF6 ) with increasing
p-toluenesulfonic acid content
low-energy LLCT absorption and the corresponding shift of the 3 LLCT state to
higher energy. The reversibility of color changes was studied upon alternate addition
of p-toluenesulfonic acid and triethylamine. On the other hand, the complexes with
crown ether pendants exhibit electronic absorption spectral changes upon addition
of metal ions, and in some cases these are accompanied by visual color changes. The
observed blue shift in the low-energy MLCT absorption band upon ion-binding is
attributed to result from the complexation of metal ions to the crown ether moiety,
which would decrease the electron-donating ability of the alkynyl ligand, leading
to the lowering of the dπ(Pt) orbital energy. These findings suggest that with a
suitable design of the ligands and complexes and a sound understanding of the
spectroscopic properties of these complexes, promising potential colorimetric and
luminescence pH and metal cation sensors could be developed.
Two luminescent alkynylplatinum(II) terpyridyl complexes, [Pt(t Bu3 trpy)
(C C-C6 H4 -X − p)](OTf)(X = NCS and NHCOCH2 I), have also been designed
and synthesized (Scheme 19.8) (Wong et al., 2004). Their photophysical properties
have been studied and both exhibited strong luminescence in various media. Similar
to the other platinum(II) terpyridyl alkynyl complexes, the emission origin has
been tentatively assigned as a dπ(Pt) → π∗ (t Bu3 trpy)3MLCT excited state, with
some mixing of a π(C CR) → π∗ (t Bu3 trpy)3 LLCT character. Human serum
albumin (HSA) has been labeled by the ready reaction of the isothiocyanate and
iodoacetamide functional groups in these two complexes to afford the respective
bioconjugates, through the formation of thiourea and thioether linkage. Both
bioconjugates in 50 mM Tris-Cl buffer (pH 7.4) are highly colored and exhibit
luminescence in the visible region upon photoexcitation. The bioconjugate resulting
from the isothiocyanate complex is found to emit with a different luminescence
color when compared to its parent label (Fig. 19.4).
Interesting and rich polymorphic behavior was also observed in the platinum(II)
terpyridine butadiynyl complex, [Pt(trpy)(C CC CH)]OTf, which was found to
exist in two forms, i.e. a dark green form and a red form (Yam et al., 2002b).
Both of which have been structurally characterized and shown to exhibit different
crystal packing arrangements. The dark green form exists as a linear-chain with
Scheme 19.8 Molecular structures of platinum(II) terpyridyl alkynyl complexes for biolabeling
studies. Reprodued with permission from (32)
500 600 700 800

Wavelength / nm
Fig. 19.4 Emission spectra of [Pt(t Bu3 trpy)(C ≡ C-C6 H4 -NCS- p)](OTf) (—) in acetonitrile solu-
tion and its HSA bioconjugate (· · · · · · ) in 50 mM Tris-Cl pH 7.4 buffer solution
the platinum atoms equally spaced, with short intermolecular Pt · · · Pt contacts of

3.388 Å. The red form, on the other hand, shows dimeric structure with alternating
Pt · · · Pt distances of 3.394 and 3.648 Å. Dissolution of both forms in acetone or
acetonitrile gave the same absorption spectrum, indicating that the same molecu-
lar species exists in the solution state. There was a dramatic color change with a
tremendous emission enhancement in solution upon varying the composition of the
solvent. Upon increasing diethyl ether content in solution, the color of the solution
changed dramatically from yellow to green to blue, with a new absorption band
formed at 615 nm concomitant with a drop in absorbance at 416 nm (Fig. 19.5),
and a tremendous emission enhancement at 785 nm in the near-infrared region was
observed (Fig. 19.5). Both the emission enhancement and new absorption band for-
mation were ascribed to be derived from a metal-metal-to-ligand charge transfer
(MMLCT) origin of aggregated species; the formation of which was facilitated by
the propensity of this class of compounds to form Pt · · · Pt and π · · · π interactions.
6
Emisiion Intensity (A.U.)
Emission Intensity (A.U.)
0.4
Absorbance
2.0 5
0.3 5 4
0.2 3
Absorbance
1.5 0.1 4 2
1
0.0
40 50 60 70 80 90 3 0
1.0 40 50 60 70 80 90
Diethyl ether %
2 Diethyl ether %
0.5
1
0.0 0
300 400 500 600 700 500 600 700 800
Fig. 19.5 UV-vis absorption (left) and emission (right) changes of [Pt(tpy)(C ≡ CC ≡ CH)]OTf
in acetonitrile with increasing diethyl ether composition. Inset (left): Plot of absorbance vs. diethyl
ether composition in acetonitrile at 615 nm (). Inset (right): Plot of corrected emission intensity
as a function of diethyl ether composition ()
It is interesting to note that such drastic color changes and emission enhance-
ment arising from the solvent-induced aggregation are dependent on the nature of
counter anions (Yam et al., 1993a). Upon solvent-induced aggregation, the solution
colors change from yellow to blue in [Pt(trpy)(C CC CH)]OTf, to magenta
in [Pt(trpy)(C CC CH)]PF6 , to pink in [Pt(trpy)(C CC CH)]ClO4 , and
to orange in [Pt(trpy)(C CC CH)]BF4 . Similarly, emission enhancement was
observed at different wavelength upon solvent-induced aggregate formation which
demonstrates that counter anions play an important role in governing the degree of
aggregation and the extent of interactions within these aggregates.
19.3.5 Tetranuclear Gold(I) Alkynylcalix [4] Crown-6 Complexes
Gold(I) alkynyl systems without phosphine ligands are usually polymeric or

oligomeric in nature, resulting in intractable material, and molecular complexes of
this type are therefore extremely scarce. Alkynylcalixarenes may serve as versatile
ligands for the construction of novel luminescent gold(I) alkynyl supramolecular
assemblies by making use of their predefined conformations and preorganized
geometries. We have recently synthesized and isolated a series of novel luminescent
tetranuclear gold(I) alkynylcalix [4] crown-6 assemblies (Scheme 19.9) (Yip et al.,
2004). The X-ray crystal structure shows a double-cage structure, in which the two
diethynylcalixcrown ligands are σ-coordinated to two gold(I) centers on the two
ends, whereas the other two Au centers are each π-coordinated to the two alkynyl
units in a η2 , η2 -bonding fashion. All these complexes show intense luminescence
upon excitation at λ ≥ 370 nm. Their solid-state emission spectra show low-energy
emission bands at about 590̃620 nm at both 77 K and room temperature. Given
the presence of short Au · · · Au distances observed in the crystal structure, the
low-energy emission is assigned as derived from states of metal-cluster-centered
(ds/dp) character that are modified by Au · · · Au interactions with mixing of metal-
perturbed intraligand π̃π∗ (C C) states.
19.3.6 Biscyclometalated Gold(III) Alkynyls
Luminescent gold(III) compounds are rare, with very few exceptions that emit at
room temperature in solution (Chan et al., 1994; Yam et al., 1993a). One of the
Scheme 19.9 Molecular structures of luminescent gold(I) alkynyl supramolecular assemblies

Scheme 19.10 Molecular structures of luminescent alkynylgold(III) complexes. Reproduced with

permission from (37)
reasons for the lack of luminescence in gold(III) species is the presence of low-
energy d̃d ligand field (LF) states. As an extension of our great interest in transi-
tion metal alkynyl complexes, together with the potential improvement or enhance-
ment of the luminescence behaviors by introduction of strong σ-donating alkynyls
to gold(III) center, we recently synthesized a novel class of luminescent alkynyl-
gold(III) complexes, [Au(C∧ N∧ C)(C CR)] (HC∧ NCH = 2, 6-diphenylpyridine)
(Scheme 19.10) (Yam et al., 2005b). All complexes exhibit an intense vibronic-
structured absorption band at 362̃412 nm in dichloromethane. Since the absorption
energy was found to be rather insensitive to the nature of the alkynyl ligand, together
with the observation of vibrational progressional spacings of 1310̃1380 cm1̃ in the
low-energy absorption band corresponding to the skeletal vibrational frequency of
the C∧ N∧ C ligand, the low-energy absorption band is assigned as metal-perturbed
intraligand (IL) π̃π∗ transition of the C∧ N∧ C ligand, involving some charge transfer
from the phenyl moiety to the pyridyl unit. This assignment is further supported by
the observation of this low-energy absorption at lower energy in the complex with
two electron-donating tert-butyl groups on the phenyl rings of the C∧ N∧ C ligand
compared with the unsubstituted analogue (Fig. 19.6). Intriguingly, this class of
alkynylgold(III) complexes exhibit intense luminescence upon photoexcitation even
at ambient temperature. A vibronic-structured emission band with band maximum
at around 480 nm was generally observed in dichloromethane solution at room tem-
perature. The luminescence has been assigned as originating from a metal-perturbed
IL 3 [π−π] state of the tridentate C∧ N∧ C ligand. Similar to the electronic absorption
studies, the complex with twotert-butyl groups on the C∧ N∧ C ligand was found to
emit at lower energy than that without such substituents (Fig. 19.6). It is notewor-
thy that their solid-state emission spectra show a structureless band (550–585 nm)
at lower energy than that in solution, which is attributed to dimeric or excimeric
emission arising from the π − π stacking of the C∧ N∧ C ligand.
This class of luminescent cyclometalated gold(III) alkynyl complexes has
been demonstrated to possess electroluminescence (EL) properties and has been
employed as electrophosphorescent emitters or dopants of OLEDs with high bright-
ness and efficiency (Wong et al., 2005b). The color of the EL has been shown
to be capable of tuning by a variation in the applied DC voltage as well as the
dopant concentration (Fig. 19.7). In one of the devices, the emission color can be
tuned from orange to blue through white by applying different DC voltages. On
R′ = H, in CH2Cl2
4 H
Normalized Emisiion Intensity

R′ = tBu, in CH2Cl2
ε × 10–4 / dm3 mol–1 cm–1
t Bu R′ = H, Thin Film
0
300 350 400 450 500 500 600 700 800
Fig. 19.6 Electronic absorption (left) and emission (right) spectra of [Au(R C∧ N∧ CR )(CC-Ph)]
at 298 K
increasing concentration of Au(III) complex

Normalized EL Intensity
Normalized EL Intensity
6V
8V
10 V 1.00
12 V
14 V
16 V 0.75
18 V
20 V
0.50
0.25
0.00
400 500 600 700 800 400 500 600 700
Fig. 19.7 Electroluminescence spectra of OLED devices with Au(III) alkynyl complexes showing
the voltage- and dopant concentration-dependent EL properties
the other hand, the EL band maxima are found to red-shift from 500 to 580 nm
in another device upon increasing the dopant concentration. A maximum external
quantum efficiency of 5.5%, corresponding to a current efficiency of 17.6 cd A−1
and luminance power efficiency of 14.5 lm W−1 , has been obtained in one of the
multilayer OLEDs.
19.4 Template Synthesis of Mesoporous Silicates

and Nanoparticles
A series of surfactant ruthenium(II) complexes that possess lyotropic liquid crys-

talline properties have been synthesized (Scheme 19.11) for the design of new met-
allomesogens (Yam et al., 2002a). Through a variation of the alkyl chain length
and the head group of the metal complex template, metal-containing mesoporous
Scheme 19.11 Molecular structures of surfactant ruthenium(II) complexes
silicates with different pore size can be prepared. These materials may have potential
applications in catalysis and optical sensor functions.
Recently, we have synthesized monodisperse silver nanocubes with edge length
of 55 ± 5 nm, for the first time, on the basis of n-hexadecyltrimethylammonium bro-
mide (HTAB)-modified silver mirror reaction in water (Yu and Yam, 2004, 2005).
The addition of HTAB could control the consumption rate of Ag(I), which exists
in the form of AgBr. The HTAB is adsorbed on the cube surfaces to generate a
Fig. 19.8 TEM images of Ag NPs at different synthesis time. Reproduced with permission
from (41)
regular interparticle spacing, which avoids random aggregation. The size of such
nanocubes varies at different time intervals, as revealed by UV-vis absorption spec-
tra and TEM images (Fig. 19.8). Their high stability in water would be favorable for
further chemical modification and systematic investigations.
Acknowledgments Vivian Wing-Wah Yam acknowledges support from the university develop-
ment fund of the university of Hong Kong, and the university of Hong Kong foundation for educa-
tional development and research limited. This work has been supported by a central allocation vote
(hku 2/05c) grant from the research grants council of Hong Kong special administrative region,
people’s republic of china.
References
Canales, F., Gimeno, M.C., Laguna, A., Jones, P.G., 1996, Aurophilicity at Sulfur Cen-
ters. Synthesis and Reactivity of the Complex [S(Au2 dppf)]; Formation of Polynu-
clear Sulfur-Centered Complexes. Crystal Structures of [S(Au2 dppf)] ·2CHCl3 ,
[(μ-Au2 dppf){S(Au2 dppf)}2 ](OTf)2 · 8CHCl3 , and [S(AuPPh2 Me)2 (Au2 dppf)](ClO4 )2
3CH2 Cl2 . Journal of American Chemical Society, 118, pp. 4839–4845.
Chan, C.W., Cheng L.K., Che, C.M., 1994a, Luminescent Donor-Acceptor Platinum(II) Com-
plexes. Coordination Chemistry Reviews, 132, pp. 87–97.
Chan, C.W., Wong, W.T., Che, C.M., 1994b, Gold(III) Photooxidants. Photophysical, Photochem-
ical Properties, and Crystal Structure of a Luminescent Cyclometalated Gold(III) Complex of
2,9-Diphenyl-1,10-Phenanthroline. Inorganic Chemistry, 33, pp. 1266–1272.
Corrigan, J.F., Fenske, D, 1997, New Copper Telluride Clusters by Light-Induced Tellurolate-
Telluride Conversions. Angewandte Chemie, International Edition, pp. 1981–1983.
Deveson, A., Dehnen, S., Fenske, D., 1997, Syntheses and Structures of Four New Copper(I)–
Selenium Clusters: Size Dependence of the Cluster on the Reaction Conditions. Journal of
Chemical Society, Dalton Transaction, pp. 4491–4498.
Fredericks, S.M., Luong, J.C., Wrighton, M.S., 1979, Multiple Emissions from Rhenium(I) Com-
plexes: Intraligand and Charge-Transfer Emission from Substituted Metal Carbonyl Cations.
Journal of American Chemical Society, 101, pp. 7415–1717.
Grohmann, A., Schmidbaur, H., in: Abel, E.W., Stone, F.G.A., Wilkinson, G. (Eds.), 1995, Com-
prehensive Organometallic Chemistry II, vol. 3 (Pergamon, Oxford).
Lo, W.Y., Lam, C.H., Yam, V.W.W., Zhu, N., Cheung, K.K., Fathallah, S., Messaoudi, S.,
Le Guennic, B., Kahlal, S., Halet, J.F., 2004, Synthesis, Photophysics, Electrochemistry,
Theoretical, and Transient Absorption Studies of Luminescent Copper(I) and Silver(I)
Diynyl Complexes. X-Ray Crystal Structures of [Cu3 (μ-dppm)3 (μ3 -η1 -C CC CPh)2 ]PF6
and [Cu3 (μ-dppm)3 (μ3 -η1 -C CC CH)2 ]PF6 . Journal of American Chemical Society, 126,
pp. 7300–7310.
Marsh, R.E., Miskowski, V.M., Schaefer, W.P., Gray, H.B., 1995, Electronic Spectroscopy of
Chloro(terpyridine)platinum(II). Inorganic Chemistry, 34, pp. 4591–4599.
McMillin D.R., Moore, J.J., 2002, Luminescence that Lasts from Pt(trpy)Cl+ derivatives (trpy =
2, 2 ; 6 , 2 -terpyridine). Coordination Chemistry Reviews, 229, pp. 113–121.
Miskowski V.M., Houlding, V.H., 1989, Electronic Spectra and Photophysics of Platinum(I1)
Complexes with α-Diimine Ligands. Solid-state Effects. 1. Monomers and Ligand π Dimers.
Inorganic Chemistry, 28, pp. 1529–1533.
Tang, W.S., Lu, X.X., Wong, K.M.C., Yam, V.W.W., 2005, Synthesis, Photophysics and Binding
Studies of Platinum(II) Alkynyl Terpyridine Complexes with Crown Ether Pendant. Potential
Luminescent Sensors for Metal Ions. Journal of Materials Chemistry, 15, pp. 2714–2720.
Wong, K.M.C., Tang, W.S., Chu, B.W.K., Zhu, N., Yam, V.W.W., 2004, Synthesis, Photophysical
Properties and Biomolecular Labeling Studies of Luminescent Platinum(II)-Terpyridyl Alkynyl
Complexes. Organometallics, 23, pp. 3459–3465.
Wong, K.M.C., Tang, W.S., Lu, X.X., Zhu, N., Yam, V.W.W., 2005a, Functionalized Platinum(II)
Terpyridyl Alkynyl Complexes as Colorimetric and Luminescence pH Sensors. Inorganic
Chemistry, 44, pp.1492–1498.
Wong, K.M.C., Zhu, X., Hung, L.L., Zhu, N., Yam, V.W.W., Kwok, H.S., 2005b, A Novel Class of
Phosphorescent Gold(III) Alkynyl-Based Organic Light Emitting Devices with Tunable Color.
Chemical Communications, pp. 2906–2908.
Wrighton, M.S., Morse, D.L., 1974, Nature of the lowest excited state in tricarbonylchloro-1,10-
phenanthrolinerhenium(I) and related complexes. Journal of American Chemical Society, 96,
pp. 998–1003.
Yam, V.W.W., 2001, Luminescent Carbon-Rich Rhenium(I) Complexes. Chemical Communica-
tions, pp. 789–796.
Yam, V.W.W., Chan, K.H.Y., Wong, K.M.C., Zhu, N., 2005a, Luminescent Platinum(II) Terpyridyl
Complexes: Effect of Counter Ions on Solvent-Induced Aggregation and Color Changes.
Chemistry - A European Journal, 11, pp. 4535–4543.
Yam, V.W.W., Wong, K.M.C., Hung, L.L., Zhu, N., 2005b, Luminescent Gold(III) Alkynyl
Complexes - Synthesis, Structural Characterization and Luminescent Properties. Angewandte
Chemie International Edition, 44, pp. 3107–3110.
Yam, V.W.W., Cheng, E.C.C., Zhou, Z.Y., 2000a, A Highly Soluble Luminescent Decanuclear
Gold(I) Complexes with a Propeller-Shaped Structure. Angewandte Chemie, International Edi-
tion, 39, pp. 1683–1685; Angewandte Chemie, 2000, pp. 1749–1751.
Yam, V.W.W., Cheng, E.C.C., Cheung, K.K., 1999, A Novel High Nuclearity Luminescent Gold(I)
Sulfido Complex. Angewandte Chemie, International Edition in English, 38, pp. 197–199;
Angewandte Chemie, 111, pp. 193–195.
Yam, V.W.W., Chong, S.H.F., Ko, C.C., Cheung, K.K., 2000b, Synthesis and Lumines-
cence Behaviour of Rhenium(I) Triynyl Complexes. X-Ray Crystal Structures of [Re(CO)3
(t Bu 2 bpy)(C C-C C-C CPh)] and [Re(CO)3 (Me2 bpy)(C C-C C-C CSiMe3 )].
Organometallics, 19, pp. 5092–5097.
Yam, V.W.W., Cheng, E.C.C., Zhu, N., 2001a, A Novel Polynuclear Gold-Sulfur Cube with an
Unusually Large Stoke’s Shift. Angewandte Chemie, International Edition, 40, pp. 1763–1765.
Yam, V.W.W., Cheng, E.C.C., Zhu, N., 2001b, The First Luminescent Tetranuclear Copper(I)
μ4 -Phosphinidene Complex. Chemical Communications, pp. 1028–1029.
Yam, V.W.W., Tang, R.P.L., Wong, K.M.C., Cheung, K.K., 2001c, Synthesis, Luminescence and
Ion-Binding Studies of Platinum(II) Terpyridyl Acetylide Complexes. Organometallics, 20,
pp. 4476–4482.
Yam, V.W.W., Choi, S.W.K., Lai, T.F., Lee, W.K., 1993a, Syntheses, Crystal Structures
and Photophysics of Organogold(III) Diimine Complexes. X-Ray Crystal Structures of
[AuIII (bpy)(mes)2 ]ClO4 and [AuIII (phen)(CH2 SiMe3 )2 ]ClO4 . Journal of Chemical Society,
Dalton Transactions, pp. 1001–1002.
Yam, V.W.W., Lee, W.K., Lai, T.F., 1993b, Synthesis and Luminescent Properties of a Novel
Tetranuclear Copper(I) Cluster containing a μ4 -Sulphur Moiety. X-Ray Crystal Structure of
[Cu4 (μ-dppm)4 (μ4 -S)](PF6 )2 • 2Me2 CO [dppm = bis(diphenylphosphino) methane]. Journal
of Chemical Society, Chemical Communications, pp. 1571–1573.
Yam, V.W.W., Lee, W.K., Lai, T.F., 1993c, Synthesis, Spectroscopy and Elec-
trochemistry of Trinuclear Copper(I) Acetylides. X-Ray Crystal Structure
of [Cu3 (μ-Ph2 PCH2 PPh2 )3 (μ3 -η1 -C Ct Bu)(μ3 -Cl)]PF6 . Organometallics, 12,
pp. 2383–2387.
Yam, V.W.W., Chong, S.H.F., Cheung, K.K., 1998, Synthesis and Luminescence
Behaviour of Rhenium(I) Diynyl Complexes Complexes. X-Ray Crystal Structures of
[Re(CO)3 (t Bu2 bpy)(C C-C CH)] and [Re(CO)3 (t Bu2 bpy)(C C-C CPh)]. Chemical
Communications, pp. 2121–2122.
Yam, V.W.W., Fung, W.K.M., Cheung, K.K., 1996a, Synthesis, Structure, Photo-physics, and
Excited-State Redox Properties of the Novel Luminescent Tetranuclear Acetylidocopper(I)
Complex [Cu4 (μ-dppm)4 (μ4 -η1 , η2 -C C-)][BF4 ]2 . Angewandte Chemie, International Edi-
tion in English, 35, pp. 1100–1102; Angewandte Chemie, 108, pp. 1213–1215.
Yam, V.W.W., Lo, K.K.W., Cheung, K.K., 1996b, A Novel Luminescent μ4 -Selenido-Bridged
Copper(I) Tetramer. Inorganic Chemistry, 35, pp. 3459–3462.
Yam, V.W.W., Lo, K.K.W., Wang, C.R., Cheung, K.K., 1996c, The First Series of Luminescent
(μ4 -Chalcogenido)silver(I) Clusters. Inorganic Chemistry, 35, pp. 5116–5117.
Yam, V.W.W., Lau, V.C.Y., Cheung, K.K., 1996d, Luminescent Rhenium(I) Carbon
Wires: Synthesis and Photophysics. X-Ray Crystal Structure of [Re(t Bu2 bpy)(CO)3
(C C-C C)Re(t Bu2 bpy)(CO)3 ]. Organometallics, 15, pp. 1740–1744.
Yam, V.W.W., Fung, W.K.M., Wong, M.T., 1997, Synthesis, Photophysics, Electrochem-
istry and Excited-State Redox Properties of Trinuclear Copper(I) Acetylides with
Bis(diphenylphosphino)-Alkyl and Aryl Amines as Bridging Ligands. Organometallics, 16,
pp. 1772–1778.
Yam, V.W.W., Lau, V.C.Y., Cheung, K.K., 1995, Synthesis and Photophysics of Luminescent
Rhenium(I) Acetylides - Precursors for Organometallic Rigid-Rod Materials. X-Ray Crystal
Structures of [Re(t Bu2 bpy)(CO)3 (t BuC C)] and [Re(t Bu2 bpy)(CO)3 Cl]. Organometallics,
14, pp. 2749–2753.
Yam, V.W.W., Li, B., Zhu, N., 2002a, Synthesis of Mesoporous Silicates with Controllable Pore
Size Using Surfactant Ruthenium(II) Complexes that Display Lyotropic Mesomorphism as
Templates. Advanced Materials, 14, pp. 719–722.
Yam, V.W.W., Wong, K.M.C., Zhu, N, 2002b, Solvent-Induced Aggregation through Metal· · ·
Metal/π · · · π Interactions – Large Solvatochromism of Luminescent Organoplatinum(II) Ter-
pyridyl Complexes. Journal of American Chemical Society, 124, pp. 6506–6507.
Yam, V.W.W., Lo, K.K.W. 1997, Luminescent Tetranuclear Copper(I) and Silver(I) Chalcogenides.
Comments on Inorganic Chemistry, 19, pp. 209–229.
Yip, S.K., Cheng, E.C.C., Yuan, L.H., Zhu, N., Yam, V.W.W., 2004, Supramolecular Assem-
bly of Luminescent Gold(I) Alkynylcalix[4]crown-6 Complexes Having Unprecedented Pla-
nar η2 , η2 -Coordination Gold(I) Centers. Angewandte Chemie International Edition, 43,
pp. 4954–4957.
Yu, D.B., Yam, V.W.W., 2004, Controlled Synthesis of Monodisperse Silver Nanocubes in Water.
Journal of American Chemical Society, 126, pp. 13200–13201.
Yu, D.B., Yam, V.W.W., 2005, Hydrothermal-Induced Assembly of Colloidal Silver Spheres into
Various Nanoparticles on the Basis of HTAB-Modified Silver Mirror Reaction. Journal of
Physical Chemistry B, 109, pp. 5497–5503.
Chapter 20
First-Principles Method for Open
Electronic Systems
Xiao Zheng and GuanHua Chen∗
We prove the existence of the exact density-functional theory formalism for open
electronic systems, and develop subsequently an exact time-dependent density-
functional theory (TDDFT) formulation for the dynamic response. The TDDFT
formulation depends in principle only on the electron density of the reduced sys-
tem. Based on the nonequilibrium Green’s function technique, it is expressed
in the form of the equation of motion for the reduced single-electron density
matrix, and this provides thus an efficient numerical approach to calculate the
dynamic properties of open electronic systems. In the steady-state limit, the conven-
tional first-principles nonequilibrium Green’s function formulation for the current is
recovered.
Density-functional theory (DFT) has been widely used as a research tool in
condensed matter physics, chemistry, materials science, and nanoscience. The
Hohenberg-Kohn theorem (1964) lays the foundation of DFT. The Kohn-Sham
formalism (1965) provides the practical solution to calculate the ground state prop-
erties of electronic systems. Runge and Gross extended further DFT to calculate the
time-dependent properties and hence the excited state properties of any electronic
systems (Runge and Gross, 1984). The accuracy of DFT or TDDFT is determined
by the exchange-correlation functional. If the exact exchange-correlation functional
were known, the Kohn-Sham formalism would have provided the exact ground state
properties, and the Runge-Gross extension, TDDFT, would have yielded the exact
properties of excited states. Despite of their wide range of applications, DFT and
TDDFT have been mostly limited to closed systems.
Fundamental progress has been made in the field of molecular electronics
recently. DFT-based simulations on quantum transport through individual molecules
attached to electrodes offer guidance for the design of practical devices (Lang
and Avouris, 2000; Heurich et al., 2002; Wang and Luo, 2003). These simula-
tions focus on the steady-state currents under the bias voltages. Two types of
approaches have been adopted. One is the Lippmann-Schwinger formalism by
Lang and coworkers (Lang, 1995). The other is the first-principles nonequilibrium
Green’s function technique (Taylor et al., 2001; Ke et al., 2004; Deng et al., 2004;
Brandbyge et al., 2002; Xue et al., 2001). In both approaches the Kohn-Sham
∗ Department of Chemistry, The University of Hong Kong, Pokfulam Road, Hong Kong, China

C Springer 2008
236 X. Zheng, G. H. Chen
Fock operator is taken as the effective single-electron model Hamiltonian, and the
transmission coefficients are calculated within the noninteracting electron model.
It is thus not clear whether the two approaches are rigorous. Recently Stefanucci
and Almbladh (2004) derived an exact expression for time-dependent current in
the framework of TDDFT. In the steady-current limit, their expression leads to
the conventional first-principles nonequilibrium Green’s function formalism if the
TDDFT exchange-correlation functional is adopted. However, they did not provide
a feasible numerical formulation for simulating the transient response of molecular
electronic devices. In this communication, we present a rigorous first-principles
formulation to calculate the dynamic properties of open electronic systems. We
prove first a theorem that the electron density distribution of the reduced system
determines all physical properties or processes of the entire system. The theorem
lays down the foundation of the first-principles method for open systems. We
present then the equation of motion (EOM) for nonequilibrium Green’s functions
(NEGF) in the framework of TDDFT. By introducing a new functional for the
interaction between the reduced system and the environment, we develop further
a reduced-single-electron-density-matrix-based TDDFT formulation. Finally, we
derive an exact expression for the current which leads to the existing DFT-NEGF
formula in the steady-state limit. This shows that the conventional DFT-NEGF
formalism can be exact so long as the correct exchange-correlation functional
is adopted. Both Hohenberg-Kohn theorem and Runge-Gross extension apply to
isolated systems. Applying Hohenberg-Kohn-Sham’s DFT and Runge-Gross’s
TDDFT to open systems requires in principle the knowledge of the electron density
distribution of the total system which consists of the reduced system and the
environment. This presents a major obstacle in simulating the dynamic processes
of open systems. Our objective is to develop an exact DFT formulation for open
systems. In fact, we are interested only in the physical properties and processes
of the reduced system. The environment provides the boundary conditions and
serves as the current source and energy sink. We thus concentrate on the reduced
system.
Any electron density distribution function ρ(r) of a real physical system is a real
analytic function. We may treat nuclei as point charges, and this would only lead
to non-analytic electron density at isolated points. In practical quantum mechanical
simulations, analytic functions such as Gaussian functions and plane wave functions
are adopted as basis sets, which results in analytic electron density distribution.
Therefore, we conclude that any electron density functions of real systems are real
analytic on connected physical space. Note that the isolated points at nuclei can
be excluded for the moment from the physical space that we consider, so long the
space is connected. Later we will come back to these isolated points, and show
that their inclusion does not alter our conclusion. Based on this, we show below
that for a real physical system the electron density distribution function on a sub-
space determines uniquely its values on the entire physical space. This is nothing but
the analytic continuation of a real analytic function. The proof for the univariable
real analytical functions can be found in textbooks, for instance, reference (Krantz
and Parks, 2002). The extension to the multivariable real analytical functions is
straightforward.
20 First-Principles Method for Open Electronic Systems 237
Lemma 1. The electron density distribution function ρ(r) is real analytic on a con-
nected physical space U . W ⊆U is a sub-space. If ρ(r) is known for all ρ(r) ∈
W, ρ(r) can be uniquely determined on entire U .
Proof. To facilitate our discussion, the following notations are introduced. Set Z+ =
{0, 1, 2, . . .}, and γ is an element of (Z+ )3 , i.e., γ = (γ1 , γ2 , γ3 ) ∈ (Z+ )3 . The dis-
placement vector r is denoted by the three-dimensional variable x = (x 1 , x 2 , x 3 ) ∈
γ γ γ γ γ γ γ
U . Denote that γ ! = γ1 !γ2 !γ3 !, x γ = x 11 x 22 x 33 , and ∂∂x γ = ∂ γ11 ∂ γ22 ∂ γ33 .
∂ x1 ∂ x2 ∂ x3
Suppose that another density distribution function ρ (x) is real analytic on U
γ γ
and equal to ρ(x) for all x ∈ W . We have ∂ ∂ρ(x) xγ = ∂ ∂ρx(x)
γ for all x ∈ W and
γ ∈ (Z+ )3 . Taking a point x 0 at or infinitely close to the boundary of W , we may
1 ∂ γ ρ(x)
expand ρ(x) and ρ(x ) around x 0 , i.e., ρ(x) = γ∈(Z+ )3 γ! ∂ x γ (x − x 0 )γ and

x0
1 ∂ γ ρ (x) γ
ρ (x) = γ∈(Z+ )3 γ! ∂ x γ (x−x 0) . Assuming that the convergence radii for the
x0
Taylor expansions of ρ(x) and ρ (x) at x 0 are both larger than a positive finite real

number b, we have ρ(x) = ρ (x) for all x ∈ Db (x 0 ) = {x : |x − x 0 | < b} since
thus
∂ γ ρ(x) ∂ γ ρ (x)
∂xγ = ∂ x γ . Therefore, the equality ρ (x) = ρ(x) has been expanded
x0 x0
beyond W to include Db (x 0 ). Since U is connected the above procedure can be
repeated until ρ (x) = ρ(x) for all x ∈ U .
We have thus proven that ρ can be uniquely determined on U once it is known
on W , and are ready to prove the following theorem.
Theorem 1. Electron density function ρ(r) for a subsystem of a connected real
physical system determines uniquely all electronic properties of the entire system.
Proof. Assuming the physical space spanned by the subsystem and the connected
real physical system are W and U , respectively. W is thus a sub-space of U ,
i.e., W ⊆ U . According to the above lemma, ρ(r) on W determines uniquely their
value on U , i.e., ρ(r) of the subsystem determines ρ(r) of the entire system.
Inclusion of isolated points, lines or planes where ρ(r) is non-analytic into the
connected physical space does not violate the theorem; so long ρ(r) is continuous
on these points, lines or planes. This can be shown clearly by performing analyti-
cal continuation of ρ(r) infinitesimally close to them. Therefore, we conclude that
ρ(r) of any finite subsystem determines uniquely ρ(r) of the entire physical system
including the nuclear sites. Hohenberg-Kohn theorem and Runge-Gross extension
state that the electron density distribution of a system determines uniquely all its
electronic properties. Therefore, we conclude that ρ(r) for a subsystem determines
all the electronic properties of the real physical system.
The above theorem guarantees the existence of an exact DFT-type method for
open systems. In principle, all we need to know is the electron density of the reduced
system. The electron density distribution in the environment can be obtained by the
analytic continuation of the electron density function at or near the boundary. In
practice the reduced system must be chosen appropriately so that this analytic con-
tinuation procedure can be conveniently performed. For instance, Fig. 20.1 depicts
one type of open systems, a molecular device. It consists of the reduced system or
Fig. 20.1 Schematic representation of the experimental setup for quantum transport through a
molecular device
device region D and the environment, the left and right leads L and R. Note that the
reduced system D contains not only the molecular device itself, but also portions
of the left and right electrodes, in this way the extension of the density function
can be facilitated by the knowledge on the electron density distribution of the bulk
electrodes.
The challenge is to develop a practical first-principles method. Taking the open
system displayed in Fig. 20.1 as an example, we develop an exact DFT formalism
for the open systems. To calculate the properties of a molecular device, we need only
the electron density distribution in the device region. The influence of the electrodes
can be determined by the electron density distribution in the device region. Within
the TDDFT formalism, a closed nonlinear self-consistent EOM has been derived
for the reduced single-electron density matrix σ (t) of the entire system (Yam et al.,
2003):
i σ̇ (t) = [h(t), σ (t)], (20.1)
where h(t) is the Kohn-Sham Fock matrix, and the square bracket on the right-hand
side (RHS) denotes the commutator. The matrix element of σ is defined as
σi j (t) = a †j (t)ai (t), (20.2)
where ai (t) and a +j (t) are the Heisenberg annihilation and creation operators for
atomic orbitals i and j at time t, respectively. Expanded in a real space basis set, the
matrix representation of σ can be partitioned as
⎡ ⎤
σL σL D σL R
σ = ⎣σ D L σ D σ D R ⎦ , (20.3)
σR L σR D σR
where σ L , σ R , and σ D represent the diagonal blocks correspond to the left lead, the
right lead and the device region, respectively, and σ L D , σ R D , σ L R , σ D L , σ D R and
σ R L are the electron coherence between two different subsystems. The only block
of finite dimensions is σD , The Kohn-Sham Fock matrix h can be partitioned in the
same way with σ replaced by h in Eq. (20.3). Thus the EOM for σD which is of our
central interest can be written down readily as

i σ̇ D = [h D , σ D ] + (h Dα σα D − σ Dα h α D ) . (20.4)
α=L .R
The reduced system D and the leads L/R are spanned by the atomic orbitals
{l} and the single-electron states {kα }, respectively. And hence Eq. (20.4) is equiva-
lent to

i σṅm = (h nl σlm − σnl h lm ) − i Q α,nm . (20.5)
l∈D α=L ,R
Here the Q α,nm on the RHS is the dissipative term due to the lead α (L or R)
whose expanded form is

Q α,nm = i h n,kα σ kα , m − σm,kα h kα ,m , (20.6)

kα
where h n,kα is the coupling matrix element between the atomic orbital n and the
single-electron state kα . σkα,m is precisely the lesser Green’s function of identical
time variables, i.e.,

σkα,m (t) = −i G <

kα,m t.t |t =t .
Based on the Keldysh formalism (1965), the analytic continuation rules of Lan-
greth and Nordlander (1991) and the NEGF formulation developed by Jauho et al.
(1994), Q α,nm (t) can be calculated by
∞ α
Q α,nm (t) = − dτ G <
nl (t, τ ) (τ, t) +
α,lm
l∈D −∞
< <
α
G rnl (t, τ ) (τ, t) − (t, τ )G lm (τ, t)
α,lm α,nl
n
<
− (t, τ )G lm (τ, t) . (20.7)
α,nl
where G r , G a and G < are the retarded, advanced and lesser Green’s function for the
reduced system D, and Σ r , Σ a , and Σ < are the retarded, advanced and lesser self-
energies, respectively. The current through the interfaces SL or SR (see Fig. 20.1)
can be expressed as
d
∂
Jα (t) = − ρ (r, t) = −
dr σk ,k (t)
α ∂t dt α α
k∈α

=i h kα,l σl,kα − σkα ,l h l,kα

l∈D k∈α

=− Q α,ll = −tr [Q α (t)] . (20.8)
l∈D
i.e., the current through Sα (α=L or R) is the negative of the trace of Qα . Combining
Eqs. (20.7) and (20.8), we recover the current expression (Eq. 20.8) of (Stefanucci
and Almbladh, 2004).
At first glance Eq. (20.5) is not self-closed since the dissipative terms Q α. remain
unsolved. According to the theorem we proved earlier, all physical quantities are
explicit or implicit functionals of the electron density in D, ρ D (t). Q α are thus also
universal functionals of ρ D (t). Therefore, Eq. (20.5) can be recast into a formally
closed form,

i σ̇ D = [h D [ρ D (t)] , σ D ] − i Q α [ρ D (t)] . (20.9)
α=L ,R
Neglecting the second term on the RHS of Eq. (20.9) leads to the conventional
TDDFT formulation in terms of reduced single-electron density matrix (Yam et al.,
2003). The second term describes the dissipative processes where electrons enter
and leave the region D. Besides the exchange-correlation functional, the additional
universal density functional Q α [ρ D (t)] is introduced to account for the dissipa-
tive interaction between the reduced system and its environment. Equation (20.9)
is thus the TDDFT formulation in terms of the reduced single-electron matrix for
the open system. In the frozen DFT approach (Wesolowski and Warshel, 1993)
an additional exchange-correlation functional term was introduced to account for
the exchange-correlation interaction between the system and the environment. This
additional term is included in h D (t) of Eq. (20.5). Yokojima et al. (2003) developed
a dynamic mean-field theory for dissipative interacting many-electron systems. An
EOM for the reduced single-electron density matrix was derived to simulate the
excitation and nonradiative relaxation of a molecule embedded in a thermal bath.
This is in analogy to our case although our environment is actually a fermion bath
instead of a boson bath. The major difference is that the number of electrons in the
reduced system is conserved in reference (Yokojima et al., 2003) while in our case it
is not. Admittedly, Q α [ρ D (t)] can be an extremely complex functional. Progressive
approximations are needed for the practical solution of Eq. (20.9).
To obtain the steady-state solution of Eqs. (20.5) or (20.9), we adopt a
similar strategy as that of reference (Stefanucci and Almbladh, 2004). As
kα
τ → ∞, Γnm (t, τ ) ≡ h nkα (t)h kα m (τ ) becomes asymptotically time-independent,
and Gs and
s rely simply on the difference of the two time-variables (Stefanucci
and Almbladh, 2004). The steady-state current can thus be expressed as

JL (∞) = − J R (∞) = − Q L ,nn (∞)
n∈D

= f () − f () T () d,
L R
(20.10)
T () = 2πη L () η R () ×

tr G rD () Γ R () G αD () Γ L () (20.11)
Here T (ε) is thetransmission coefficient, f α (ε) is the Fermi distribution func-

tion, and ηα (ε) = k∈α δ(ε − εkα ) is the density of states for the lead α (L or R).
Equation (20.10) is exactly the Landauer formula (Datta, 1995; Landauer, 1970) in
the DFT-NEGF formalism (Taylor et al., 2001; Ke et al., 2004). The only difference
is that Eq. (20.10) is derived within the TDDFT formalism in our case while it is
evaluated within the DFT framework in the case of the DFT-NEGF formulation
(Taylor et al., 2001; Ke et al., 2004). In other words, the DFT-NEGF formalism
can be exact so long as the correct exchange-correlation functional is used! This
is not surprising, and is simply a consequence of that (i ) DFT and TDDFT can
yield the exact electron density and (ii) the current is the time derivative of the total
charge.
Just as the exchange-correlation functional, the exact functional form of Q α on
density is rather difficult to derive. Various approximated expressions have been
adopted for the DFT exchange-correlation functional in practical implementations.
Similar strategy can be employed for Q α One such scheme is the adiabatic approx-
imation, which implies that the dissipation functional Q α depends instantaneously
on the electron density of the reduced system D, i.e.,

Q α [ρ D (t)] ≈ Q αAD ρ tD , (20.12)
where Q α is a functional of the density function ρD (t) over both time and space and
Q αAD is a functional of ρD (t) over space only. For those ρD (t) corresponding to the
steady-state solutions of Eq. (20.9), the exact functional values of Q αAD [ρ D (t)] can
be acquired by the TDDFT-NEGF formalism with the correct exchange-correlation
functional being adopted, i.e.,

Q αAD ρ tD |ρ t SS = Q αT D D F T −N E G F [ρ tD ], (20.13)
D∈ ρD
where {ρ D SS
} denotes the set of all accessible steady-state density functions on the
reduced system D. The explicit functional form of Q αAD can thus be obtained by fit-
ting to the data available from Eq. (20.13) and then smoothly extrapolating to ρD (t)
not in {ρ D
SS
}. In this way Q α becomes an explicit functional of ρD and the Eq. (20.9)
spontaneously leads to the same correct steady-state solution as that offered by the
TDDFT-NEGF formalism.
Another scheme is the wide-band limit (WBL) approximation (Jauho et al.,
1994), which consists of a series of approximations imposed on the leads: (i ) their
band-widths are assumed to be infinitely large, (ii) their linewidths Λαk (t, τ ) defined
by πηα (εkα )Γ kα (t, τ ) are regarded as energy independent, i.e.,
Λαk (t, τ ) ≈ Λα (t, τ ) ≈ Λα ,
and (iii) the energy shifts are taken as level independent, i.e., δεkα (t) ≈ δεα (t) ≈ δεα
for L or R. The physical essence of the transport problem is captured under these
reasonable hypotheses [18]. In the practical implementation, the effects of the spe-
cific electronic structures of the leads can be captured by enlarging the device region
to include enough portions of the electrodes. Following the procedures similar to

those in reference (Jauho et al., 1994), we obtain that
†
Q α = P α (t) + P α (t) + Λα · σ D , (20.14)
where the curly bracket on the RHS denotes the anticommutator. Taking t = 0 as
the switch-on instant, P α (t) can be written down as
α
α 2i (−) t eiδ (t −τ ) (+)
P (t) = U (t) dτ U (τ )+
π 0 t −τ
0 α

i eiδ τ τ,0
dτ G (−τ ) Λα − 2Λα . (20.15)
−∞ t −τ D
where G rD is the retarded Green’s function of D before the switch-on instant. The
propagators for the reduced system D are defined as
⎧ ⎫
⎨ t ⎬
U (±) (t) = exp ±i h D (τ ) dτ ± Λα t (20.16)
⎩ 0 ⎭
α=L ,R
Equations (20.14) ∼ (20.16) constitute the WBL formulation of the TDDFT-

NEGF formalism where Q α exhibit an implicit dependence on ρD (t).
To summarize, we have proven the existence of the exact TDDFT formalism
for the open electronic systems, and have proposed a TDDFT-NEGF formulation
to calculate the quantum transport properties of molecular devices. Since TDDFT
results in formally exact density distribution, the TDDFT-NEGF formulation is in
principle an exact theory to evaluate the transient and steady-state currents. In partic-
ular, the TDDFT-NEGF expression for the steady-state current has the exact same
form as that of the conventional DFT-NEGF formalism (Taylor et al., 2001; Ke
et al., 2004; Deng et al., 2004; Brandbyge et al., 2002; Xue et al., 2001), and this
provides rigorous theoretical foundation for the existing DFT-based methodologies
(Lang, 1995; Taylor et al., 2001; Ke et al., 2004; Deng et al., 2004; Brandbyge et al.,
2002; Xue et al., 2001) calculating the steady currents through molecular devices.
In addition to the conventional exchange-correlation functional, a new density func-
tional is introduced to account for the dissipative interaction between the reduced
system and the environment. In the WBL approximation, the new functional can
be expressed in a relatively simple form which depends implicitly on the electron
density of the reduced system. Since the basic variable in our formulation is the
reduce single-electron density matrix, the linear-scaling techniques such as that of
reference (Yam et al., 2003) can be adopted to further speed up the computation.
Acknowledgments Authors would thank Hong Guo, Jiang-Hua Lu, Jian Wang, Arieh Warshel
and Weitao Yang for stimulating discussions. Support from The Hong Kong Research Grant
Council (HKU 7010/03P) and the committee for Research and Conference Grants (CRCG) of
The University of Hong Kong is gratefully acknowledged.
References
Brandbyge M. et al., 2002, Phys. Rev. B 65, p. 165401.

Datta S., 1995, Electronic Transport in Mesoscopic Systems (Cambridge University Press).
Deng W.-Q., Muller R. P. and Goddard W. A. III, 2004, J. Am. Chem. Soc. 126, p. 13563.
Heurich J., Cuevas J. C., Wenzel W. and Schön G., 2002, Phys. Rev. Lett. 88, p. 256803.
Hohenberg P. and Kohn W., 1964, Phys. Rev. 136, p. 64.
Jauho A.-P., Wingreen N. S. and Meir Y., 1994, Phys. Rev. B 50, p. 5528.
Ke S.-H., Baranger H. U. and Yang W., 2004, J. Am. Chem. Soc. 126, p. 15897.
Keldysh L. V., 1965, JETP 20, p. 1018.
Kohn W. and Sham L. J., 1965, Phys. Rev. 140, p. 1133.
Krantz S. G. and Parks H. R., 2002, A Primer of Real Analytic Functions (Birkhauser, Boston).
Landauer R., 1970, Philos. Mag. 21, p. 863.
Lang N. D., 1995, Phys. Rev. B 52, p. 5335.
Lang N. D. and Avouris Ph., 2000, Phys. Rev. Lett. 84, p. 358.
Langreth D. C. and Nordlander P., 1991, Phys. Rev. B 43, 2541.
Runge E. and Gross E. K. U., 1984, Phys. Rev. Lett. 52, p. 997.
Stefanucci G. and Almbladh C.-O., 2004, Europhys. Lett. 67 (1), p. 14.
Taylor J., Guo H. and Wang J., 2001, Phys. Rev. B. 63, p. 245407.
Wang C.-K. and Luo Y., 2003, J. Chem. Phys. 119, p. 4923.
Wesolowski T. A. and Warshel A., 1993, J. Phys. Chem. 97, p. 8050.
Xue Y., Datta S. and Ratner M. A., 2001, J. Chem. Phys. 115, p. 4292.
Yam C. Y., Yokojima S. and Chen G. H., 2003, J. Chem. Phys. 119, p. 8794; Phys. Rev. B 68,
p. 153105.
Yokojima S., Chen G. H., Xu R. and Yan Y., 2003, Chem. Phys. Lett. 369, p. 495; J. Comp. Chem.
24, p. 2083.
Author Index
Beling, C. D., 119 Lin, K. J., 141

Lin, M. C., 211
Chan, W. K., 119, 187 Lin, Y. H., 141
Chen, C. D., 211 Ling, C. C., 119
Chen, C. H., 181 Liu, C., 41, 151
Chen, F. R., 141 Liu, C. J., 211
Chen, G. H., 237
Chen, Y. T., 211 Man, K. Y., 187
Cheng, H. M., 29, 41, 151
Cheng, I. C., 211 Qian, T. Z., 99
Cheng, K. W., 187
Cheung, C. K., 119 Ren, W. C., 29
Chu, C. J., 211
Chu, M. W., 181 Shen, S. Q., 83
Cong, H. T., 151 Sheng, P., 99
Cui, X. D., 83 Shieh, D. B., 211
Datta, S., 59 Tam, K. H., 119

Ding, L., 83 Tang, D. M., 151
Djurišić, A. B., 119 Tang, Y. B., 151
Tang, Z. K., 49, 193
Endo, M., 3, 9 Tsai, L. C., 211
Tse, C. W., 187
Fung, S., 119
Ge, W. K., 83 Wang, C., 135

Geng, H. Z., 15 Wang, J. N., 83
Wang, X. P., 99
He, H. T., 83 Wong, K. M. C., 219
Hou, J. G., 203 Wu, C. S., 211
Wu, H. C., 211
Kai, J. J., 141 Wu, Y. N., 211
Wu, Y. P., 211
Lee, T. C., 141
Lee, Y. H.15 Yam, V. W. W., 219
Lei, S. B., 135 Yang, C. L., 83
Leung, Y. H., 119 Ye, J. T., 49, 193
Li, F., 41
Li, I. L., 49 Zhai, J. P., 49
Li, Z. M., 49 Zhang, F. C., 83
Lin, H. Y., 211 Zhang, H. L., 41
Lin, J. J., 141 Zheng, X., 237
245
Subject Index
AlN, 149–174 Conductivity, 62, 63, 140, 144, 145, 149, 150
Aluminophosphate, 50 CoPc, 205–207
Ballistic transport, 61, 62 Diffusive transport, 62, 109

DNA molecule, 209, 212–215
Capacitor, 79
Carbon nanotubes, 3 EELS, 152, 166, 170, 179–184
anode material, 5, 41, 42, 44, 45, 47 Electrochemical potential, 63, 64, 67, 68, 75
band gap, 20, 21 Electron gas, 83–86, 89–92, 95, 96, 179–181
capacitor, 5, 6, 29 Electronic transport, 139, 146
carbon nanotube-filled nylon, 11 Electron spin, 83, 84, 86, 87
catalytically grown, 9 Energy dissipation, 60, 69, 99, 100, 103–107,
conductance, 23, 24 114, 115
conductivity, 15, 20–25 Equilibrium, 60, 63, 67–69, 73, 83, 105, 107,
double-walled, 10 191, 235, 236
electrode material, 5, 41, 44, 45, 47
energy storage, 5, 11, 12, 41, 46 Field emission, 123, 141, 149, 151, 166–170,
field emission, 11, 12, 16 174, 182
large scale production, 5, 9, 10
lithium ion battery, 5, 41, 44, 47
material quality factor, 16, 20, Gold sulfido, 219, 220
22–25
microcatheter, 11, 12 Hydrodynamics, 99, 100, 110, 114
multi-walled, 4, 9, 10, 29 Hydrothermal Growth, 49, 50, 117, 118
radial breathing mode, 19, 20, 22, 31–34,
36, 38, 54, 55 Immiscible fluid, 99, 101, 102
Raman, 16, 19, 20, 22, 25, 29–38, 45, 46, InGaAs, 83, 84, 94, 96
50–52, 54, 55 Iodine molecular chain, 191, 192
resistance, 15–17, 23, 24
semiconducting, 20–23, 34, 35 Kerr rotation, 84, 89–91, 95
SEM image, 4, 17, 25, 42, 43, 46
single-walled (SWCNTs), 4, 9, 10, 16, 29,
Mass spectra, 52–54
31, 33–37, 49, 50, 55
Maxwell’s demon, 59–62, 71, 77
solid electrolyte, 41
TEM image, 5, 10, 17, 31, 33
Transmittance, 15–17, 23–25 Nanorod, 117–119, 122–128, 151–155,
transparent conducting films, 15 167, 174
Circular photogalvanic effect, 83–89, 95 Nitride, 149–152, 174
Conductance, 62–66, 72, 73, 76, 89, 204,
209, 211 Open electronic system, 235, 236, 242
247
248 Subject Index
Photoluminescence, 118–120, 122, 123, 152, SEM image, 124, 142, 146, 153, 155, 159–161,
169, 223 163–165, 167, 170–172, 211, 213
Photovoltaic, 185, 187–189 Silicon nanowire, 209–214
Platinum terpyridyl, 224–226 Spin-galvanic effect, 83
Pyrolysis reaction, 54 Spin photocurrent, 83–89, 95
STM, 133–137, 201–207
Quantum manipulation, 201, 202
Raman, 119, 120, 192–197 TEM image, 125, 153, 156, 160–166, 170,
Resistance, 61, 85, 139–147, 151, 169, 185, 173, 231, 232
188, 209, 213, 214 Tetranuclear copper, 218–222
Rhenium diimine, 223 Trinuclear copper, 221
RuO2 , 140–143, 147
Zeolite, 49, 53, 191–193, 197
Semiconducting, 143, 149 Zinc oxide, 117, 123–128

Zikang Tang, Ping Sheng Nanoscale Phenomena Basic Science To Device Applications Lecture Notes in Nanoscale Science and Technology 2007

Uploaded by

Copyright:

Available Formats

Zikang Tang, Ping Sheng Nanoscale Phenomena Basic Science To Device Applications Lecture Notes in Nanoscale Science and Technology 2007

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Zikang Tang, Ping Sheng Nanoscale Phenomena Basic Science To Device Applications Lecture Notes in Nanoscale Science and Technology 2007

Uploaded by

Copyright:

Available Formats

Nanoscale Phenomena

Basic Science to Device Applications

Volumes Published in this Series:

Zikang Tang and Ping Sheng

Hong Kong University of Science and Technology

ISBN-13: 978-0-387-73047-9 e-ISBN-13: 978-0-387-73048-6

Library of Congress Control Number: 2007931614

c 2008 Springer Science+Business Media, LLC

Printed on acid-free paper.

Zikang Tang and Ping Sheng

Part I Nano Structured Carbons and their Applications

1 Science and Technology of Nanocarbons

2 Catalytically-Grown Carbon Nanotubes and Their Current

3 Transparent Conducting Films by Using Carbon Nanotubes

4 Raman Spectroscopy on Double-Walled Carbon Nanotubes

5 Interface Design of Carbon Nano-Materials for Energy Storage

6 Formation Mechanism of 0.4 nm Single-Walled Carbon Nanotubes

Part II Quantum Dots and Molecular Spintronics

7 Nanodevices and Maxwell’s Demon

8 Manipulating Electron Spins in an InGaAs/InAlAs Two-

9 Continuum Modelling of Nanoscale Hydrodynamics

10 Defect in Zinc Oxide Nanostructures Synthesized by a

Part III Nano Materials Design and Synthesis

11 Towards Surface Science Studies of Surfaces Formed by Molecular

12 Electronic Transport Through Metal Nanowire Contacts

13 Synthesis and Properties of Quasi-One-Dimensional Nitride

14 Electron Energy-Loss Spectroscopy for Nannomaterials

15 Fabrication of Photovoltaic Devices by Layer-by-Layer

16 Optical Properties of Arrays of Iodine Molecular Chains Formed

Part IV Molecular Electronics

17 Quantum Manipulation at Molecule Scale

18 Silicon-Based Nano Devices for Detection of DNA Molecules

19 From Simple Molecules to Molecular Functional Materials

20 First-Principles Method for Open Electronic Systems

Author Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 245

Subject Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 247

C. D. Beling, Department of Physics, F. R. Chen, Department of Engineering

X. D. Cui, Department of Physics, The M. W. Chu, Center for Condensed

Chen Wang, National Center for C. J. Chu, Institute of Physics,

Hong-Tao Cong, Shenyang National H. T. He, Department of Physics and

Supriyo Datta, School of Electrical J. G. Hou, Hefei National Laboratory

Lu Ding, Department of Physics and J. J. Kai, Department of Engineering

Hong-Zhang Geng, Department of Z. M. Li, Department of Physics,

J. J. Lin, Institute of Physics & S. Q. Shen, Department of Physics,

J. N. Wang, Department of Physics Vivian Wing-Wah Yam, Department

Keith Man-Chung Wong, Centre for J. T. Ye, Department of Physics,

H. C. Wu, Department of Chemistry, X. Zheng, Department of Chemistry,

∗ Faculty of Engineering, Shinshu University, 4-17-1 Wakasato, Nagano-shi 380–8553, Japan

Z. Tang, P. Sheng (eds), Nano Science and Technology. 3

1.2 Carbon Nanotubes

synthesis of carbon nanotube via a catalytic chemical vapor deposition (CVD)

1.3 Anode Material for Lihitum Ion Battery

1.4 Electrode Material for Electric Double Layer Capacitor

Fig. 1.2 Tree of nanocarbons with basic properties and applications

1.5 Conclusion and Prospect

in nanoscale, will function as a starting point of industrial innovation. The technical

One-dimensional carbon nanotubes have attracted a particular interest due to their

2.2 Large Scale Production of Carbon Nanotubes

In order to use carbon nanotubes in novel devices, it is necessary to produce these

G max = e2 / h ≈ 25.8 K (7.9)