Plasma-Surface Modification of Biomaterials

Topics in Applied Physics
Volume 100
Topics in Applied Physics is part of the SpringerLink service. For all customers with standing
orders for Topics in Applied Physics we offer the full text in electronic form via SpringerLink
free of charge. Please contact your librarian who can receive a password for free access to the full
articles by registration at:
springerlink.com → Orders
If you do not have a standing order you can nevertheless browse through the table of contents of
the volumes and the abstracts of each article at:
springerlink.com → Browse Publications
Topics in Applied Physics
Topics in Applied Physics is a well-established series of review books, each of which presents a
comprehensive survey of a selected topic within the broad area of applied physics. Edited and written
by leading research scientists in the field concerned, each volume contains review contributions
covering the various aspects of the topic. Together these provide an overview of the state of the art
in the respective field, extending from an introduction to the subject right up to the frontiers of
contemporary research.
Topics in Applied Physics is addressed to all scientists at universities and in industry who wish to
obtain an overview and to keep abreast of advances in applied physics. The series also provides easy
but comprehensive access to the fields for newcomers starting research.
Contributions are specially commissioned. The Managing Editors are open to any suggestions for
topics coming from the community of applied physicists no matter what the field and encourage
prospective editors to approach them with ideas.
Managing Editors
Dr. Claus E. Ascheron Dr. Hans J. Koelsch
Springer-Verlag GmbH Springer-Verlag New York, LLC
Tiergartenstr. 17 233, Spring Street
69121 Heidelberg New York, NY 10013
Germany USA
Email: claus.ascheron@springer.com Email: hans.koelsch@springer.com
Assistant Editor
Adelheid H. Duhm
Springer-Verlag GmbH
Tiergartenstr. 17
69121 Heidelberg
Germany
Email: adelheid.duhm@springer.com
Giacomo Messina Saveria Santangelo (Eds.)
Carbon
The Future Material for
Advanced Technology Applications
With 245 Figures, 6 in Color, and 24 Tables
123
Giacomo Messina
Saveria Santangelo
Faculty of Engineering
Department of Mechanics and Materials
University Mediterranea of Reggio Calabria
89060 Reggio Calabria, Italy
messina@ing.unirc.it
santange@ing.unirc.it
Library of Congress Control Number: 2006924284
Physics and Astronomy Classification Scheme (PACS):

81.05.Uw, 81.07.-b, 60., 70., 80.
ISSN print edition: 0303-4216

ISSN electronic edition: 1437-0859
ISBN-10 3-540-29531-3 Springer Berlin Heidelberg New York
ISBN-13 978-3-540-29531-0 Springer Berlin Heidelberg New York
This work is subject to copyright. All rights are reserved, whether the whole or part of the material is concerned,
specifically the rights of translation, reprinting, reuse of illustrations, recitation, broadcasting, reproduction on
microfilm or in any other way, and storage in data banks. Duplication of this publication or parts thereof is
permitted only under the provisions of the German Copyright Law of September 9, 1965, in its current version, and
permission for use must always be obtained from Springer. Violations are liable for prosecution under the German
Copyright Law.
Springer is a part of Springer Science+Business Media
springer.com
© Springer-Verlag Berlin Heidelberg 2006

Printed in Germany
The use of general descriptive names, registered names, trademarks, etc. in this publication does not imply, even in
the absence of a specific statement, that such names are exempt from the relevant protective laws and regulations
and therefore free for general use.
Typesetting: DA-TEX · Gerd Blumenstein · www.da-tex.de
Production: LE-TEX Jelonek, Schmidt & Vöckler GbR, Leipzig
Cover design: design & production GmbH, Heidelberg
Printed on acid-free paper 57/3100/YL 543210
To Lorenzo Maria, my adored son,
with love
Saveria
To my sons Gloria and Alfonso,

with love
Giacomo
Preface
Carbon is a surprisingly versatile element, able to hybridise in three differ-

ent states, sp1 , sp2 and sp3 . The changes in local bonding of carbon atoms
account for the existence of extremely diverse allotropic phases, exhibiting
a quite broad range of physical and chemical properties. This element can
crystallise as diamond (sp3 hybridisation) or graphite (sp2 hybridisation)
and give rise to many noncrystalline phases (generally containing a mixture
of sp1 , sp2 and sp3 hybridisations), such as fullerenes, carbon nanotubes and
disordered, nanostructured and amorphous carbons.
Strong tetrahedral σ bonds are responsible for the extreme physical prop-
erties of diamond, a wide gap semiconductor having the largest bulk modulus
of any solid, the highest atom density, the largest room-temperature thermal
conductivity, the smallest thermal expansion coefficient and the largest limit-
ing electron and hole velocities of any semiconductor. Graphite, whose sheets
(graphenes), featured by strong intralayer trigonal σ bonding, are held to-
gether by weak interlayer van der Waals’ forces, is an anisotropic metal.
These two crystalline allotropic forms of carbon have been known and
utilised for centuries, together with other carbonaceous materials, generically
identified as coals. The use of hydrocarbons and organic materials is more
recent. The true turning point in the use of carbon as a material for advanced
technology applications, however, occurs only in the second half of the last
century.
In 1953, the discovery, at Union Carbide Co., of the possibility of grow-
ing diamond films by chemical vapour deposition (CVD), at low pressures,
opened the way to the development of low-cost deposition techniques. The
growth of CVD diamond films at subatmospheric pressures, firstly reported
by Derjaguin and Fedoseev in 1967 and later implemented by Matsumoto’s
group, opened new perspectives in technological applications of diamond.
The enthusiasm for these achievements caused further efforts to be addressed
to the development of low-temperature CVD techniques and stimulated the
search for alternative growth methods, such as the QQC process, based
on a multiplexed laser–solid interaction, announced in the 1990s by QQC
Inc. (USA).
Around 1969, Aisenberg and Chabot demonstrated that carbon films with
properties ranging between those of diamond (diamond-like carbon, DLC)
VIII Preface
and those of graphite (graphite-like carbon, GLC) can be deposited by the

use of energetic carbon species. The term amorphous carbon (a-C) identifies
a carbon matrix with any possible mixture of sp1 -, sp2 - and sp3 -hybridised
sites and no crystalline long-range order. Indeed, a wide nomenclature for
the classification of these films does exist. Hydrogenated amorphous carbon
(a-C:H) denotes an a-C film containing up to even 60% hydrogen. The term
DLC, initially introduced to identify exclusively a-C films with hardness and
transparency properties close to those of diamond, is nowadays commonly
used to designate a larger class of materials, including both hydrogen-free and
hydrogenated a-Cs, containing a significant fraction of sp3 hybridised carbon
sites and exhibiting a wide spectrum of properties and potentialities. Finally,
the name tetrahedral amorphous carbon (ta-C) is attributed to hydrogen-
free a-Cs containing up to 95% sp3 sites, sometimes also termed amorphic
diamond (a-D), whose correspondent hydrogenated form is known as ta-C:H.
In 1985, spectrometric measurements on interstellar dust led to the discov-
ery of new allotropic forms of carbon, consisting of pentagonal and hexagonal
rings arranged to form closed-cage molecules. However, the scientific interest
in fullerenes, started only in 1990 after the invention of the method for their
production in macroscopic amounts, was really fuelled by the awarding of the
Nobel Prize in chemistry for this discovery in 1996.
In 1989, Liu and Cohen theoretically predicted the existence of a crys-
talline β-C3 N4 phase with bulk modulus and hardness higher than diamond.
The attempts of synthesising such a hypothetical superhard compound drew
attention upon carbon nitride (CN). In spite of the lack of any agreed success
on this direction, in the last years the interest in amorphous CN compounds
has been renewed, due to the wide range of possible applications.
The observation, in 1993, of single-walled carbon nanotubes (SWCNTs),
consisting of single graphene sheets wrapped into hollow cylinders having
nanometer-sized diameter, followed to the discovery, about two years ear-
lier, of multiwalled carbon nanotubes (MWCNTs), comprising up to tens of
coaxial cylinders, caused enormous interest in many fields of research.
The exciting discoveries of the last decades have given great impulse to
the research on carbon-based materials, offered new opportunities and issued
new challenges. The lengthy history of diamond as a material for advanced
technology began in the 1920s with the development of photoconductive UV
detectors, followed in the 1940s by the fabrication of ionising radiation detec-
tors, both based on the use of geological diamonds. In 1979, the experiments
of Himpsel, followed by those of Pate in 1986, demonstrated the high quantum
efficiency for photoelectron emission from natural diamond crystals, starting
the consideration of diamond as an emitting material in cold cathode sources
and devices. Nowadays, a great variety of technological applications in fields
as diverse as mechanics, lithography, optics, chemistry, electronics, takes full
advantage of the extraordinary properties of higher and higher quality syn-
thetic diamond films, achieved thanks to the continuous advances in prepara-
Preface IX
tion methods. Surprisingly, however, there are even manifold applications of

“damaged diamond”. These benefit from the dramatic modifications of elec-
trical, mechanical, optical and chemical properties induced on virgin material
by ion-beam exposure. The realisation of ohmic contacts in diamond-based
electronic devices and the production of microstructures on single-crystal
diamond membranes for micromachining exploit the local complete graphiti-
sation process. Partial graphitisation, instead, is employed for the realisation
of electrodes for electrochemistry and of conductive patterns on diamond.
The cheaper production costs, along with the significant fraction of sp3
bonding conferring on DLC many of the profitable properties of diamond,
make it a valuable material for widespread technology applications. The first
exploitation of DLC has been as a hard protective coating. During the last 30
years, taking advantage of its low friction, the use of DLC as coating has been
enlarged to the solid lubricant technology. Its good blood compatibility sug-
gests of extending its utilisation as biomaterial in blood-contacting-devices.
Nowadays, DLCs coat a surprisingly large number of unexpectedly different
objects, spanning from microelectromechanical devices to dental prostheses,
from optical windows to razor blades, from orthopedic pins and screws to car
parts, from medical guide-wires to disks, heads and slides used in magnetic
storage technology.
Recently, there has been considerable interest in the application of low-
threshold field-emitting carbon films for cold cathode devices and flat panel
displays. The interest in amorphous hydrogen-free and hydrogenated CN
compounds is, instead, mainly concerned with their possible application as
electrodes, IR detectors and gas sensors. Group-IV carbon-based binary al-
loys, like amorphous silicon and germanium carbon-alloys, are used for opto-
electronic devices and for the construction of solar cells.
Fullerenes, whose family comprises a large variety of isomers, are used as
sorbents, as superconductors, as chromatographic stationary phases, as mate-
rials for building optical and electronic devices, hydrogen cells, photosensitive
elements and electrochemical sensors. Thanks to their astonishing properties,
fullerenes and their derivatives appear to be a good alternative to polymer
molecules currently used in nanolithography, as well as promising candidates
as antioxidants, neuroprotectors and inhibitors of bacterial growth.
The unique structural features of SWCNTs and MWCNTs are reflected
in their singular optical, electronic and mechanical properties. Thanks to the
wide variety of geometrical-structure-dependent behaviours, CNTs possess
great potential in many research fields, spanning from physics to chemistry,
from engineering to materials science, from biology to life science. Although
gaining the control of diameter and chirality of CNTs still represents the
goal needed for the defect-free CNT production and the full success of the
CNT-based technological applications, they are presently regarded as build-
ing blocks for various nanoscale devices. Current applications relate to mainly
electronic devices, such as field emitters, transistors and sensors, while one of
X Preface
the future challenges consists in the creation of highly integrated CNT-based

cicuits. The recent fabrication of the world’s smallest motor based on CNTs
has opened the way to nanoelectromechanical systems, while the realisation
of the first AFM tip using CNTs has laid the foundations for a real revolution
in high-resolution imaging techniques.
It appears now clear that carbon-based materials constitute a topic of

huge scientific interest and great strategic importance in an interdiscipli-
nary approach spanning applied physics, materials science, biology, mechan-
ics, electronics and engineering. Development of current materials, advances
in their applications and discovery of new forms of carbon are the themes
addressed by the frontier research in these fields.
The book covers all the fundamental topics concerned with amorphous
and crystalline carbon-based materials, such as synthetic diamond, diamond-
like carbon, carbon alloys, carbon nanotubes and cluster-assembled materi-
als. The latest scientific results and developments in the various technological
application areas of these materials are reviewed. Both theoretical and exper-
imental aspects related to their optical, electrical, elastic, thermomechanical,
structural, vibrational and electronic properties are extensively discussed.
The contributions discuss subjects of great currency and huge scientific
and technological interest, such as the use of carbon-based materials for en-
ergy storage, for fabrication of sensors operating in harsh environments and
as biocompatible coatings for medical implants. A broad spectrum of both
well-assessed and still controversial as well as newly developed aspects con-
cerning growth and characterisation techniques are addressed, ranging from
modelling of deposition processes to semiempirical methods for its control
and optimisation, from investigation of diamond-nucleation and nanotube-
growth mechanisms to more comprehensive explanation of the origin of the
Raman features commonly regarded as the fingerprint of amorphous carbons.
An accurate picture is drawn of both the present developments in research
on carbon and the future prospects that may constitute a reference point for
a new generation of researchers, who accept the challenge issued by the scien-
tific community in this fascinating field. The volume, comprising 24 contribu-
tions, is the first comprehensive state-of-the-art review of this fast-evolving
field of research. Its preparation, to which many of the most world-famous
researchers engaged in the field have contributed, has acted as a friendly and
cooperative opportunity to share the latest outstanding and exciting results
in the hot topics of research on carbon. The 75 contributors, most of whom
often promote scientific events and organise international meetings on carbon
and related subjects, belong to a multitude of disciplines. This witnesses the
interdisciplinary nature of the fundamental and technology-centred studies
of carbon, a material entering everyday life, on which worldwide attention is
focused for its extraordinary future potential.
The contributions are presented in alphabetic order because any classifi-
cation relative to material kind (diamond, carbon alloys, carbon nanotubes,
Preface XI
etc.) or to contribution type (fundamental studies, technological applications,

etc.) or subject (growth techniques, characterisation tools, etc.) might be re-
ductive for the reader. Most of the contributions, in fact, deal with aspects
common to more than one of these topics.
We end this preface by acknowledging all the scientists who participated
enthusiastically and painstakingly to the realisation of this volume and to
the Springer staff for the valuable assistance provided.
Reggio Calabria, Giacomo Messina

26 August 2005 Saveria Santangelo
Contents
Aid of Scaling Laws in the Achievement of a Well-Controlled

Film Deposition Process
Giacomo Messina, Saveria Santangelo . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
A Spectroscopic Approach to Carbon Materials for Energy
Storage
Giuseppe Zerbi, Matteo Tommasini, Andrea Centrone,
Luigi Brambilla, Chiara Castiglioni . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
1.1 Carbon-Based Materials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
2 Prototypical Lattices in Three, Two and One Dimensions . . . . . . . . . 26
2.1 Diamond . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
2.2 Graphite . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
2.3 Polyacetylene . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
2.4 Polyynes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
3 A Spectroscopic Approach to Disordered Carbon Materials . . . . . . . 33
3.1 1D Systems – Saturated Carbon Materials . . . . . . . . . . . . . . . . . 33
3.2 Polyconjugated Polymers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
3.3 2D sp2 Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
4 Edge Effects in Graphitic Domains for Structure Diagnosis . . . . . . . 36
5 Approaching the Structure of Carbonaceous Materials by
Vibrational Spectroscopy: Imagination and Reality . . . . . . . . . . . . . . 41
5.1 Carbon-Based Materials for Hydrogen Storage . . . . . . . . . . . . . . 44
6 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
Biocompatibility, Cytotoxicity and Bioactivity of Amorphous
Carbon Films
Sandra E. Rodil, René Olivares, Higinio Arzate, Stephen Muhl . . . . . . . 55
1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
2 Experimental Details . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
2.1 Film Deposition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
2.2 Film Characterization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
2.3 Cell Preparation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
XIV Contents
2.4 Cytotoxicity Test . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59

2.5 Bioactivity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
2.5.1 Morphological Assay . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
2.5.2 Protein Synthesis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
3 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
3.1 Film Properties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
3.2 Cytotoxicity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
3.3 Bioactivity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
3.3.1 Morphological Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
3.3.2 Proteins . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
4 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
Characterisation of the Growth Mechanism during PECVD
of Multiwalled Carbon Nanotubes
Martin S. Bell, Rodrigo G. Lacerda, Kenneth B.K. Teo,
William I. Milne . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
2 Production of Carbon Nanotubes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82
3 Plasma Composition during PECVD . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
4 Characterisation of the Growth Mechanism . . . . . . . . . . . . . . . . . . . . . 85
5 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90
Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92
Correlation Between Local Structure and Film Properties
in Amorphous Carbon Materials
Giovanni Fanchini, Alberto Tagliaferro . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95
1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95
2 Electronic States in Carbon . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95
3 Hybridisation and Local Structure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96
4 The Various Forms of Amorphous Carbon . . . . . . . . . . . . . . . . . . . . . . 97
5 Phase Matching and Its Effects on Materials . . . . . . . . . . . . . . . . . . . . 99
6 Optoelectronic and Mechanical Properties . . . . . . . . . . . . . . . . . . . . . . 102
7 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104
Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104
Defects in CVD Diamond Films from Their Response
as Nuclear Detectors
Marco Marinelli, Enrico Milani, Aldo Tucciarone, Gianluca Verona
Rinati . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107
1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107
2 CVD Diamond Nuclear Detectors: Realization and Physics . . . . . . . 108
Contents XV
3 Analysis of the Charge Collection Spectrum . . . . . . . . . . . . . . . . . . . . 111

3.1 Qualitative Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111
3.2 Quantitative Analysis: The General Model . . . . . . . . . . . . . . . . . 115
3.3 The Use of Detector Thickness . . . . . . . . . . . . . . . . . . . . . . . . . . . 117
3.4 The Use of Penetration Depth . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119
4 Time-Domain Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123
4.1 The Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123
4.2 Qualitative Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125
4.3 Quantitative Analysis: Computer Simulation . . . . . . . . . . . . . . . 126
5 Temperature Effects: Depumping . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 129
6 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133
Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 135
Effects of Nanoscale Clustering in Amorphous Carbon
J. David Carey, S. Ravi P. Silva . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137
1 Introduction and Bonding in Carbon . . . . . . . . . . . . . . . . . . . . . . . . . . . 137
2 Disorder in Amorphous Carbon . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 139
3 Intracluster Effects in Amorphous Carbon . . . . . . . . . . . . . . . . . . . . . . 142
4 Intercluster Interactions in Amorphous Carbon . . . . . . . . . . . . . . . . . . 147
5 Field Emission from Amorphous Carbon . . . . . . . . . . . . . . . . . . . . . . . 148
6 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 150
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 150
Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 151
Elastic and Structural Properties of Carbon Materials

Investigated by Brillouin Light Scattering
Marco G. Beghi, Carlo S. Casari, Andrea Li Bassi, Carlo E. Bottani . . 153
1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 153
2 Experimental Technique . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 156
3 Derivation of the Elastic Constants . . . . . . . . . . . . . . . . . . . . . . . . . . . . 158
4 Brillouin Scattering from Carbonaceous Materials . . . . . . . . . . . . . . . 161
5 Ultra-Thin Carbon Films . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 162
6 Nanotubes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 166
7 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 170
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 170
Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 173
Electrical Resistivity and Real Structure of Magnetron-
Sputtered Carbon Films
Alexei A. Onoprienko . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 175
1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 175
2 Experimental . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 176
3 Results and Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 177
3.1 Substrate Temperature Effects . . . . . . . . . . . . . . . . . . . . . . . . . . . . 177
XVI Contents
3.2 Substrate Bias Effects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 181

4 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 184
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 185
Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 186
Formation, Atomic Structures and Properties of Carbon
Nanocage Materials
Takeo Oku, Ichihito Narita, Atsushi Nishiwaki, Naruhiro Koi,
Katsuaki Suganuma, Rikizo Hatakeyama, Takamichi Hirata, Hisato
Tokoro, Shigeo Fujii . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 187
1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 187
2 Synthesis Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 189
3 Fullerene Clusters and Metallofullerenes . . . . . . . . . . . . . . . . . . . . . . . . 190
4 Onions and Nanotubes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 196
5 Carbon Nanocapsules . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 200
6 Properties of Carbon Nanomaterials . . . . . . . . . . . . . . . . . . . . . . . . . . . 204
6.1 Photoluminescence of Carbon Nanocapsules . . . . . . . . . . . . . . . . 204
6.2 Magnetic Properties of Carbon Nanocapsules . . . . . . . . . . . . . . . 205
6.3 Possibility of H2 Gas Storage in Carbon Nanocages . . . . . . . . . 207
6.4 One-Dimensional Self-Organization of Nanocapsules . . . . . . . . . 209
7 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 212
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 212
Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 215
Hard Amorphous Hydrogenated Carbon Films and Alloys
Fernando L. Freire Jr. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 217
1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 217
2 Film Growth and Chemical Composition . . . . . . . . . . . . . . . . . . . . . . . 220
3 Film Microstructure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 224
4 Mechanical and Nanotribological Properties . . . . . . . . . . . . . . . . . . . . 229
5 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 234
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 235
Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 237
Ion Microscopy on Diamond
Claudio Manfredotti . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 239
1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 240
2 IBIC . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 241
3 Lateral IBIC . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 246
4 IBIL . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 251
5 XBIC . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 257
6 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 263
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 264
Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 265
Contents XVII
Measurements of Defect Density Inside CVD Diamond Films

Through Nuclear Particle Penetration
Renato Potenza, Cristina Tuvé . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 267
1 Diamond as Radiation Detector . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 267
1.1 Properties of Diamond . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 267
1.2 Lattice Defects in Synthetic Diamond Crystals . . . . . . . . . . . . . 269
1.3 Application of Diamond to Beam and Beam Profile Monitoring271
2 Mechanism of Conduction in Circuits Including Diamond . . . . . . . . . 273
2.1 Modified Hecht’s Model for Charge Transport Inside Diamond 275
2.1.1 Absorption of Carriers During Their Transport Along
the Electric Field Lines . . . . . . . . . . . . . . . . . . . . . . . . . . . . 275
2.1.2 Edge Effects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 276
2.2 Effect of Nonuniform Bragg’s Deposit of Charge into Diamond 278
2.3 Nonuniform Distributions of Charge and Defects . . . . . . . . . . . . 279
3 Charge Collection Efficiencies of Diamond Detectors
Under Ion Bombardment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 280
3.1 Experimental Procedure to Measure the Charge Collection
Efficiencies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 282
3.2 Corrections to the Efficiencies η± Needed to Into Take
Account the Pulse Height Defect of Diamond for Detectors
Heavy Ionizing Particles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 283
4 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 285
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 285
Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 286
Laser Ablation-Deposited CNx Thin Films
Enza Fazio, Enrico Barletta, Francesco Barreca, Guglielmo Mondio,
Fortunato Neri, Sebastiano Trusso . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 287
1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 287
2 Experiments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 288
4 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 300
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 301
Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 302
Modeling of the Transport Properties of Diamond Radiation
Sensors
Stefano Lagomarsino, Silvio Sciortino . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 303
1 Models of the Polycrystalline Diamond Band Gap . . . . . . . . . . . . . . . 303
1.1 Band A . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 304
1.2 Trapping Centers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 304
1.3 Carrier Lifetimes and Charge Collection Distance . . . . . . . . . . . 305
1.4 Location of the Trapping and Recombination Centers in the
Diamond Band Gap . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 306
1.5 Unipolar Conduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 308
XVIII Contents
2 A General Model for Transport Properties of pCVD Diamond:

Underlying Assumptions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 309
3 Conductivity Under Exposure to Ionizing Radiation . . . . . . . . . . . . . 311
4 Thermal Relaxation of Uniform Trap Level Distributions . . . . . . . . . 316
4.1 Radiation Induced Conductivity Transient at Different
Fading Times . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 317
4.2 Persistent Radiation-Induced Conductivity . . . . . . . . . . . . . . . . . 319
5 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 323
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 325
Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 326
Nucleation Process of CVD Diamond on Molybdenum
Substrates
Giuliana Faggio, Maria G. Donato, Stefano Lagomarsino, Giacomo
Messina, Saveria Santangelo, Silvio Sciortino . . . . . . . . . . . . . . . . . . . . . . . 329
1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 329
2 Experimental . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 330
3.1 Raman and Photoluminescence Analysis . . . . . . . . . . . . . . . . . . . 331
3.1.1 Discontinuous Samples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 332
3.1.2 Continuous Samples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 334
3.2 Statistical Study of the Nucleation Process . . . . . . . . . . . . . . . . . 336
4 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 340
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 341
Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 343
Optical Characterisation of High-Quality Homoepitaxial
Diamond
Maria G. Donato, Giuliana Faggio, Giacomo Messina, Saveria
Santangelo, G. Verona Rinati . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 345
1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 345
2 Experimental . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 348
3 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 348
3.1 Optical and SEM Characterisation . . . . . . . . . . . . . . . . . . . . . . . . 348
3.2 Raman Characterisation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 350
3.2.1 First-Order Raman Scattering . . . . . . . . . . . . . . . . . . . . . . 350
3.2.2 Second-Order Raman Scattering . . . . . . . . . . . . . . . . . . . . . 351
3.3 Photoluminescence Characterisation . . . . . . . . . . . . . . . . . . . . . . . 352
4 Discussion and Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 355
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 357
Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 358
Contents XIX
Pulsed Laser Deposition of Carbon Films: Tailoring Structure

and Properties
Paolo M. Ossi, Antonio Miotello . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 359
1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 359
3 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 362
3.1 Low-Pressure Deposited Films . . . . . . . . . . . . . . . . . . . . . . . . . . . . 362
3.2 High-Pressure Deposited Films . . . . . . . . . . . . . . . . . . . . . . . . . . . 367
4 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 373
4.1 Low-Pressure Deposited Films . . . . . . . . . . . . . . . . . . . . . . . . . . . . 373
4.2 High-Pressure Deposited Films . . . . . . . . . . . . . . . . . . . . . . . . . . . 376
5 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 377
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 378
Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 379
Raman Spectra and Structure of sp2 Carbon-Based
Materials: Electron–Phonon Coupling, Vibrational Dynamics
and Raman Activity
Chiara Castiglioni, Fabrizia Negri, Matteo Tommasini, Eugenio Di
Donato, Giuseppe Zerbi . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 381
1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 381
2 Raman Spectra of Polyconjugated Materials . . . . . . . . . . . . . . . . . . . . 383
3 Electron–Phonon Coupling and Raman Features . . . . . . . . . . . . . . . . 385
3.1 Electron–Phonon Coupling in Polyacetylene . . . . . . . . . . . . . . . . 385
3.2 Electron–Phonon Coupling in Graphenes . . . . . . . . . . . . . . . . . . 390
3.3 Electron–Phonon Coupling in Carbon Nanotubes . . . . . . . . . . . 395
4 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 399
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 400
Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 402
Raman Spectroscopy and Optical Properties of Amorphous
Diamond-Like Carbon Films
Leonid Khriachtchev . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 403
1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 403
3 Raman Diagnostics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 405
4 Effects Induced by Interference of Light . . . . . . . . . . . . . . . . . . . . . . . . 409
5 Optical Characterization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 413
6 Some Applications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 416
7 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 418
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 419
Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 420
XX Contents
Raman Spectroscopy of CVD Carbon Thin Films Excited by

Near-Infrared Light
Margit Koós, Miklós Veres, Sára Tóth, Miklós Füle . . . . . . . . . . . . . . . . . 423
1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 423
1.1 The Raman Effect . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 423
1.2 Raman Spectra of Carbon Materials . . . . . . . . . . . . . . . . . . . . . . 426
2 Infrared Excited Raman Spectroscopy of Amorphous Carbon Thin
Films . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 429
2.1 a-C:H Thin Films Prepared from Benzene . . . . . . . . . . . . . . . . . 429
2.2 a-C:H Thin Layers Prepared From Methane . . . . . . . . . . . . . . . . 438
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 443
Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 444
The Role of Hydrogen in the Electronic Structure of
Amorphous Carbon: An Electron Spectroscopy Study
Lucia Calliari, Massimiliano Filippi, Nadhira Laidani, Gloria
Gottardi, Ruben Bartali, Victor Micheli, Mariano Anderle . . . . . . . . . . . 447
1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 447
2 Experimental . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 449
2.1 Hydrogen Incorporation in the Structure of a-C . . . . . . . . . . . . 449
2.2 Temperature-Induced Hydrogen Evolution from a-C:H . . . . . . . 450
3 Hydrogen Incorporation in the Structure of a-C . . . . . . . . . . . . . . . . . 450
4 Temperature-Induced Hydrogen Evolution from a-C:H . . . . . . . . . . . 455
5 Concluding Remarks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 459
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 461
Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 462
UV-Induced Photoconduction in Diamond
Emanuele Pace, Antonio De Sio, Salvatore Scuderi . . . . . . . . . . . . . . . . . . 463
1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 463
2 Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 464
3 Properties of Diamond . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 467
4 Synthesis of Diamond . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 468
5 Experimental Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 470
6 Electro-Optical Properties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 472
6.1 Polycrystalline Diamond Detectors . . . . . . . . . . . . . . . . . . . . . . . . 474
6.2 Single-Crystal Diamond Detectors . . . . . . . . . . . . . . . . . . . . . . . . 481
7 Single-Pixel Detectors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 485
7.1 Photodetectors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 485
7.2 Photoconductor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 485
7.3 Photodiode . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 487
8 Pixel Array Detectors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 489
9 Applications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 491
9.1 UV Lasers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 493
9.2 Photolithography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 494
Contents XXI
9.3 Space Astronomy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 495

10 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 497
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 498
Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 503
Vibrational Spectroscopy in Ion-Irradiated Carbon-Based
Thin Films
Giuseppe Compagnini, Orazio Puglisi, Giuseppe A. Baratta, Giovanni
Strazzulla . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 505
1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 505
2 Irradiation of Crystalline Carbon and Carbon Alloys . . . . . . . . . . . . . 506
3 Irradiation of Hydrocarbons (Oligomers, Polymers and Frozen
Gases) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 510
4 Irradiation of sp-Rich Amorphous Carbon Phases . . . . . . . . . . . . . . . 513
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 518
Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 519
Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 521
Aid of Scaling Laws in the Achievement
of a Well-Controlled Film Deposition Process
Giacomo Messina and Saveria Santangelo
Dipartimento di Meccanica e Materiali, Università “Mediterranea” di Reggio

Calabria, Località Feo di Vito, I-89060 Reggio Calabria, Italy, and INFM, Unità
di Ricerca di Roma “Tor Vergata”
santangelo@ing.unirc.it
Abstract. Scaling laws are profitably applied in all the scientific disciplines. Ap-
proximated solutions constitute a valuable aid for the analysis of complex systems
and physical processes and, generally, for handling problems involving many vari-
ables. We indicate the path for deriving a scaling law for the growth parameters,
referring to the synthesis of hydrogenated amorphous carbon-nitrides (a-CN:H)
by reactive sputtering as an example. Thanks to its generality, the semiempirical
method utilised here can be applied to all carbon-based materials and all deposition
techniques. We demonstrate the existence of an analytical dependence of the film
physical properties, as resulting from Raman analysis and Rutherford backscatter-
ing measurements, on a single dimensionless combination of the growth parame-
ters. We give an empirical rule for tailoring the film characteristics through specific
changes of the deposition conditions. We thus achieve the capability of easily pre-
dicting and controlling the final issue of the synthesis process.
1 Introduction
The search for approximated solutions, which are useful for describing the
behaviour of complex systems and processes, is a topic common to all the
scientific disciplines. In this context, the majority of efforts are generally de-
voted to the derivation of scaling laws, which play a key role in handling
problems involving a great number of variables. In the last years, a semi-
empirical method [1], based on a modified application of the theorem on
“physically similar systems” of dimensional analysis [2], has been proposed
as a physical approach to approximation [3, 4, 5]. According to this method,
dimensionless variable combinations (‘Q-arguments’), aimed at describing ef-
fectively the process, are generated. These combinations can be regarded as
the only independent variables and represent scaling laws for the variables
involved. Once the problem is reformulated in terms of Q-arguments, practi-
cal approximants to the physical laws governing the process are derived by a
fitting procedure to the experimental data. The process issue is, hence, finally
described by analytical functions depending on few parameters, rather than
on the many variables initially present.
Scaling laws constitute a quite helpful tool for practically controlling and
easily improving complex systems whose behaviour depends on several mech-
G. Messina, S. Santangelo (Eds.): Carbon, The Future Material for Advanced Technology Ap-
plications, Topics Appl. Phys. 100, 1–21 (2006)
2 Giacomo Messina and Saveria Santangelo
anisms of different nature. They are, hence, profitably utilised in a very wide
variety of fields [6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19], ranging from
biology to engineering, from space science to chemistry, from medicine to
materials science, from genetics to physics. They also offer a valuable aid in
studying complicated processes, such as film deposition [20, 21, 22, 23, 24],
whose evolution and final issue are determined by manyfold concomitant fac-
tors. Finding the scaling laws for the growth process allows us to achieve a
full control of it, with the consequent possibility of optimising the preparation
technique. Once the capability of tailoring the film characteristics is acquired,
high-quality films can be straightforwardly synthesised [5, 25, 26, 27, 28]. The
derivation of scaling laws, however, generally represents a difficult task whose
success strongly depends on how the process is modelled. For this the reason,
it must be faced in the more general context of the process approximation
and requires a deep understanding of the mechanisms involved.
This contribution, where the film growth approximation is addressed, in-
tends to furnish the experimentalist with a powerful tool for acquiring the
capability of readily predicting and easily driving the deposition issue, on the
basis of the empirically gained know-how. By application to the previously
developed method [1], a scaling law for the film growth process is derived
and approximated solutions are therein attained. The preparation of hydro-
genated amorphous carbon–nitrides by reactive sputtering is considered as an
example. Nonetheless, the general method utilised can be extensively applied
to all growth techniques and materials. We investigate the influence of the
deposition conditions on the physical properties of the grown films. By deriv-
ing the relationships approximating the physical laws governing the process,
empirical rules are achieved for driving it and tailoring the synthesised film
characteristics through specific changes of the growth conditions. We use Ra-
man spectroscopy and Rutherford backscattering analysis, but the results of
any other postgrowth diagnostic technique could be employed for deriving the
scaling law by the use of the same operative procedure. In particular, experi-
mental details are given in Sect. 2. The operative scaling law derivation pro-
cedure we develop and our outstanding results obtained are briefly reviewed
in Sect. 3. In Sect. 4, the very simple model of the reactive-sputtering growth
process, utilised for the scaling law derivation, is illustrated. From the discus-
sion, a deeper insight is achieved on the role played by internal microscopic
variables in determining the final issue of the a-CN:H film sputter-deposition.
Conclusions are briefly given in Sect. 5.
2 Experimental
Two sets of a-CN:H films are prepared by graphite sputtering in N2 –H2 –
He–Ar atmosphere using a conventional 13.56 MHz diode system. Samples of
series ‘D’ are utilised for deriving the scaling law. Samples of series ‘T’ are
deposited afterwards, with the purpose of testing its validity. The deposition
Scaling Laws for Film Deposition 3
Table 1. Values of the dimensionless argument Q corresponding to the growth

conditions of a-CN:H films utilised for deriving (series ‘D’) and testing (series ‘T’)
the scaling law. ΦN2 and ΦH2 indicate the reactive-gas (N2 and H2 ) flow rates, while
ΦHe and ΦAr standing for the inert-gas (He and Ar) flow rates. Note that flow rates
of inlet gases, chamber pressure (p) and rf power (W ) are varied so that at least
the value of one of these parameters is different from one deposition to another
Film Q ΦN2 ΦH2 ΦHe ΦAr p W

# sccm sccm sccm sccm mTorr W
D1 0.036 20.0 7.0 30.0 70.0 20 300
T1 0.079 8.4 3.0 70.0 30.0 38 300
D2 0.169 10.0 5.0 30.0 70.0 21 300
D3 0.347 10.0 7.0 30.0 70.0 25 300
D4 0.434 10.0 7.0 30.0 70.0 20 300
T2 0.578 10.0 7.0 30.0 70.0 20 400
D5 0.603 7.0 5.0 30.0 70.0 21 300
D6 0.715 4.2 3.0 70.0 30.0 38 220
T3 0.865 3.0 2.2 70.0 30.0 38 180
D7 0.974 4.2 3.0 70.0 30.0 38 300
temperature is 100 ◦ C, while the gas flow rates (ΦN2 , ΦH2 , ΦHe , ΦAr ), chamber
pressure (p) and rf power (W ) vary as indicated in Table 1.
The film stoichiometry is investigated by Rutherford backscattering (RBS)
and elastic recoil detection analysis (ERDA) measurements. Raman spec-
troscopy is employed for studying bonding and structural properties. Details
about film deposition and analysis, including spectra fitting and interpreta-
tion, can be found in [29, 30, 31, 32, 33].
3 Results
3.1 Film Properties
Figure 1 shows the Raman spectra of the films considered. As discussed [31,
32], the shape evolution of the D band and G bands reflects the structural and
bonding modifications, produced by the film stoichiometry changes. The most
meaningful parameters derived from Raman analysis are shown in Table 2
together with the film chemical composition.
Since six deposition parameters are independently varied during the
preparation of the samples, the representation of these results leads to hyper-
surfaces in 7-dimensional spaces. In principle, projecting such hypersurfaces
could indicate the role of each growth variable. Nevertheless, the simulta-
neous variation of the remaining deposition parameters actually causes a
confused spreading of the projected points, with a consequent lack of a clear
correlation with the changes in film stoichiometry and related properties.
Fig. 1. Raman spectra of a-CN:H

films belonging to (a) series D, used
for deriving, and (b) series T, used
for testing scaling law. Excitation
energy is 2.41 eV. Spectra decompo-
sition results are shown in case of
sample D6, as an example. Dotted
and dashed lines represent the best
fits to spectrum and D band and
G band, respectively
This, as demonstrated in Fig. 2, where the G band frequency position (ωG )

is plotted as a function of rf power and chamber pressure, hinders us from
guessing the possible effect of any variation of the growth conditions on the
film characteristics.
3.2 Operative Scaling Law Derivation Procedure
A simpler picture might derive from finding a dependence of Raman de-

scriptive parameters on a limited number (L) of dimensionless arguments
in place of the initial N (L) growth variables. The search for such argu-
ments is pursued through a semiempirical method [1], based on an extended
application of the theorem on “physically similar systems” of dimensional
analysis [2]. A two-way and trial-and-error approach (Fig. 3), involving the
repeated comparison with the experimental data, is utilised in place of the
standard one-way and necessary procedure. After replacement of dimension-
Table 2. Results of the structural and compositional analysis carried out on the
films. Samples are listed in increasing order of Q. ωG and ID /IG stand for G band
frequency position and D/G intensity ratio measured in the Raman spectra, respec-
tively. LC denotes the average size of Csp2 clusters estimated from ID /IG values [34].
The film C, H and N contents, as well as O contamination, resulting from ERDA
and RBS measurements, are reported
Film ωG ID /IG LC C content H content N content O content

# cm−1 nm % % % %
D1 1575.2 0.89 1.27 69.9 19.6 7.0 3.5
T1 1575.1 1.18 1.46 75.8 12.9 7.6 3.7
D2 1575.0 1.03 1.37 73.5 15.4 7.4 3.7
D3 1574.1 0.91 1.29 74.6 14.2 7.5 3.7
D4 1573.6 0.96 1.32 71.4 17.9 7.1 3.6
T2 1573.2 0.91 1.29 71.4 17.9 7.1 3.6
D5 1572.3 0.94 1.31 74.6 14.2 7.5 3.7
D6 1569.9 1.06 1.39 73.0 16.1 7.3 3.6
T3 1563.9 1.20 1.48 79.4 8.7 7.9 4.0
D7 1563.3 1.23 1.50 79.0 7.2 9.8 4.0
Fig. 2. Dependence of the G band frequency position (ωG ) on rf power (W ) and

chamber pressure (p). The ωG points shown are obtained by projecting the 7-
dimensional ωG hypersurface. Their confused spreading in the W, p, ωG space is
due to the simultaneous variation of the flow rates of the inlet gases
Fig. 3. Scheme of the flow chart used for deriving the scaling law Q. Stage 1: ini-
tialisation and testing of the argument Q. Stage 2: definition of a new argument as
power product of further dimensionless variable combinations Qk . M represents the
maximum number of arguments Qk entering the new formulation of Q. Stage 3: ad-
justing of the exponents αk and testing of the scaling law. The experimental data
are approximated to a suited function of Q and exponent adjusting is pursued, in
a sufficiently high iteration number I, by minimising the standard deviation σ
ate Raman descriptive parameters with appropriate dimensionless Raman

arguments, arbitrary dimensionless combinations (Q1 , Q2 , ..., QL ) of the
growth parameters (G1 , G2 , ..., GN ) are generated on the basis of physical
considerations, by the aid of dimensional analysis.
The operative procedure [35] followed comprises three stages:
Stage 1. A dimensionless argument Q1 is firstly generated (argument Q
initialisation). Assuming that the Raman argument R depends on Q, the
experimental data are approximated to a suited function of Q. If the stan-
dard deviation σ does not exceed the prior-established maximum allowed
value σmax (argument Q testing), combination Q1 represents the scaling law
searched for.
Stage 2. If not, a new dimensionless argument Q2 is generated (new ar-
gument Q definition), and R is assumed to depend on Q1 and Q2 through
the power product Q1 Qα 2 (exponent initialisation).
Fig. 4. Progressive process picture simplification achieved during the various stages
of the scaling law derivation. The dependence of the dimensionless Raman variables
α β
Ω (a–c) and ID /IG (d–f ) on arguments Q1 , Q1 Qα 2 and Q1 Q2 Q3 (where Q1 =
−1 −1 −1 −1
W · p · Φrg , Q2 = ΦAr · Φ and Q3 = ΦN2 · ΦH2 ). The numerical value, −2/3, of
exponent α is empirically determined by separately minimising the data spread in
plots (b) and (e). The numerical value, −3, of exponent β is attained by analogous
procedure from plots (c) and (f ). As the scaling law derivation progresses, Ω and
ID /IG points progressively order, so they finally show clear trends. The dashed
lines represent the best fits to the (c) Ω and (f ) ID /IG data relative to samples of
series D
Stage 3. A new function of Q is chosen for approximating the experimental

data. The value of the exponent α is optimised (exponent adjusting) by min-
imising the corresponding standard deviation σ(α). If σ(α) ≤ σmax (scaling
law testing), the procedure is stopped. The combination so attained repre-
sents the scaling law for the growth parameters. Otherwise, stages 2 and 3
β
are repeated. A further argument Q3 is generated and the form Q1 Qα 2 Q3
hypothesised for the scaling law. If σ(β) ≤ σmax , the procedure is stopped
and the effective dependence of the argument R on the growth parameters
involved is determined. The same procedure as for R has to be followed for
any other Raman argument R and/or for the film fractional content of any
composing element ξ (with k ξk = 1). However, in principle, the introduc-
tion of different arguments (Q1 , Q2 , ...) is necessary and different exponents
(α , β , ...) are, as a consequence, attained.
3.3 Choice of Q-arguments
According to the procedure of Sect. 3.2, the G band frequency position is

first replaced with the dimensionless argument Ω = (ωsp2 − ωG )/ωsp2 (where
ωsp2 = 1575 cm−1 is the value, ωG tends to in disordered CN-alloys with low C
contents [31]) and the D/G intensity ratio (ID /IG ) is considered in place of the
average size of Csp2 clusters (LC ). Suitable dimensionless combinations of the
deposition variables are then generated on the basis of the current assessments
on the synthesis of C-based films [36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46].
It is worth noting that the same procedure as for Ω is independently
followed for ID /IG , as well as for the fractional film N-content (ξN = 0.070 −
−0.098). However, since both Raman arguments reflect the features of C-
nanostructures, that, in turn, are strongly influenced by the N-incorporation
level, the same arguments are tentatively considered and successfully tested
by comparison with the experimental Ω, ID /IG and ξN data.
The first argument introduced,
Q1 = W · p−1 · Φ−1
rg
(where Φrg = ΦN2 + ΦH2 stands for the reactive gas flow rate), ideally relies
on impact ion energy for depositions operated in reactive atmosphere [24]. It
is chosen with the purpose of taking into account the significant role played
by impact ion energy in determining the film characteristics [37, 39, 40, 46]. It
has been demonstrated [24] that, if the films are prepared by changing solely
W , p and Φrg , the choice of Q1 as a scaling law for the growth parameters
allows stopping the procedure at stage 1. In the presence of a larger number
of independently varying deposition variables instead, Q1 introduces only a
partial simplification (Fig. 4a,d), calling for the successive stage running.
Hence, a second argument,
Q2 = ΦAr · Φ−1
(where Φ = Φrg + ΦHe + ΦAr denotes the flow rate of all inlet gases), is
generated. The introduction of Q1 is intended to take into account the strong
influence exerted by the changes of Ar gas fraction on the deposition rate
and C-atom nanostructure formation [47, 48, 49, 50]. The comparison with
the experimental data in stage 3 (Fig. 4b,e) shows that the combination
W · p−1 · Φ−1
rg · (ΦAr · Φ
−1 −2/3
) remarkably reduces the spread of both Ω and
ID /IG points.
Trying to further simplify the process picture, stages 2 and 3 are repeated.
A third argument,
Q3 = ΦN2 · Φ−1
H2 ,
ideally taking into account the antagonist action carried out by N2 against H2
in surface reactions involving growth precursor species [51, 52, 53], is finally
generated.
3.4 The Scaling Law for the Growth Conditions
At the end of the above described procedure, the scaling law,

−2/3
Q = W · p−1 · Φ−1
rg · ΦAr · Φ2/3 · Φ−3 3
N2 · ΦH2 (1)
(with rf power, chamber pressure and gas flow rates expressed in W, mTorr
and sccm, respectively), is derived. This law is able to effectively account
for the variations of the fractional film N-content and of considered Raman
arguments with the growth conditions.
At this stage, (1) actually accounts for all the changes of ξN , Ω (Fig. 4c)
and ID /IG (Fig. 4f) produced by the variations of W , p, ΦAr , Φ, ΦN2 and ΦH2
in the ranges 220−−300 W, 20−−38 mTorr, 30−−70 sccm, 107−−127 sccm,
4−−20 sccm and 3−−7 sccm (under which the films of series D are prepared).
In order to further test the effectiveness of Q, a new series of samples is
prepared. The deposition parameters of these films (series T) are chosen so
as to obtain values of Q falling within the range investigated (even if a wider
range is actually considered for some of the growth variables). The ξN , Ω and
ID /IG points relative to additional samples line up along the curves drawn
by points relative to films of series D, confirming the validity of the scaling
law derived. This is demonstrated in Fig. 5 for the film N-content (xN ) and in
Figs. 6 and 7 for the dimensionate Raman variables ωG and LC . Argument Q,
therefore, finally turns out to be able to account for all the changes of xN ,
ωG and LC produced by the variations of W , p, ΦAr , Φ, ΦN2 and ΦH2 in
the ranges 180 − −400 W, 20 − −38 mTorr, 30 − −70 sccm, 105 − −127 sccm,
4 − −20 sccm and 2 − −7 sccm.
3.5 Resulting Simplification of the Process Picture
As demonstrated in Fig. 4, a progressive process picture simplification is

achieved once the problem is reformulated in terms of Q-arguments. This
result is of crucial importance for the achievement of a well-controlled depo-
sition process.
First, ξN , Ω and ID /IG points, initially covering 7D hypersurfaces as an
effect of the variation of W , p, ΦAr , Φ, ΦN2 and ΦH2 , now line up along
2D curves. The reason is that, as shown, ξN , Ω and ID /IG depend on the
growth parameters through a single dimensionless combination of them. In
second place, by fitting such curves to proper functions, the dependence of
ξN , Ω and ID /IG (i.e., of xN , ωG and LC ) on the deposition variables can be
explicitly determined. The choice of the functions approximating the physical
laws governing the process is, to some extent, arbitrary. Nevertheless, the film
N-content is satisfactorily fitted to
xN (%) 7.26 + 0.85 exp[((Q/0.8)0.1 )].

Fig. 5. Dependence of film N content (xN ) on the dimensionless argument Q. Open

and filled symbols refer to a-CN:H films utilised for deriving (series D) and testing
(series T) the scaling law, respectively. The dashed line represents the physical
approximant, xN 7.26 + 0.85 exp[((Q/0.8)0.1 )], to the data
The frequency position of the G band and average size of Csp2 clusters of
the films under study are well reproduced by the analytical functions
ωG (cm−1 ) 1565.8 − 3.0Q + 9.5 exp[−(Q/0.8)8]

and
LC (nm) 1.49 − 7.59Q2 exp(−5.49Q2),
whose form, empirically demonstrated by fitting the experimental data, is

theoretically inferred from extremely simplified models [54] that account for
the influence of the microscopic growth variables on the deposition process.
Finally, if the growth parameters are scaled within the ranges above indicated,
according to (1), the N-incorporation level into the film and the features of
the resulting C-nanostructures do not correspondingly undergo significant
changes. In other words, xN , ωG and LC are invariant for all the simultaneous
variations of W , p, ΦAr , Φ, ΦN2 and ΦH2 leaving Q unchanged. This further
implies that the same xN , ωG and LC can be obtained even starting from
different configurations, provided that the deposition parameters selected
give the same value of Q.
Once a clear correlation is established between variations of the growth
conditions and corresponding changes of the film physical properties, the
final issue of sample preparation can be foreseen and, besides, the a-CN:H
characteristics can be tailored by properly tuning the deposition parameters.
Fig. 6. Dependence of the G band frequency position (ωG ) on the dimensionless

argument Q. Symbols are the same as in Fig. 5. The dashed line represents the
physical approximant, ωG 1565.8 − 3.0Q + 9.5 exp[−(Q/0.8)8 ], to the data. Note
that ωG data are the same as shown in Fig. 2. The comparison between the two plots
clearly demonstrates the simplification attained after the problem reformulation in
terms of argument Q
Fig. 7. Dependence of the Csp2 cluster average size (LC ) on the dimensionless
argument Q. Symbols are the same as in Fig. 5. The dashed line represents the
physical approximant, LC 1.49 − 7.59Q2 exp(−5.49Q2 ), to the data
4 Discussion
4.1 Reason for the Scaling Law Existence
The presented results clearly show that, in the films considered, stoichiome-
try and related structural properties are determined by the six independent
deposition parameters through a single combination of them. The existence
Fig. 8. Average film growth rate as a function of the dimensionless argument Q.

Symbols are the same as in Fig. 5; a line is drawn as a visual help
of such an effective combination suggests the possibility that, upon variation

of the growth conditions, a mutual compensation takes place between the
effects generated on film composition and C-nanostructures by the physical
agents (impact ion energy of precursor species, influence of sputtering gas
and antagonism of H2 and N2 in surface reactions), which arguments Q1 , Q2
and Q3 are related to. For instance, the simultaneous increase of rf power and
flow rate of sputtering gas will reflect onto higher impact energy of the pre-
cursor species and improved sputtering efficiency. The former factor favours
C-enrichment of the growing layer, while, contrarily, the latter hinders C-
deposit due to the etching process carried out by Ar-atoms. Analogously,
raising the nitrogen flow rate, while simultaneously decreasing chamber pres-
sure, will cause a N-incorporation enhancement and, again, an impact ion
energy increase. The former agent promotes the formation of nucleation cen-
tres for Csp2 clusters, while the latter, conversely, favours the development
of tetrahedrally bonded C-phase. Therefore, in both the cases, as a balance
of the opposite effects generated by the variation of the microscopic variables
involved into the growth process, the stoichiometry and the structural and
bonding properties of the deposited film may finally undergo no appreciable
change.
The cited examples allow us to qualitatively understand the reason for
the existence of scaling law (1) and the related invariance of the film charac-
teristics. Below, an alternative path for deriving the scaling law is proposed,
based on the formulation of a simple theoretical model of the growth process.
In this frame, the changes undergone by the microscopic variables internally
ruling the film deposition are correlated with the variations externally carried
out on the growth parameters, so as the scaling law existence is semiquanti-
tatively explained.
4.2 Modelling of the Growth Process
It is important to remember that an approximate, rather than an exact solu-

tion to the complex problem of film deposition by reactive sputtering is here
searched for. This concept should guide the reader in the present subsection,
where an extremely simplified model is proposed.
An exhaustive discussion about the mechanisms involved in a-C-based
film growth can be found in [46]. Here, far from regarding all the variables
actually governing the film deposition, the discussion is limited to those pa-
rameters whose variation (Table 1) is responsible for the property changes
observed in the samples under study (Table 2).
As commonly established [36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46], impact
ion energy and amount of C-species trapped into the growing layer have a
relevant influence in the synthesis of a-C films. Since in a-CN:H preparation
the C-deposit is controlled by the irreversible surface reactions, the average
energy with which the N- and H-containing precursor species (CHx , C2 Hx ,
CNx , C2 Nx , Hx CN, . . . [55, 56, 57, 58]) impinge onto the substrate does enter
the problem. On the basis of these assessments, it has been demonstrated [24]
that the dimensionless combination of the deposition variables, which are able
to account for the modifications of the characteristics, featuring sputtered a-
CN:H films prepared by varying only W , p and Φrg , can be written as a
power product of impact energy of radicals containing reactive species (Ei,r )
and deposition flux towards the substrate (FC→s ),
Q = (Ei,r )a · (FC→s )c , (2)
with a and c exponents to be empirically determined. Having elementarily

modelled FC→s under proper assumptions [24], combination (2) simplifies to
W · p−1 · Φ−1
rg , i.e., to argument Q1 . However, in present case, because of the
variation of a larger number of growth parameters, FC→s requires a different
and more sophisticated modelling.
As known, in addition to impact ion energy, the structural and bond-
ing properties of a-C based materials largely depend on the deposition
rate [48, 49, 50], which, in turn, in film preparation by sputter techniques,
is strongly influenced by the sputtering gas action. The sputtering gas exerts
a twofold action, resulting in simultaneous competing deposition and etching
processes [59, 60], which alternatingly prevail depending on the changes made
in the growth conditions. Such an alternation is evidenced, as in present case
(Fig. 8), by a nonmonotonic variation of the film growth rate. The reason
for this competition is that, as the sputtering gas fraction increases, on one
hand, sputtering efficiency improves and C-flux towards the plasma enhances.
Meanwhile, on the other hand, high-energy Ar-atoms etch the growing film
layer causing a diminishing deposition rate.
Then, nitrogen and hydrogen bring about a competitive action in sur-
face reactions [51, 52, 53]. The antagonism between the two reactive gases
is due to the different sticking coefficients of N- and H-containing precur-

sor species. Radicals having energy enough penetrate into the growing layer,
while the remaining ones stick to its surface and form an adsorbate layer.
Nitrogenated radicals possess, on average, lower impact energies, i.e., higher
sticking coefficients, sn . They are, hence, generally adsorbed onto the growing
layer [38]. Contrarily, more energetic hydrogenated radicals (to which lower
sticking coefficients, sh , pertain) penetrate into the growing layer with con-
sequent densification [46]. This reflects onto the local C-bonding and film
stoichiometry. Since terminal Cspx Nspy groups act as nucleation centres for
Csp2 phase [31, 33], the adsorption of nitrogen promotes the film clusteri-
sation degree enhancement. The penetration of hydrogen, instead, favours
the formation of hydrocarbon groups Csp3 Hx [31, 33, 61]. In addition, since
hydrogen is preferentially released in ion-induced desorption (dehydrogena-
tion) [46], H-adsorption promotes the film C-enrichment. N-incorporation,
instead, progresses at expense of the film C-content (see [61] in this book at
p. 221). Therefore, any variation of the ratio sn /sh results in changes of film
stoichiometry, as well as of the nature of nanostructures formed by C-atoms.
These are the reasons why a comprehensive model of FC→s , besides to
the role of the permanence time (trg ) of reactive gases into the chamber [24],
should primarily take into account the action of the fluxes of Ar-atoms to-
wards target (FAr→t ) and substrate (FAr→s ), as well as of the flux of reactive
gases towards the substrate (Frg→s ). The effects should be, then, further
considered of the antagonism between N2 and H2 , ruled by the ratio (sn,h )
between the sticking coefficients of nitrogenated and hydrogenated radicals.
FAr→t feeds the C-flux towards the plasma, favouring C-deposit. Contrar-
ily, being responsible for the simultaneous etching process, FAr→s hinders
deposition. Acting on the precursor species formation, Frg→s controls C-
incorporation into the growing layer. Finally, sn,h decides the type of growth
Table 3. Proportionality relationships between microscopic (internal) growth vari-

ables and measurable (external) deposition parameters. Ei,r denotes the average
impact energy of radicals containing reactive species. FAr→s and FAr→t respectively
indicate fluxes of Ar atoms towards substrate and target. Frg→s stands for flux of
reactive gases towards the substrate. Finally, trg denotes the permanence time of
reactive gases into the deposition chamber, while sn,h indicates the ratio, sn /sh ,
between sticking coefficients of nitrogenated to hydrogenated precursor species
Internal variable Proportionality with external parameters From refs.

Ei,r W 1/2 · p−1 · Φ−1
rg · Φ [36], [41], [46]
FAr→t W 3/4 · p1/4 · ΦAr · Φ−1 [36], [46], [62]
FAr→s W 1/4 · p3/4 · ΦAr · Φ−1 [54]
Frg→s W 1/4 · p3/4 · Φrg · Φ−1 [54]
trg p · Φ−1 [24]
sn,h ΦN2 · Φ−1
H2 [54]
precursors incorporated. Accordingly, the flux of precursor species reaching

and sticking to the substrate is here modelled as:
FC→s ∝ (FAr→t )b · (FAr→s )−d · (Frg→s )f · (trg )h · (sn,h )−l , (3)
with b, d, f , h and l positive exponents to be empirically determined. FAr→s

and sn,h enter expression (3) with negative exponents, which ideally take
into account the hindering action, carried out by etching process and N-
incorporation, against C-deposit. Such a novel expression generalises that
previously derived [24].
4.3 Physical Meaning of the Scaling Law
As accurately discussed in [54], in a-CN:H preparation by rf diode systems,

under the deposition conditions of Table 1, the above-considered microscopic
process variables are related to the macroscopic growth parameters through
the proportionality relationships reported in Table 3. Replacing such expres-
sions in (2) and (3) allows us to write Q explicitly in terms of measurable
and externally tunable parameters. The latter enter the expression of Q con-
sequently obtained with the exponents reported in Table 4. The numerical
values of these exponents can be empirically determined through the experi-
mental data fitting, so as to finally obtain:
−2/3 −3
(W 2/3 · p−4/3 · Φ−4/3
rg · Φ4/3 ) · (W 1/3 · p1/3 · ΦAr · Φ−2/3 · Φ1/3 3
rg · ΦN2 · ΦH2 ).
4/3
This product, whose first term corresponds to Ei,r , while the second being
a power of FC→s , gives combination (1).
All the deposition parameters but ΦAr enter both the two microscopic
factors singled out within the expression of Q, and, what is more, each of
them influences more than one internal variable. Hence, varying the growth
conditions causes many microscopic variables to change simultaneously and,
sometimes, the opposite effects, which are produced by these changes on film
stoichiometry and the nature of C-nanostructures, to compensate each other.
For this reason the film properties may appear insensitive to the changes of
the deposition conditions or, more generally, their modifications may be not
readily foreseeable.
4.4 The Scaling Law for the Microscopic Growth Variables
Once the exponents, with which the external deposition parameters enter
combination obtained by replacing (3) in (2), have been empirically deter-
mined, the explicit relationship between Q and the microscopic growth vari-
ables can be further derived by the aid of relationships of Table 4, so we
finally attain
Q = (Ei,r )4/3 · (FAr→t )5/6 · (FAr→s )−3/2 · (Frg→s )1/3 · (trg ) · (sn,h )−3 . (4)
Table 4. Exponents with which the deposition parameters enter (2), as explic-
itly written, in terms of measurable parameters, after replacement of relationships
reported in Table 3. W and p denote rf power and chamber pressure, while Φrg
and ΦAr indicate the flow rates of reactive gases and sputtering gas, respectively. Φ
stands for the flow rate of all the inlet gases. ΦN2 and ΦH2 denote N2 and H2 flow
rates
Deposition parameter Corresponding exponent Numerical value
W a/2 + (3b − d + f ) · c/4 1
p −a + (b − 3d + 3f + 4h) · c/4 −1
Φrg −a + f c −1
2
ΦAr c · (b − d) −
3
2
Φ a − c · (b − d + f + h)
3
ΦN2 −lc −3
ΦH2 lc 3
As hypothesised, Q depends inversely on FAr→s and sn,h . This finding indi-

rectly confirms the property of the assumptions on which the model for-
mulated is based. In particular, (4) demonstrates that, as hypothetically
supposed by expression (3), manifold factors may simultaneously influence
FC→s . Favouring and feeding the precursor species formation, longer trg and
larger Frg→s make more intense the carbonaceous flux towards the substrate.
FAr→t assists C-flux towards the plasma, thus its increase results in film
C-enrichment. Contrarily, raising FAr→t , as a result of etching enhancement,
hinders the development of the growing layer. The balance between the simul-
taneous deposition and etching processes determines the film growth rate [61].
However, all the internal deposition variables cooperatively participate in the
occurrence of these processes. This is clearly demonstrated in Fig. 8, where
the average deposition rate, as deduced from the results of joint IR and
RBS film-thickness measurements [54], is plotted as a function of Q. A non-
monotonic trend is obtained, as expected. Finally, since the incorporation
of nitrogen progresses at the expense of film C-enrichment [61], contrarily
promoted by hydrogen due to the preferential desorption from the surface
adsorbate layer, decreasing sn and/or increasing sh , i.e., lowering sn,h , ulti-
mately results in films with higher C-contents.
Combination (1) provides the rules according to which the external de-
position parameters, W , p, ΦAr , Φ, ΦN2 and ΦH2 , can be scaled, within the
ranges considered, with no significant changes in film stoichiometry and na-
ture of C-nanostructures. Correspondingly, (4) can be regarded as the scaling
law for the internal growth variables, Ei,r , FAr→t , FAr→s , Frg→s , trg and sn,h .
The existence of the latter reflects onto (and, thus, explains fully and defi-
nitely) the existence of the former.
4.5 A Figure of Merit for a-CN:H Film Deposition
It is finally important to point out another relevant outcome emerging from

the search for an approximated solution to the a-CN:H film growth process.
According to expression (2), in order to increase Q it is sufficient that only
one of the two factors, that Ei,r - or that FC→s -related, increases due to vari-
ations in the growth conditions. On one hand, as known, increasing impact
ion energy gives rise to higher sp3 /sp2 C-bonding (i.e., fourfold-/threefold-
coordinated C-atom) fractions. On the other hand, enhancing FC→s allows
the deposition of C-richer films. Since all the elements in the films exhibit a
coordination number lower than C (Nsp1 is 1- or 2-coordinated [63]; Nsp2 , if
present, is 2- or, in very few cases, 3-coordinated [64]; H is 1-coordinated), the
increase of relative C-content causes similarly the enhancement of the film
average coordination number (χav ). As evidenced by the G band downshift
(Fig. 6), films with improved characteristics are actually obtained at higher
Q values. Hence, Q can be regarded as a “figure of merit” for hydrogenated
amorphous carbon–nitride growth by sputtering.
When deposition is operated in the low Q regime, the N-incorporation
level remains roughly constant (Fig. 5). Under this condition, the slow ωG
variation (Fig. 6) reflects the C-bonding evolution produced by the forma-
tion of different hydrocarbon groups [29, 31, 33]. Porous, softer and H-richer
films are attained, with Csp2 cluster islands exhibiting progressively smaller
dimensions going towards lower Q (Fig. 7). By operating at higher Q val-
ues, instead, the enhanced terminal Cspx Nspy group formation promotes
the Csp2 matrix development and hinders H-incorporation [29, 31, 33]. Corre-
spondingly, LC gradually increases, while ωG drops as an effect of the changed
film stoichiometry and related χav increase.
5 Conclusion
The results of Raman analysis and Rutherford backscattering measurements
performed on a-CN:H films, prepared by graphite sputtering in N2 –H2 –He–Ar
atmosphere, are utilised for the derivation of a scaling law for the deposition
parameters. This is accomplished by means of a semiempirical approach,
based on an extended application of theorem on “physically similar systems”
of dimensional analysis.
The influence of the growth conditions on the film stoichiometry as
well as on the resulting features of the nanostructures formed by C-atoms
is investigated. By elementarily modelling the reactive sputtering process,
the crucial role played by the variables internally ruling film deposition is
clarified. We demonstrate that impact energy of radicals containing reac-

tive species (Ei,r ), fluxes of Ar-atoms towards substrate (FAr→s ) and tar-
get (FAr→t ), flux of reactive gases towards the substrate (Frg→s ), perma-
nence time of reactive gases into the deposition chamber (trg ) and ratio be-
tween sticking coefficients of nitrogenated to hydrogenated precursor species
(sn,h ) enter in determining the film properties through their combination
−3/2
Q = Ei,r · FAr→t · FAr→s · Frg→s · trg · s−3
4/3 5/6 1/3
n,h . Such a combination represents the
scaling law for the process on a microscopic scale. In the frame of the growth
process approximation, the changes undergone by the internal variables are
therewith correlated with the variations externally operated on rf power (W ),
chamber pressure (p) and flow rate of all the inlet gases (Φ), of sputtering-gas
(ΦAr ) and of reactive gases (Φrg = ΦN2 + ΦH2 ). As a result, the scaling law,
−2/3
Q = W · p−1 · Φ−1rg · ΦAr · Φ2/3 · Φ−3 3
N2 · ΦH2 , for the measurable and externally
tunable deposition parameters is derived.
The argument Q is then shown capable of accounting for all the changes
of the film N-content (xN ), of the frequency position of the G band (ωG )
and of the average size of Csp2 clusters (LC ), produced by the variations
of W , p, ΦAr , Φ, ΦN2 and ΦH2 in the ranges 180 − −400 W, 20 − −38 mTorr
(2.7−−5.1 Pa), 30−−70 sccm, 105−−127 sccm, 4−−20 sccm and 2−−7 sccm.
By fitting the experimental data relative to the films under study, the physical
approximants xN (%) 7.26 + 0.85 exp[((Q/0.8)0.1 )], ωG (cm−1 ) 1565.8 −
3.0Q + 9.5 exp[−(Q/0.8)8 ] and LC (nm) 1.49 − 7.59Q2 exp(−5.49Q2) are
attained. These represent the empirical rules whose knowledge is required
to be able to tailor the film characteristics through specific changes of the
deposition conditions.
Acknowledgements
We wish to thank Prof. A. Tagliaferro and Dr. G. Fanchini of the Physics

Department, Polytechnics of Torino, who kindly provided a-CN:H samples.
References
[1] G. Messina, A. Paoletti, S. Santangelo, A. Tucciarone: Microsys. Technol. 1,
23 (1994) 1, 2, 4
[2] P. W. Bridgman: Dimensional Analysis (Yale University Press, New Haven
1931) 1, 4
[3] G. Messina, S. Santangelo, A. Paoletti, A. Tucciarone: Microelectron. Eng. 34,
147 (1997) 1
[4] G. Messina, A. Paoletti, S. Santangelo, A. Tucciarone: Nuovo Cimento D 20,
1201 (1998) 1
[5] S. Santangelo, G. Messina, G. Fanchini, A. Tagliaferro: Diam. Relat. Mater.
13, 1391 (2004) 1, 2
[6] J. Bragard, G. Lebon: Transport Porous Med. 16, 253 (1994) 2
[7] K. S. Chan: Scripta Metall. Mater. 32, 235 (1995) 2

[8] B. A. Klinger, J. Marshall: Dynam. Atmos. Oceans 21, 227 (1995) 2
[9] C. J. Allègre, E. Lewin: Earth Planet. Sc. Lett. 132, 1 (1995) 2
[10] J. C. Bowman: J. Fluid Mech. 306, 167 (1996) 2
[11] N. Kleeorin: Studia Geograph. Geod. 42, 419 (1998) 2
[12] Y. Luo: Chem. Phys. Lett. 289, 350 (1998) 2
[13] G. B. West: Physica A 263, 104 (1999) 2
[14] K. Goulianos: Nucl. Phys. B 71, 368 (1999) 2
[15] D. Huber, D. Pfenniger: Astrophys. Space Sci. 276, 523 (2001) 2
[16] L. Kari: Appl. Acoust. 63, 583 (2002) 2
[17] E. van Nimwegen: Trends Genet. 19, 479 (2003) 2
[18] A. Salabat, K. Nasirzadeh: J. Mol. Liq. 349, 103 (2003) 2
[19] D. Tondeur, L. Luo: Chem. Eng. Sci. 59, 1799 (2004) 2
[20] S. Thurner, C. Windischberger, E. Moser, P. Walla, M. Barth: Physica A 326,
511 (2003) 2
[21] H. Kasajima, E. Nagano, T. Suzuki, S. G. Kim, W. T. Kim: Sci. Technol. Adv.
Mater. 4, 553 (2003) 2
[22] G. E. Glicksman, A. O. Lupulescu: J. Cryst. Growth 264, 541 (2004) 2
[23] G. Messina, S. Santangelo, G. Fanchini, A. Tagliaferro: Vacuum 67, 537 (2002)
2
[24] G. Messina, S. Santangelo, A. Tagliaferro, A. Tucciarone: J. Non-Cryst. Solids
318, 322 (2003) 2, 8, 13, 14, 15
[25] H. S. Kwok, H. S. Kim, D. H. Kim, W. P. Shen, X. W. Sun, R. F. Xiao: Appl.
Surf. Sci. 595, 109 (1997) 2
[26] P. Leceur, M. B. Korzenski, A. Ambrosini, B. Mercey, P. Camy, J. L. Doualan:
Appl. Surf. Sci. 186, 403 (2002) 2
[27] L. Brendel, A. Schindler, M. von den Driesch, D. E. Wolf: Comput. Phys.
Commun. 147, 111 (2002) 2
[28] P. M. Ossi, A. Miotello: Pulsed laser deposition of carbon films: Tailoring
structure and properties, in G. Messina, S. Santangelo (Eds.): Carbon, The
Future Material for Advanced Technology Applications, vol. 100, Topics Appl.
Phys. (Springer, Heidelberg, Berlin 2006) 2
[29] G. Messina, S. Santangelo, G. Fanchini, A. Tagliaferro, C. E. Nobili, G. Ot-
taviani, A. Paoletti, A. Tucciarone: Diam. Relat. Mater. 11, 1166 (2002) 3,
17
[30] G. Messina, A. Paoletti, S. Santangelo, A. Tagliaferro, A. Tucciarone: J. Appl.
Phys. 89, 1053 (2001) 3
[31] G. Fanchini, G. Messina, A. Paoletti, S. C. Ray, S. Santangelo, A. Tagliaferro,
A. Tucciarone: Surf. Coat. Tech. 257, 151 (2002) 3, 8, 14, 17
[32] G. Messina, S. Santangelo, G. Fanchini, A. Tagliaferro: GNSR2001: State of
Art and Future Development in Raman Spectroscopy and Related Techniques
(IOS Press, Amsterdam 2002) pp. 244–255 3
[33] G. Fanchini, A. Tagliaferro, G. Messina, S. Santangelo, A. Paoletti, A. Tuc-
ciarone: J. Appl. Phys. 91, 1155 (2002) 3, 14, 17
[34] A. C. Ferrari, J. Robertson: Phys. Rev. B 61, 14095 (2000) 5
[35] G. Messina, S. Santangelo: Proceedings of GNSR2003 Conference (Morlacchi,
Perugia 2004) pp. 114–120 6
[36] Y. Catherine: Diamond and Diamond-like Films and Coatings, Proc. of NATO-
ASI 1990 (Plenum, New York) 8, 13, 14
[37] P. J. Fallon, V. S. Veerasamy, C. A. Davis, J. Robertson, G. A. J. Amaratunga,
W. I. Milne, J. Koshinen: Phys. Rev. B 48, 4777 (1993) 8, 13
[38] J. Robertson: Diam. Relat. Mater. 3, 361 (1994) 8, 13, 14
[39] Y. Lifshitz: Diam. Relat. Mater. 5, 388 (1996) 8, 13
[40] M. Chhowalla, J. Robertson, C. W. Chen, S. R. P. Silva, C. A. Davis,
G. A. J. Amaratunga, W. I. Milne: J. Appl. Phys. 81, 139 (1997) 8, 13
[41] J. Robertson: Amorphous Carbon: State of the Art (World Scientific, Singapore
1998) p. 32 8, 13, 14
[42] D. Schneider, C. F. Meyer, H. Mai, B. Schöneich, H. Ziegele, H. J. Scheibe,
Y. Lifshitz: Diam. Relat. Mater. 7, 973 (1998) 8, 13
[43] R. G. Lacerda, F. C. Marques, F. L. F. Jr.: Diam. Relat. Mater. 8, 495 (1999)
8, 13
[44] M. P. Siegal, P. N. Provencio, D. R. Tallant, R. L. Simpson: Appl. Phys. Lett.
76, 2047 (2000) 8, 13
[45] K. Yamamoto, K. Wazumi, T. Watanabe, Y. Koga, S. Iijima: Diam. Relat.
Mater. 11, 1130 (2002) 8, 13
[46] J. Robertson: Mat. Sci. Eng. R 37, 129 (2002) 8, 13, 14
[47] F. Parmigiani, E. Kaym, H. Seki: J. Appl. Phys. 64, 3031 (1988) 8
[48] R. Kurt, R. Sanjines, A. Karimi, F. Lévy: Diam. Relat. Mater. 9, 566 (2000)
8, 13
[49] S. Santucci, L. Lozzi, L. Valentini, J. M. Kenny, A. Menelle: Diam. Relat.
Mater. 11, 1188 (2002) 8, 13
[50] J. M. Ting, H. Lee: Diam. Relat. Mater. 11, 1119 (2002) 8, 13
[51] S. R. P. Silva, B. Rafferty, G. A. J. Amaratunga, J. Schwan, D. F. Franceschini,
L. M. Brown: Diam. Relat. Mater. 5, 401 (1996) 8, 13
[52] S. R. P. Silva, J. Robertson, G. A. J. Amaratunga, B. Rafferty, L. M. Brown,
J. Schwan, D. F. Franceschini, G. Mariotto: J. Appl. Phys. 81, 2626 (1997) 8,
13
[53] C. Godet, N. M. J. Conway, J. E. Bourée, K. Bouamra, A. Grosman, C. Ortega:
J. Appl. Phys. 91, 4154 (2002) 8, 13
[54] G. Messina, S. Santangelo: Diam. Relat. Mater. 14, 1331 (2005) 10, 14, 15,
16
[55] R. Kaltofen, T. Sebald, G. Weise: Thin Solid Films 118, 308 (1997) 13
[56] W. Jacob: Thin Solid Films 326, 1 (1998) 13
[57] R. L. C. Wu, K. Miyoshi, W. C. Lanter, J. D. Wrbanek, C. A. D. Joseph: Surf.
Coat. Tech. 573, 120–121 (1999) 13
[58] N. Mutsukura, K. Akita: Diam. Relat. Mater. 9, 761 (2000) 13
[59] J. Hong, A. Granier, A. Goullet, G. Turban: Diam. Relat. Mater. 9, 573 (2000)
13
[60] B. Druz, Y. Yevtukhov, V. Novotny, I. Zaritsky, V. Kanarov, V. Polyakov,
A. Rukavishnikov: Diam. Relat. Mater. 9, 668 (2000) 13
[61] F. Freire, Jr.: Hard amorphous hydrogenated carbon films and alloys, in
G. Messina, S. Santangelo (Eds.): Carbon, The Future Material for Advanced
Technology Applications, vol. 100, Topics Appl. Phys. (Springer, Heidelberg,
Berlin 2006) 14, 16
[62] C. D. Martino, G. Fusco, G. Mina, A. Tagliaferro, L. Vanzetti, L. Calliari,

M. Anderle: Diam. Relat. Mater. 6, 559 (1997) 14
[63] S. Muhl, J. M. Mendez: Diam. Relat. Mater. 8, 1809 (1999) 17
[64] N. M. Victoria, P. Hammer, M. C. D. Santos, F. Alvarez: Phys. Rev. B 61,
1083 (2000) 17
Index
a-C:H:N, 1–5, 10, 11, 13, 14, 16, 17 hydrogenated amorphous carbon-
sp2 -bonded clusters, 5, 10, 18 nitrides, 1
cluster size, 5, 8, 10, 11, 18
sp3 /sp2 bonding ratio, 17 microscopic growth variables, 12–15
approximated solutions, 1, 2, 10, 11, 13 nitrogen incorporation, 10, 12–17
physical approximants, 1, 10, 11, 18

carbon coordination, 17
process approximation, 2, 13–15, 18
threefold and fourfold coordination,
process modelling, 13–15
17
Q-arguments, 1, 6, 8, 9
dimensionless arguments, 4–6, 8
Raman dimensionless arguments, 6–8
figure of merit for growth process, 17 Raman spectroscopy, 1–3, 17
film stoichiometry, 3, 11, 12, 14–17 reactive sputtering, 2, 13
G band, 3, 4, 8, 10, 11, 17, 18 scaling laws, 1, 2, 15, 17, 18

derivation method, 1, 4, 7–9, 17
hydrogen desorption, 14, 16 validity range, 9
A Spectroscopic Approach to Carbon
Materials for Energy Storage
Giuseppe Zerbi1,2 , Matteo Tommasini1,2 , Andrea Centrone1,2 ,

Luigi Brambilla1,2 , and Chiara Castiglioni1,2
1
Dipartimento di Chimica, Materiali e Ingegneria Chimica “G. Natta”,
Politecnico di Milano, Piazza Leonardo da Vinci 32, I-20133 Milano, Italy,
and INSTM, Unità di Ricerca di Milano
2
Centro d’Eccellenza “Nano-Engineered MAterials and Surfaces” (NEMAS),
Politecnico di Milano, Via Ponzio 34/3, I-20133 Milano, Italy
giuseppe.zerbi@polimi.it
Abstract. In this Chapter we present the contribution offered by vibrational spec-

troscopy in the characterization of carbon-based materials. The necessary concep-
tual tools will be introduced in Sect. 2 and Sect. 3; the basic structural and dynam-
ical properties of the different crystalline or ordered aggregation forms of carbon
will be described. With such a background we pave the way to the spectroscopic
characterization (Raman and infrared) of disordered carbon materials, such as those
recently proposed for hydrogen storage. The application of the concepts and data
described to the field of new carbon materials for energy storage is presented.
1 Introduction
Carbon-based materials are at present of great technological relevance as pos-

sible substances that are capable of storing hydrogen as new energy vectors
in the automotive field. In the past few years the interest in carbon-based
materials for energy storage has reached very high levels as a consequence
of the news reported by some laboratories [1, 2, 3] that such materials could
absorb, even at room temperature, astonishingly large amounts of gaseous
hydrogen. If these reports were true the “energy problem” which plagues hu-
manity could have been easily and happily solved. Unfortunately, the data on
hydrogen absorption reported were not reproduced by other authors and were
later shown to be unacceptable [4, 5, 6]. Moreover, many other papers report
little or negligible hydrogen absorption in different carbon nanostructures
(for a general discussion and a list of references see [7]).
We became concerned with the failure of the many attempts to squeeze,
in a reversible way, hydrogen inside some kind of carbonaceous materials
and noticed that most of the experiments were not based on some sort of
rationale in the choice of the materials. Indeed, the physics of the absorption
necessarily must be strictly related to the yet unknown structure, at the
molecular level, of such carbon-based materials. On the other hand, it is well
known that the knowledge, at the molecular level, of amorphous carbon-based
24 Giuseppe Zerbi et al.
materials cannot be easily obtained because they are generally insoluble and
noncrystalline solids.
The above observations have inspired a group of laboratories to take a
“molecular approach” to the problem of hydrogen absorption [8]. The project
focussed on producing carbon-based materials of suitable structure by syn-
thesis in controlled situations [8, 9], followed by a molecular characteriza-
tion [10] and the development of first-principle calculations that may guide
the interpretation of the experimental data [11, 12, 13, 14]. Two classes of
polyaromatic compounds were considered (for a general discussion see [15]),
namely 1. polyaromatic dendrimers and 2. carbon-based materials (PAH)
obtained from a controlled graphitization [9] of structurally well-character-
ized and monodisperse polyaromatic precursors. Details on these studies are
reported in Sects. 4 and 5 of this Chapter.
Again, such research turned out to be unsuccessful since the amount of
hydrogen absorbed is still hopelessly small and sometimes negligible. This
poses the basic scientific demand (of general interest) to relate the molecular
nature and the electronic properties of the hydrogen molecule to the fact that
it can indeed wander inside a carbon cage, but it does not find the conditions
to stick inside such a cage.
We believe that the approach to the problem must systematically proceed
first with the understanding of both the ordered and disordered structures
the carbon atoms are able to generate. The structures must be identified by
means of experimental physical techniques, which are obviously many when
the carbon materials are crystalline and ordered, while they reduce to a few
techniques, sometimes vague, when the materials become disordered. First-
principle calculations may provide the clue, or a hint, to the understanding of
such structural data. This line of thought inspires the discussion that follows
in this Chapter.
Our approach aims at presenting how molecular and lattice dynamics
(and, by consequence, infrared and Raman spectroscopy) can play a role in
the field of carbon-based materials. We wish to present the state of the art
and the recent development of molecular dynamics and vibrational infrared
and Raman spectroscopy (theory and experiments) (i) as a probe of ordered
and disordered carbon-based materials and (ii) as a probe of hydrogen pushed
into the interatomic spaces in such carbon-based materials.
1.1 Carbon-Based Materials
The definition of “carbon-based materials” is indeed quite vague, and we shall

try to draw some boundaries dictated by the kind of analysis we are going to
discuss. In the formation of any kind of covalently bonded network of carbon
atoms, each carbon may act in one of its three states of hybridization, namely
Csp3 , tetrahedral, Csp2 , plane trigonal and Csp, linear. The duty of organic
chemistry is to strictly control the reaction process directing carbon atoms
to preassigned positions in a given molecule under construction and with a
A Spectroscopic Approach to Carbon Materials for Energy Storage 25
preassigned hybridization state (hence in a given valence state). The molecu-

lar architecture and the derived physical and chemical properties follow from
the result of the reaction process.
Recently physics has moved into the field of carbon and proposed very
interesting and potentially useful new ways to obtain networks of carbon
atoms generated by collisions of carbon atoms subjected to welldefined ki-
netic and thermodynamic conditions [16]. Various kinds of diamond-like or
graphite-like materials have been produced which show very interesting new
mechanical and electrical properties. The precise control of the physical con-
ditions have also allowed for the generation of structurally ordered carbon
networks like fullerenes [17] and carbon nanotubes [18], which are at present
matter of strong interest in science and technology.
While a whole set of physical techniques of characterization can be ex-
ploited in the case of simple organic materials practically in any physical
phase, the chore of structural characterization, at the so-called molecular
level, of structurally highly disordered, insoluble and sometimes unstable sys-
tems becomes an exciting challenge to physics and chemistry. We review and
discuss some of the aspects of the role of molecular dynamics and vibra-
tional infrared and Raman spectroscopy and its achievements in the study of
disordered carbon-based materials.
A warning is necessary at the beginning of this Chapter. Our whole ex-
perience in molecular and lattice dynamics and spectroscopy has developed
based on a molecular approach [19, 20]. In this approach, the vibrating ob-
jects are made with atoms of a known mass, connected by valence bonds
in a given chemically acceptable architecture and subjected to an intra- or
intermolecular potential which can be conveniently described by a “valence
force field” originating from electrons building up the chemical bonds. Inter-
molecular interactions will generally be represented by some kind of Van der
Waals forces [21, 22, 23]. In other words, we will stick as much as possible to
a molecular modelling of both our ordered or disordered systems made up by
carbon atoms.
Our chemical background allows us, when nature makes it possible, to
look at the systems as infinite lattices or as finite systems. Consideration of
finite systems opens the window to the very rich class of oligomeric molecules
which have been made (or can be made on purpose) by chemists as models
of larger, but chemically extremely similar systems. Consideration of the
dynamics of finite (even if very large) molecules [24] does not allow us to
overlook the existence of surface effects (or end effects), which may affect, to
different extents, the electronic properties, the dynamics and the spectra of
the systems studied. As a typical case of identification of end effects we quote
the case of the spectra of the oligomers of pyrrole up to polypyrrole [25].
The examples presented in this Chapter witness the relevance of oligomers
in such a kind of studies. The striking case is that of the use of infrared
spectroscopy on a series of polymethylene chains, CH3 –(CH2 )n –CH3 , which
has allowed plotting, with great accuracy, all the phonon dispersion curves
of polyethylene considered as a one-dimensional lattice [26, 27]. This goal has
never been reached by any, even modern, neutron scattering experiments.
2 Prototypical Lattices in Three, Two

and One Dimensions
In the introductory paragraph we referred to the three states of hybridiza-
tion of carbon atoms as basic units that can generate ordered or disordered
networks uniquely made up by carbon atoms. Before proceeding we feel it
necessary to introduce in the discussion some additional “basic lattices” that
will be encountered later in our discussion. Let us consider the extreme struc-
tures of various ordered organizations of carbon atoms all in a unique state
of hybridisation.
2.1 Diamond
The ordered linking of Csp3 generates diamond, which has been treated by
hundreds of papers from solid state physics to applied industrial technol-
ogy [28]. Diamond is an indirect band-gap material with an energy gap of
5.49 eV. In diamond Csp3 atoms are joined by single σ bonds. The three-
dimensional (3D) lattice of diamond consists of two interpenetrating face-
centered cubic lattices, displaced one from the other by 1/4 of the principal
diagonal of the cube. Each atom of one lattice is at the centre of a tetrahe-
dron formed by its four neighbours of the other lattice. The crystallographic
translational unit cell contains eight atoms and displays the whole cubic sym-
metry Oh . The smallest translational unit cell (primitive cell) contains only
two atoms. These two atoms generate six branches in the phonon dispersion
relation [29]. The phonons that propagate along the symmetry directions of
the first Brillouin zone belong to well-defined symmetry species and determine
the number of phonon branches experimentally observed (by neutron scatter-
ing) along a chosen symmetry direction. At Γ the three optical branches are
triply degenerate (F2g symmetry), while the triply degenerate translational
mode at ν = 0 is of F2u symmetry.
For our kind of discussion most of the dynamic properties of diamond are
already available [30, 31]. Accurate phonon dispersion curves are available,
and they have been nicely accounted for by a six-parameter valence force field
refined by least squares fitting over an extremely large number of phonon
frequencies derived from neutron scattering data [29]. As seen in Table 1, the
statistical dispersions of the calculated valence force constants are generally
very small, thus supporting the physical meaning of the calculated numbers.
The information we feel relevant for this Chapter derives from the values
of the valence force constants reported in Table 1:
Table 1. Six-parameter valence force field and relative uncertainties for diamond
crystal. Units: stretching and stretching–stretching force constants are expressed in
(mdynes/Å), bending and bending–bending in (mdyne Å/rad2 ), stretching–bending
in (mdyne/rad)
KR = 3.831 ± 0.023
HΛ = 0.872 ± 0.121
FR = 0.164 ± 0.017
FRΛ = 0.392 ± 0.012
FΛ = −0.015 ± 0.010
FΛ = 0.173 ± 0.043
(i) The diagonal and off-diagonal valence force constants are fully compara-
ble with those derived from similar studies of n-alkane chains and poly-
ethylene [26, 27], thus showing that the ‘chemistry’ of the C–C bonds is
very similar (as expected) in the two classes of materials.
(ii) The intramolecular potential for σ-bonded C–C systems is very localized
mostly to its first neighbour, and it feels only a slight perturbation by
the second neighbour (FΛ = 0.173 ± 0.043 mdyne Å/rad2 ). This is
again in agreement with the finding of molecular dynamics of n-alkane
chains [26, 27], of the lattice dynamics of polyethylene [32] and of similar
simple polyolefins [33].
The highest optical frequency of F2g symmetry is observed in the Raman
spectrum at 1338 cm−1 . For a perfect crystal no other one-phonon transitions
are predicted to be observable in the optical spectra. However, as soon as the
crystal contains defects, the translational symmetry is broken and many of
the lattice modes gain some dipole moment change, thus giving rise to weak
absorption bands in the infrared. One-phonon and two-phonon densities of
states are available both from calculations and experiments [29].
Oligomeric molecules for diamond were not known for quite a while. The
first cage molecule, adamantane [34], was discovered in 1933 and only recently
has research on nanostructured materials revealed that a whole class of oligo-
adamantanes or diamondoids can be obtained from petroleum [35, 36, 37].
Their physical properties are under intense study, and their molecular dy-
namics and spectra are in too early a stage to teach us something relevant
for the understanding of ‘end effects’ of finite diamond lattices. Contrary to
what we will discuss below for Csp2 and Csp systems, we expect that the
dynamical perturbations by the end effects for diamond will die off quickly
moving into the crystal because only the nearest neighbours are involved in
the vibrations, as mentioned above.
2.2 Graphite
Next we consider Csp2 , which can form a two-dimensional (2D) lattice of

graphite or a one-dimensional (1D) lattice as in the case of polyacetylene. In
such a state of hybridization each carbon atom must be linked to its neigh-
bours with three σ bonds in a trigonal planar geometry and with one π bond
located at either side of the plane of the σ bonds. Because of the existence of
a large set of interacting 2pz orbitals, a whole new physics has been developed
in order to account for the delocalization in 2D or 1D lattice of such electrons
and for the variety of new physical processes that are generated. The first
consequence of such delocalization is that the intramolecular interactions of
π electrons extend over large distances as revealed in many experiments and
as indicated by first-principle calculations.
The prototypical case of a Csp2 lattice in 2D is graphite, which at present
is matter of active theoretical and experimental studies. Graphite is a layered
material consisting of superimposed sheets made up by carbon atoms in sp2
hybridization. Each sheet has a honeycomb structure that can be viewed as
the limiting case of a polycondensed aromatic system of infinite size. The in-
termolecular interactions between layers are of Van der Waals type, whereas
the bonds within each layer are covalent σ and π bonds. The π component
gives rise to an extended 2D polyconjugated system with a peculiar semi-
metallic electronic structure [38, 39, 40]. The Fermi surface degenerates to a
point in reciprocal space, named K. This is the corner of the hexagon that
defines the first Brillouin zone (six equivalent K points are therefore present).
The conduction properties of graphite and related compounds such as nano-
tubes [18] are given by the Bloch electrons in the vicinities of the K point.
Similarly to what happens in polyacetylene and its oligomers [41] (which will
be discussed below), a strong electron–phonon coupling is present in graphite
also [42]. This is responsible for the Kohn anomaly recently demonstrated in
graphite [43]. The coupling of π → π ∗ electronic transitions with phonons
allows us to successfully use resonant Raman spectroscopy to characterize
graphite and the related compounds. In particular, for graphite a peculiar
phonon with wavevector K (named A phonon) is normally Raman-silent
since it does not satisfy the q = 0 selection rule. Because of the strong elec-
tron–phonon coupling of the A phonon [42, 43], as soon as the q = 0 selection
rule is removed (by the presence of defects and/or confinement such in mi-
crocrystalline graphite [44, 45]) the A phonon becomes resonantly activated
in the Raman [13, 46, 47]. This gives rise to the so-called D band of defective
graphite and related compounds, observed in the Raman in the 1300 cm−1
frequency range [41]. Moreover, the frequency of the D band shows a disper-
sion with respect to the energy of the laser in resonance condition [13, 44, 45].
It has been recently proposed [42, 43] that the coupling of π → π ∗ electronic
transitions with the A phonons (with a wavevector close to the K point)
is ultimately responsible of the observed D peak behaviour. The interested
reader can found additional information about the phonon structure and the
electron–phonon coupling in graphite and nanotubes in the Chapter by Cas-
tiglioni et al. (pp. 381–402).
2.3 Polyacetylene
The simplest 1D lattice made up by a linear sequence of Csp2 is polyacety-

lene (PA), (−CH=CH−)n , which appeared in modern soft matter physics and
chemistry only relatively recently, and in the year 2000 generated a Nobel
Prize in chemistry to Shirakawa, McDiarmid and Heeger. For sake of simplic-
ity, with the acronym PA we refer here to the most stable, and most known,
form of PA, namely trans-polyacetylene as obtained from thermal isomeriza-
tion of cis-polyacetylene (which is the form chemistry provides first during the
synthesis) [48,49,50]. The molecular structure of trans-polyacetylene has been
the subject of many works. The best oriented samples were obtained by the
Durham route with a very high draw ratio [51], which has also been improved
by a modification of the original preparation of the Shirakawa polymer which
reached a draw ratio of 6.5 and which shows a very high anisotropic electrical
conductivity when doped (up to 105 Ω−1 cm−1 , almost like copper) [52].
Polyacetylene, in a first approximation, can be considered a 1D lattice
with a band gap Eg ≈ 1.4 eV consisting of a translational unit (−CH=CH−)
that repeats itself in a linear zig-zag polymer chain. The existence of a band
gap tells us that PA is not a metal, but is instead a semiconductor. Accepting
the language of chemistry, PA is a long polyene chain with conjugated π elec-
trons, while physics describes PA as a dimerized Peierls-distorted sequence
of CH units.
Because of strong electron delocalization, short-chain polyene molecules
(including the large class of carotenoids [53]) become extremely interesting
models for understanding physical properties of PA. Figure 1 shows the vi-
bronic spectra of a series of conjugated oligoenes where (in a one-electron
picture) the strong dependence of the HOMO–LUMO energy gap with the
number of conjugated double bonds appears clearly. By increasing the num-
ber of conjugated double bonds, the Kuhn’s limit of a metallic system is not
reached, but the energy gap opens at a certain length of the chain because
Peierls distortion sets in [54].
As discussed in several review articles, in these polyconjugated 1D systems
the motions of electrons are strongly coupled with the nuclear motions. It has
been shown that for specific normal modes (which can be found by lattice
dynamics calculations) a strong electron–phonon coupling takes place [56],
which strongly enhances the Raman cross section, thus generating an ex-
tremely strong and characteristic scattering line in the frequency range of
the skeletal stretching, near 1500 cm−1 [56]. The molecular approach taken
Fig. 1. UV–Vis absorption spectra of oligoenes of increasing chain lengths (n =

3, 5, 6, 8, 10 represents the number of double C=C bonds). Horizontal axis is
wavelengths (nm), left vertical axis is the absorbance (10−3 units) for n = 3, 5, 6;
right vertical axis is the absorbance for n = 8, 10. Taken from [55]
in this work has brought to the interpretation of the electron–phonon coupling

and Raman enhancement in terms of the so-called R− mode1 [57], which can
be related to the approach presented by the school of Horovitz [58, 59].
The electron–phonon coupling depends on the HOMO–LUMO energy
gap, hence on the effective conjugation length (ECL) of the molecule. In
the dynamical treatment in terms of a molecular model, such a mode is indi-
cated by R− . It has been shown experimentally and accounted for by theory
and first principle calculations that the frequency and the intensity of the
R− mode is conjugation length-dependent. Since a real sample of PA from
any reaction process consists of a mixture of molecules (or chain segments)
with varying ECL, by selecting various exciting laser lines in a Raman exper-
iments it is possible to match the energy gap of various effective conjugation
lengths (resonance condition), thus enabling us to probe the “conjugational
polydispersivity” of any real sample [56].
At present, a reliable valence force field for PA is available [41]. Obviously
the use of such a force field for shorter oligomeric models has to consider that
the effective electron–phonon coupling is a function of the conjugation length,
as clearly shown experimentally and theoretically in the case of a series of
polyenals CH2 =CH–(CH=CH)n −CHO [60].
PA can be considered a prototypical system of a very large class of poly-
conjugated 1D systems (oligomers and polymers). While PA is very unsta-
1
The symbol R− replaces for typographical restrictions the symbol commonly
used in the previous literature, namely the letter “ya” of the Cyrillic alphabet.
ble, other 1D systems are chemically stable and can be suitably treated
for attaining technologically relevant properties, such as electrical conduc-
tivity, electrochromism, photochromism, electroluminescence, photoconduc-
tivity, nonlinear optical responses, etc. [49, 50, 61]. Of particular interest in
this discussion are the 1D lattices made up by aromatic rings (e.g., polypara-
phenylene (PPP) [62], polyparaphenylene vinylene (PPV) [63, 64], polythio-
phenes [65,66], rylenes [67] and, more recently, functionalized acenes [68,69]).
For a review on the vibrational properties of polyconjugated systems see [70].
At this point of our analysis we need to highlight a structural feature
that strongly determines π electron delocalization in these polyconjugated
systems. Conjugation means a distribution of π electrons along the carbon
skeleton that is made possible through the overlapping of adjacent 2pz or-
bitals, which exist in this class of molecular systems. Maximum coupling takes
place when the two adjacent 2pz orbitals are parallel and any rotation about
the single C–C bond decreases the extent of overlapping. In more technical
structural words, if ϕ is the dihedral angle of the two 2pz orbitals about
the single C-C bond, coupling is maximum at ϕ = 0◦ and is minimum at
ϕ = 90◦ . A cosine law may be guessed for the extent of overlapping as func-
tion of ϕ; the consequent electronic and dynamical properties derive from the
ϕ dependence. In turn, the value of ϕ depends on the conformation of the
system [71].
The geometrically driven coupling of 2pz orbitals allows us to classify poly-
conjugated systems as ‘flexible’ (F) and ‘rigid’ (R). Typical cases are PPP
and PPV, which are certainly F, while the classes of rylenes, linear polyacenes
and generally of ladder polyconjugated polymers and, finally, graphite and
its oligomers can be classified as R molecules. The conformational structure
of F molecules in the solid is the result of energy minimization of inter- and
intramolecular interactions, which may change dramatically when the sys-
tem is dissolved in various solvents or it is mixed in a blend with another
material. Innumerable are the calculations presented in the literature, with
increasingly sophisticated theoretical approaches, which try to evaluate the
various intramolecular contributions that give rise to the potential barrier
supervising the intramolecular conformational flexibility of F systems. The
weakness of such calculations lies in the fact that they can easily treat mole-
cules isolated in vacuo, but have a much harder time dealing with the reality
of systems which do interact in three dimensions (either as crystals or in
liquid and solution states).
For R molecules one can safely think, in a first approximation, that the
electronic properties (and the many relevant consequent physical properties)
remain constant per each given chain length; on the contrary for F systems,
even for a fixed chain length, the conformational flexibility becomes a puzzle
in the effort of rationalizing structure and properties of a given system. Bond
length alternation (BLA) loses its precise meaning, and necessarily the con-
cept of a torsional dependent effective BLA (EBLA) needs to be introduced.
In polyconjugated systems intramolecular 2pz –2pz coupling has been con-

sidered mostly for a single molecule in vacuo (as in most of the cases con-
sidered by first-principles calculations). Nowadays increasing importance is
being given to intermolecular 2pz –2pz interactions mostly in the case of R sys-
tems. Indeed self-assembling in solution and formation of liquid crystal phases
have been recently considered for flat rigid oligomeric molecules that show in-
teresting properties (e.g., intermolecular electrical conductivity), which occur
because of intermolecular 2pz –2pz interactions [72, 73, 74].
2.4 Polyynes
The last structural unit to be considered is the linear repetition of Csp units
named by chemists as polyynes, (−C≡C−)n . Not much is known on the
dynamical and spectroscopic properties of polyynes, and their chemistry has
enjoyed a recent revival of which spectroscopy will certainly profit; see, for
instance, [75, 76].
Depending on the hybridization state of the terminal carbon atoms, fi-
nite linear carbon chains can in principle consist of alternating single and
triple bonds, or consecutive double bonds (cumulenes). The first case is
typical of H−(C≡C)n −H linear structures (terminal carbons in sp state),
whereas the cumulenic case is typical of H2 Cn H2 structures (terminal car-
bons in sp2 state). The molecules of allene and acetylene are the first in the
series of cumulenes and polyynes, respectively. We have seen that in poly-
acetylene the Peierls distortion sets in, giving an alternate structure of single
and double bonds. It is likewise reasonable to suppose a similar behaviour
also for an infinite linear chain of carbon atoms. Under this regard, for the
infinite linear chain, the cumulenic structure is expected to be less stable
than the alternance of single and triple bonds (acetylenic structure). In fact,
ring-shaped clusters of carbon atoms have been theoretically studied show-
ing signatures of Peierls distortion and Kohn anomaly and softening of the
dimerization phonon [77], similarly to polyacetylene. Recently, thanks to a
new synthetic procedure [78, 79], solutions of a distribution of linear carbon
chains H−(C≡C)n −H (3 < n < 8) have become available for Raman and
SERS characterization [80]. While the Raman/SERS data for each single
and isolated chain length are not yet available, the theoretical analysis of the
experimental data (through density functional theory calculations) shows a
definite softening of the most active vibration in the Raman with respect
to the increasing length of the linear chain. This is consistent with previous
independent theoretical studies [77].
3 A Spectroscopic Approach to Disordered Carbon

Materials
In the previous section we considered carbon-based materials that possess
a highly ordered and well-defined structure. Single crystals with supposedly
perfect structure have been produced and carefully studied for diamond and
graphite. The structure of trans-polyacetylene is sufficiently well known, while
an ‘infinite’ polyyne chain has not yet been obtained. Moving into the world
of real materials, we must accept the unavoidable existence of structural small
or large defects, and the coexistence, in a given solid material, of amorphous
(or disordered) and crystalline domains; sometimes the system is fully dis-
ordered. Following our line of thought we consider it an unavoidable step
in the structural analysis of carbon-based materials to dwell on our present
knowledge of the structure of these systems. In other words, first we need
to know how to handle a few structural defects. Then one can proceed from
slightly disorded to fully amorphous systems.
3.1 1D Systems – Saturated Carbon Materials
The understanding of defects in 1D systems received a great deal of attention

during the golden age of chemistry and physics of classical polymers which,
in most of the cases, consist of σ-bonded carbon atoms in sp3 hybridization
(and this makes life easier). All conceivable theoretical and experimental tech-
niques have been applied to understand the conformational structure and the
corresponding physico-chemical and mechanical behaviour of common poly-
mers such as polyethylene, polypropylene, polytetrafluoroethylene, etc., in
their ideally perfect or defect-containing structures. The energies necessary
for generating conformational defects labelled as G (gauche), T (trans) and
of their combinations (GG, GTG, GTG , GGTGG, etc.) have been evalu-
ated, and physical tools to detect them in the various polymers have been
developed. In a similar way, chemical defects as well as stereochemical defects
have been treated. From the viewpoint of molecular and lattice dynamics of
polymers as 1D systems, a considerable effort has been made to offer vi-
brational spectroscopies (infrared and Raman) as useful tools for identifying
such defects in real polymeric materials [81, 82]. Necessarily in disordered
lattices of any dimension, the overall symmetry of the system is lost, group
theory becomes useless, phonons become very perturbed and all vibrational
transitions may become spectroscopically active. The concept of vibrational
density of states g(ω) for a disordered system becomes then essential, and nu-
merical tools were developed for the calculation of the vibrational eigenvalues
of extremely large dynamical matrices [82, 83].
As shown in Sect. 2.1, for Csp3 systems intramolecular interactions are
very localized and die off as soon as the second neighbour. This situation
makes the calculation of 1D systems with defects much simpler and faster,
since the dynamical matrices turn out to be practically band matrices with
a very few codiagonals. One can then afford huge calculations in reasonably
short computer times [84]. We remind the reader that the comparison of
the calculated g(ω) with the experimental spectrum suffers of an extremely
relevant limitation, namely g(ω) is not transition-moment-weighted in the
infrared or in the Raman, and we have no ground to predict whether a given
strong singularity or a weak feature in the calculated g(ω) may give rise
to a strong or weak transition in the vibrational spectra. Correlations with
some known model molecules, with the results from ab initio calculations
and with the studies on infrared and Raman intensities [85, 86] on simple
systems turned out to be of help for the analysis of real systems. The dynam-
ical situations previously elaborated by solid state physics for very simple
defect-containing 3D inorganic lattices [87, 88] have been found also for or-
ganic 1D classical polymers. Depending on the atomic masses, geometry and
vibrational potentials, highly localized out-of-band or gap modes as well as
strongly collective resonance modes have been calculated and identified as
originating from defects in 1D polymers [82, 83].
From the study of localized and/or quasi-localized modes, spectroscopic
signals correlated with specific defects have been identified and widely used
in the diagnostic analysis of the conformational, chemical and stereochem-
ical disorder of a given real classical organic polymer system of industrial
relevance. The spectra of disordered 1D classical polymers have been treated
in the literature and have been used for detailed structural analysis of the
simplest polymers of industrial relevance, such as polyethylene, polytetraflu-
oroethylene and polyvinylchloride [82, 83].
3.2 Polyconjugated Polymers
The problem becomes more complex when moving to sp2 polyconjugated

systems. As discussed in Sect. 2.3, because of delocalization, interactions
extend to large intramolecular (and, who knows, intermolecular) distances.
EBLA must be considered because of the torsional flexibility of F systems,
cross links as well as Csp3 may truncate delocalization, and many other
chemical and structural impurities of unknown type may occur. It has to be
remembered that even the simplest PA in a real sample is thought (but never
unquestionably shown) to contain cross links and CH2 defects that strongly
perturb delocalization.
As to the knowledge of the interaction distance along the 1D polyconju-
gated chain, great expectations were put in first-principle calculations which
necessarily must take into account the molecular reality of the systems in
one, two and three dimensions. Affording such enormous calculations re-
quires powerful computing capabilities; on the other hand, data from sim-
plistic approaches may lead to wrong conclusions. As a simple example we
quote the problem tackled a few years ago on the effect on delocalization by a
CH2 group separating two sections of polyene chains. Old-fashioned simplified
quantum mechanical calculations (CNDO, MNDO) have indicated that CH2

completely breaks conjugation and builds a huge barrier to the tunnelling of
π electrons through the barrier [89].
Concerning the experimental indication of long range interactions in 1D
polyconjugated systems we consider of extreme interest the experiments
on nonlinear optics of a series of polyacetylene-like chains with growing
lengths [90, 91] from which the final answer is that certainly conjugated dou-
ble bonds interact at least up to ≈ 60 neighbours [91].
3.3 2D sp2 Systems
For graphitic-like systems the issue of local and/or collective normal modes
needs to be updated. We neglect in this discussion the case of defects due
to the chemical synthesis of the material, namely chemical groups or atoms
(functional groups) dangling from the main graphitic lattice or stemming out
of the peripheral bonds. When the ideal view of an infinite graphite lattice is
abandoned and a real graphitic domain is considered, the number of atoms
at the periphery becomes sizeable, no matter of the shape of the domain.
Dynamics tells us that the normal modes at the periphery are many and
may behave like gap modes (highly localized or quasi-resonance modes) or
resonance modes (collective motions) depending on the dynamical conditions
which define the intramolecular coupling and depending on their energies
which may occur away, near or inside a phonon branch [82, 83]. The sec-
ond kind of possible defects of a lattice of graphite is the distortion of the
sp2 skeleton by some kind of strain induced during the chemical reaction.
In this case the work of imagination becomes almost free to conceive many
kinds of either small or large defects. The simplest case is consideration of
a few Csp3 atoms which necessarily generate localized or extended corruga-
tion; one can go further by introducing strained rings or even holes where
dangling bonds are lingering ready to react. Our fundamental questions are
the following:
(i) Is there any unquestionable experimental direct evidence of such corru-
gated structures?
(ii) Assuming that corrugations and wrinkles occur, where and how could
one locate their normal modes in infrared and/or Raman and/or neutron
scattering spectroscopies?
Efforts have been made to work out theoretical models [92,93,94], but match-
ing with unquestionable experimental data has yet to be found. In the opin-
ion of the writers, plenty of work still needs to be done on the dynamics
and experimental spectra of these systems in order to provide clear relation-
ships between the structures and the relative signatures in the infrared and
Raman. From the viewpoint of molecular vibrational dynamics-high-energy
local modes (such as C–H stretchings) arising as defects from boundaries or
holes, in a first approximation, can be overlooked and the attention should
be focused on C–C stretchings, C–H in-plane bendings, CCC bendings and

CCCC out-of-plane modes (from 1600 cm−1 downward). These modes may
be generated by the defects, which are necessarily coupled to the phonons of
the main lattice, thus generating collective resonance modes.
It is obvious that in- and out-of-plane motions in harmonic approximation
for a planar graphitic-like lattice are orthogonal and no coupling takes place.
From previous experience with aromatic molecules [95, 96], we know that
when corrugating defects are introduced such separation is lost, all modes
can couple and the intensities are shared between two or among several cou-
pled modes (just as expected for resonance defect modes). Since, in absence
of chemical substitution (i.e., in absence of highly polar bonds) the vibrations
of the CC skeleton cannot generate large transition dipole moments, char-
acteristic signals in infrared are not likely to occur. On the other hand, CC
bonds are highly polarizable and their stretching may show up in the Raman
spectra as weak features. Calculations of transition moments in the infrared
and cross sections in the Raman may be the clue to locate meaningful and
characteristic vibrational transitions that may provide a partial picture of the
disorder inside the carbon systems under discussion in this Chapter. There
are no studies that have clearly shown the activation in the Raman of vibra-
tions localized on sp3 -induced corrugations. It has to be considered that in
any case the Raman spectra of defective graphite has a strong band around
1300 cm−1 , the so-called D band. The spectral region is unfortunately too
close to the usual frequencies found for the C–C stretching vibrations char-
acteristic of a sp3 carbon system. Therefore the future search for Raman
signatures of skeletal sp3 -induced corrugations in graphitic domains has nec-
essarily to take into account the fact that for visible laser excitation the
Raman cross section originating from the sp2 component overrides the Ra-
man cross section originating from any sp3 -induced corrugation. This is why
recent Raman investigations on such graphitic systems have been pursued
using UV excitations. In the following section we will focus on the D band
and its behaviour, and we will introduce a theoretical explanation of the Ra-
man signal that is based on model molecules mimicking confined sp2 regions
of graphite.
4 Edge Effects in Graphitic Domains

for Structure Diagnosis
A clean contribution to the study of carbon materials supposed to consist
of graphitic domains is given by the class of polycyclic aromatic hydrocar-
bons (PAHs) that have been recently made available from new synthetic
processes [15]. The electronic and vibrational spectra of PAH molecules with
well-defined shape (hence symmetry) and size have been studied, revealing
many interesting aspects related to the confinement of π electrons in two-di-
mensional domains with a size in the nanometer range [41]. The fundamental
(a) (b)
Fig. 2. (a) Raman frequencies of G and D line of a microcristalline graphite sample,

as obtained with different exciting laser energies (from [44, 45]); (b) Modeling of
the Raman frequency dispersion of a disordered graphite sample, according to [13]
understanding of the resonant Raman spectroscopy of PAHs [14] opens the

way to the interpretation of the multiwavelength Raman spectra of nanostruc-
tured graphitic materials. In particular, it becomes possible to effectively use
the D band as a powerful diagnostic tool for obtaining information about
the distribution of domains in a nanostructured graphitic material. Disor-
dered and nanostructured graphite is a very wide class of materials (ranging
from annealed amorphous carbons, microcystalline graphites, graphite fibers
and whiskers, and carbons obtained from controlled pyrolysis of molecular
precursors). All these materials are characterized by the presence of a non-
negligible amount of sp2 carbons that give rise to islands that can be cor-
related (in terms of structure and also in terms of their Raman response)
to an ideally perfect crystal of graphite. Their first-order spectra show two
main signatures: first, a band located at about 1580 cm−1 , which is known
as the graphite (G) band, since it is the only feature observed in the first-
order Raman spectrum of highly ordered crystalline graphite. The second
band (D band) appears at about 1350 cm−1 and it is a characteristic feature
of disordered samples of graphite (e.g., nanocrystalline and microcrystalline
graphites) and amorphous carbon materials containing sp2 graphitic islands.
The fundamental experiments made by Pócsik et al. [44, 45] on microcrys-
talline graphites and those performed by Ferrari et al. [97, 98] on films of
amorphous carbons showed a marked resonance behavior and frequency dis-
persion of the D line when changing the energy of the exciting laser line used
in the Raman experiments. This behaviour is illustrated in Fig. 2.
The presence of the G band can be immediately related to the existence
of structures like graphite in the samples (it corresponds to the Raman active
E2g phonon at the Γ point in the first Brillouin zone of graphite). On the
other hand, the understanding of the physical mechanism lying behind the
activation of the disorder induced D band has been a matter of debate in the
literature. An original interpretation of the origin of the D line in graphitic

materials has been proposed based on a molecular approach [99] built on the
spectroscopic data obtained from a wide series of samples of PAHs identified
as realistic models of nanosized sp2 domains in disordered carbon materials.
The correlation experimentally found between the Raman bands of
graphite and the Raman-active modes of PAHs has been confirmed and ra-
tionalized on the basis of quantum chemistry [14]. In particular, the analysis
of the theoretical results clarifies the origin of the large Raman cross section
for the few (D) bands near 1300 cm−1 observed (and calculated) for all the
PAHs examined. The relevant fact is that the existence of nonequivalent C–C
bonds in PAHs is essential for the activation of normal modes in the D-band
region. The existence of nonequivalent C–C bonds is the very consequence of
the confinement of π electrons in a domain of a few nanometers.
The extension of these fundamental findings on PAHs to nanostructured
graphite is straightforward: nanosized sp2 clusters of condensed aromatic
rings have to be associated to some equivalent PAHs characterized by a sim-
ilar structural relaxation driven by the confinement of π electrons in a finite
domain. The spectroscopic signature of this relaxation is just the appearance
of the D line in the Raman spectrum. Because of the effects of conjugation,
the optical absorption of these nanodomains (thus the resonant Raman condi-
tion) are size-dependent. In particular, a decreasing trend of the energy of the
lowest optical transition is found while increasing the size of the nanodomain
(graphite being a zero-gap material corresponding to the limiting case of infi-
nite size). On the other hand, because of the strong electron–phonon coupling
(see the Chapter by Castiglioni et al. in this book, pp. 381–402) it turns out
that also the frequency of the vibrational mode in the D-band region for
each nanodomain is size-dependent. In other words, the vibronically coupled
π → π ∗ optical excitations in graphitic nanodomains are size-dependent be-
cause of π-conjugation; this fact induces a size dependence also on the vibra-
tional frequency of the Raman-active vibrations (in the D-band region) via
a strong and vibration-selective (A phonon) electron–phonon coupling [42].
The consequence for multiwavelength resonant Raman spectroscopy of the
D band of nanostructured graphite is dramatic: For each given laser only
the nanodomains characterized by a suitable optical gap (i.e., size) are con-
tributing to the signal in the D region; moreover their vibrational frequency
for the above reasons will be size-dependent. In particular, a softening of the
D-band frequency is found for an increasing size of the nanodomains. This
explains the increase of the D-band frequency with respect to the energy of
the resonating laser line found in the experiments carried out by Pócsik et
al. [44, 45].
This kind of behaviour of the D band in the multiwavelength Raman
spectra of nanostructured graphite closely resembles the behaviour of the
ECC band in polyacetylene and related 1D polyconjugated compounds. It is
not by accident that the ECC vibration is characterized by a strong electron–
phonon coupling that drives a gap opening through the Peierls distortion
exactly as the relaxation of the geometry of graphite along the D-band phonon
(A phonon) induces a gap opening at the K point at the border of the first
Brillouin zone [42].
The case of microcrystalline graphite could at first seem fairly different,
since no apparent presence of sp2 nanodomains can be foreseen. Recent stud-
ies carried out through STM and Raman spectroscopy [100] show that in
this case the variety of different π electron confinement situations does not
come from a plethora of nanodomains, but more precisely from the peculiar
electronic structures close to the edges of the graphitic planes. Such edge
electronic states can be mapped by STM and, thanks to ab initio calcula-
tions, they can be correlated with the frontier orbitals of model molecules
such as rylenes, whose Raman spectra show precisely the expected behavior
in a Pócsik diagram [100]. According to STM measurements, such edge states
extend for a few nanometer far from the edge and are extremely sensitive to
the topology of the edge [100].
The theoretical analysis illustrated above allows us to discuss the multi-
wavelength Raman behaviour of two pyrolyzed materials obtained in different
conditions from a molecularly well-defined precursor [9]. Figure 3 illustrates
the experimental results and the comparison with the outcome from a model
of the resonant Raman response produced by a distribution of graphitic nan-
odomains (according to the method described in [13]).
In Fig. 3a we can observe that in the case of the sample pyrolysed at
800◦ C, the frequencies of the D line undergo shifts with the energy of the
exciting laser. The observed frequencies follow the empirical law experimen-
tally determined by Pócsik et al. [44, 45]. As illustrated by the simulation
reported in Fig. 3a, it is possible to predict this behavior using as input of
the model [13] a homogeneous distribution of nanodomains ranging from very
small size (6 carbons) up to a large size (384 carbons).
A completely different behavior is experimentally found for a sample ob-
tained from the same precursor but pyrolysed at a lower temperature (600◦ C,
see Fig. 3b). In this case one does not observe frequency shifts with laser ex-
citation in the visible range. The behavior is markedly different from that of
disordered graphite, and it can be correctly predicted only if a different dis-
tribution of nanodomains is used as input in modeling the Raman response.
In particular, the observed behaviour is found when domains containing at
least 150 carbon atoms are selected in the model of the Raman response (see
Fig. 3b). Accordingly, we can conclude that this sample is structurally more
controlled and it keeps memory of the precursor molecule. This is consistent
with the milder pyrolysis conditions (600◦ C as opposed to 800◦ C).
(a)
(b)
Fig. 3. Multiwavelength Raman spectra of a pyrolytic carbon sample obtained

from hexaphenylbenzene (HPB) pyrolysed at 800◦ C (a) and 600◦ C (b) [9]. Com-
parison with Pócsik data [44,45] and with the results from a modeling of the Raman
response according to [13]. The Raman frequencies of the D peak are reported in
a Pócsik diagram [44, 45] for the three excitations used (1.96, 2.41, 2.71 eV, corre-
sponding, respectively, to the red, green and blue spectra represented) to compare
with the dispersion of the D peak of microcrystalline graphite
5 Approaching the Structure of Carbonaceous Materials

by Vibrational Spectroscopy: Imagination and Reality
It is well known that experimental solid state physics has tremendously con-
tributed to the development of the field of carbon-based materials by invent-
ing various techniques (e.g., arc discharge, chemical vapour deposition, laser
ablation, plasma-assisted vapour deposition, molecular beam deposition, etc.)
for the generation of clusters of carbon atoms that are then deposited onto
a surface. The physical properties of such deposits are of relevant interest in
modern technology. These pioneering works have also generated well-defined
and important molecules such as fullerene, fulleroids and nanotubes. All these
techniques are discussed in detail in specialized publications.
Our previous lengthy systematic approach towards the understanding of
the structure of disordered carbon-based materials is aimed at discovering
the structural reason why hydrogen cannot be ‘adsorbed’ in such systems.
Necessarily when a hydrogen molecule enters such complicated structures it
must find a ‘cavity’ where weak interatomic interactions favour its settling
inside the cavity itself. The description of the amorphous surrounding may
be represented by the following two extreme cases.
(i) Amorphous carbon materials with a different degree of local organiza-
tion, such as the models proposed in [101] or [102], are the results of
sophisticated large-scale atomistic simulations. The systems consist of a
large variety of structural entities that find carbon atoms bonded to each
other in various disordered and irregular geometries (quite often away
from their canonical states of hybridization). During the numerical cal-
culations the system may have reached a minimum energy structure as
a balance of all sorts of internal strains and stresses. In these models we
find a whole variety of strained small rings, cross links, curled chains of
various lengths; necessarily a random distribution of voids (or cavities) of
various sizes occurs in such materials. If this is the ‘real’ disorder of the
materials we are considering, spectroscopy can give little help in charac-
terizing the system, since all possible kinds of vibrational motions may
occur, thus generating a very broad vibrational spectrum with no partic-
ular features. Fortunately, the system may evolve locally towards islands
of more stable forms of aggregation, characterized by carbon atoms in
a given hybridization state. These can be considered as clusters of di-
amond, graphite, or segments of polyacetylene or polyyne. In this case
spectroscopic signatures should be found, and are indeed found, which
nicely correlate with the known spectroscopic behaviour of the parent
crystalline systems [103, 104, 105] as discussed in the previous pages of
this Chapter. By suitable annealing such metastable systems may evolve
towards graphitisation [97, 98].
(ii) An alternative picture is to consider a graphitic system consisting mostly
of intermingled graphitic nanodomains with holes and imperfections em-
(I) (II) (III)
Fig. 4. Sketches of the chemical structures of HPB (I ), HBC (II ) and halogenated
HPB (III )
bedded in ‘amorphous’ surroundings. This has been the inspiring idea un-
derlying the project denominated “Molecular Approach to Carbon based
Materials for Energy Storage” – MAC-MES [8].
While materials obtained from vapour phase deposition, arc discharge,
etc., and also simply as carbon soot, etc., may possess a structure as in (i)
(which we already know does not adsorb hydrogen), systems as in (ii) were
obtained only recently and provided new spectroscopic probes for the explo-
ration of their structure and of their capability to adsorb hydrogen [7,8,9,10].
As a contribution towards a systematic understanding of the structural fea-
tures of disordered carbon-based materials, we have recently reported on
the infrared and Raman spectra of materials obtained from thermodynam-
ically controlled pyrolytic processes of model aromatic molecules suitably
selected and prepared as precursors of structurally controlled graphitic ma-
terials. Hexa-phenyl-benzene (HPB, (I) in Fig. 4) has been taken as the most
relevant precursor molecule to be pyrolyzed at various temperatures under
controlled conditions [10].
The concepts underlying our work were the following: It was previously
shown [15] that polyphenyl precursors, with appropriate topology, can be
smoothly transformed into flat polycyclic aromatic hydrocarbons by an in-
tramolecular cyclodehydrogenation. Typical is the case of HBC, which when
treated with FeCl3 , produces hexa-pery-hexabenzocoronene ((II) in Fig. 4).
It was thought that a pyrolytic treatement may cause two main processes:
(a) intramolecular dehydrogenation thus generating molecules of type II, and
(b) intermolecular dehydrogenation coupling various HBC units which could
then undergo intramolecular dehydrogenation as in (a). Suitable halogena-
tion (with halogen X, X=Cl, Br, I) in para position of each benzene ring in
HPB ((III) in Fig. 4) could favour the intermolecular linking with elimination
of the corresponding acid HX before entering the process as in (a). Pyrolysies
were carried out at 450, 500, 600 and 800◦ C for various lengths of time (from
1 to 5 days). The typical morphology of the carbonaceous graphitic material
obtained is given in Fig. 5.
First, from the observation of dominant G and D lines in the Raman

spectra, it was verified that the systems obtained were mostly graphitic (see
Sect. 4). Then the chemical structure of the boundaries of the graphitic do-
mains obtained were studied in terms of the dynamics and spectra of the
peripheral C–H bonds, which show localized out-of-plane deformation modes
(decoupled from the skeletal modes). These modes become specific of the lo-
cal chemical structure as given in Table 2 [10]. As an example of application
of these spectroscopic correlations, we show the structural evolution of the
boundaries of the graphitic domains as revealed by the infrared spectra of
HPB heated at 450, 500 and 600◦C (Fig. 6).
Table 2. Top: description of SOLO, DUO, TRIO and QUATRO structures. Bot-
tom: patterns of absorption bands in the infrared to be associated to specific sets
of adjacent aromatic C–H groups belonging to “fused” benzene rings (frequency
ranges are given in cm−1 )
SOLO 860–910 – –
DUO 800–810 810–860 –
TRIO 750–770 770–800 800–810
QUATRO 730–750 750–770 –
A further interesting structural observation has been derived from these

studies. Let us assume that during the pyrolysis HPB molecules are not
broken up, but are only forced to ‘condense’ and/or link to each other. Ideally
we can construct two main repeating ‘motives’, which show borders we label
as ‘regular’ (reg) and ‘irregular’ (irreg, see Fig. 7). In absence of any region-
selective reactions, simple statistical reasoning shows that the growing of reg
structures becomes less and less probable favouring the formation of irreg
structures. As shown in Fig. 7, irreg clusters necessarily contain ‘holes’ where
six atoms are missing and are replaced by six hydrogens. Whenever the size of
the clusters increases, the relative population of TRIO structures decreases in
favour of SOLO and DUO. In principle, we can distinguish reg or irreg clusters
by the relative population of SOLO and DUO, namely for reg structures the
ratio (number of DUO/number of SOLO) is always one. For irreg structures
with holes, the population of SOLO is larger than the population of DUO. As
an example, we compare in Fig. 6 the infrared spectra of HPB heated for one
day at 450, 500 and 600◦ C. The SOLO and DUO bands increase relative to
Fig. 5. SEM images of HPB-800-1d at increasing magnifications
Fig. 6. Infrared spectra of HPB pyrolysed at increasing temperatures (450, 500,

600◦ C)
the TRIO band, thus indicating that the sizes of the clusters have increased.
5.1 Carbon-Based Materials for Hydrogen Storage
The last question which needed to be answered is whether the graphitic mate-
rials suitably prepared as described previously in Sect. 5 were able to adsorb
hydrogen. Experiments were done with the usual techniques [4, 5]. In addi-
tion, a spectroscopic method based on Raman scattering was developed that
provides many interesting details on the physics of the ‘adsorption’ of hydro-
gen in these materials [7]. A special cell for Raman scattering experiments
Fig. 7. Examples of regular and irregular structures with different relative

TRIO/SOLO/DUO weigths: (a) regular, TRIO/SOLO/DUO = 5/2/2; (b) irregu-
lar, TRIO/SOLO/DUO = 2/1/1; (c) regular, TRIO/SOLO/DUO = 5/2/2; (d) ir-
regular, TRIO/SOLO/DUO = 2/2/1
Fig. 8. Raman spectra of hydrogen adsorbed on a pyrolysed carbon (HPB-550-3d)

recorded at 40 ± 2 K at pressures ranging from 4 to 62.5 bar (from [7])
Fig. 9. The effect of a uniform static electric field along the bond axis of the
hydrogen molecule. The field is generated by the array of point charges depicted
in the top-left panel. Comparison with the polarization effects on the bond lengh,
vibrational frequency and infrared absorption induced by the benzene molecule.
Data from [106], according to density functional calculations (B3LYP/6-311G**).
Notice that these calculations show only the softening (downshift of the vibrational
frequency) of the H–H bond induced by interactions with aromatic carbon atoms.
Actually the theory of the interaction also predicts an upward shift [7], as also
revealed experimentally in the Raman spectrum of H2 under pressure (see Fig. 8)
was designed and built; this cell is capable of operating at very low temper-
ature (40 K) and both in high vacuum or high pressure (up to 100 bars)
with relatively high resolution. We could easily record the vibrorotational
and rotational Raman spectra of hydrogen as function of temperature and
pressure in absence or in presence of carbon-based materials described in (i)
and (ii) in Sect. 5. Typical Raman spectra of hydrogen adsorbed on a pyrol-
ysed carbon are shown in Fig. 8. As discussed in [7], we do observe signatures
of interaction of hydrogen with carbon materials both in cases (i) and (ii) for
a small relative amount of the total hydrogen introduced in the cell. Clear
indications are found that no charge transfer phenomena are involved in the
interaction with the absorbing material (contrary to [3]), but only a simple
physisorption takes place; no indication of chemisorption has been noticed.
The lesson we learned from these studies is that hydrogen does not find
the suitable energetic and sterical conditions to be squeezed into the ‘cavi-
ties’ of any carbon-based materials (so far known or tested), including carbon
nanotubes, even at low temperatures and high pressures. This information
from experiments weakens the validity of several theoretical elaborations pre-
sented in the literature. From our side, attempts at understanding such failure
consisted first in the application of classical potentials for describing the in-
teraction of H2 with carbons. It turned out that the the potential function
is extremely flat with a very shallow potential minimum, which indicates a
very weak interaction with the surrounding carbon atoms within the cavity.
Next, a quantum mechanical modelling of the interaction of the hydrogen
molecule with its surrounding has been attempted; interatomic distances, vi-
brational frequencies, infrared and Raman vibrational intensities have been
calculated for hydrogen interacting with benzene (taken as a model of aro-
matic carbons). The results from this computation have been compared with
the results obtained by acting on the hydrogen molecule with a static electric
field mimicking the interacting carbon substrate (see Fig. 9 and [106]). The
final relevant information which derives from these works is that in order to
keep hydrogen inside the cavity, a stronger interaction, which may induce
a larger dipole in the H–H bond, has to be obtained by introducing highly
polarizing atoms in the cavity. The idea was empirically conceived during the
synthetic part of the work, and precursors containing various nitrogen atoms
were prepared and then pyrolyzed [9, 10]. Unfortunately, no real positive re-
sults were obtained on these materials.
The results from these studies have suggested to set aside the idea that
carbon cages (of the average size existing in the material discussed in the
previous pages) may be able to store hydrogen. Smaller carbon cages or
new cavities with highly charged atoms (i.e., those that are strongly po-
larizing) should be conceived. Recently a whole new class of more ordered
organic/inorganic materials with tailored cages has been synthesized [107],
and the Raman spectra of deuterium adsorbed in one of these materials has
been obtained and studied [108]. It is very pleasing to realize that from an in-
terdisciplinary collaboration between chemists, physicists and spectroscopists
new routes for future studies on hydrogen storage have been indicated.
6 Conclusions
We hope the reader has developed a feeling about the complexity and the
yet unsolved problem in understanding the structural properties of the car-
bon-based materials that are now being considered as relevant for technolog-
ical applications. Basically we suggest that, so far, hydrogen cannot yet be
stored in a useful amount for automotive applications in the carbon-based
materials presently available. However, new hopes arise with new classes of
organic/inorganic cage molecules. The problem of hydrogen storage has mo-

tivated the development and use of the physical techniques presented here,
which may find a broader application in the field of the characterization of
materials.
Acknowledgements
We acknowledge all the members of the MAC-MES consortium [8] for fruit-
ful exchange of ideas and constructive collaboration. In particular, we thank
Prof. K. Müllen and Dr. C. Mathis for their contributions in the field of
graphenes and Dr. L. Gherardi of Pirelli Labs for many discussions on
hydrogen absorption. Part of this work was supported by a grant from
MURST – Italy (FIRB project “Carbon-based micro- and nanostructures”,
RBNE019NKS) and by the MAC-MES contract (European Commission,
Fifth Framework Programme, Research Project “MAC-MES: Molecular Ap-
proach To Carbon Based Materials for Energy Storage” – G5RD-CT2001-
00571).
References
[1] A. Chambers, C. Park, R. Terry, K. Baker, N. M. Rodriguez: J. Phys. Chem.
B 102, 4253 (1998) 23
[2] P. Chen, X. Wu, J. Lin, K. L. Tan: Science 285, 91 (1999) 23
[3] A. C. Dillon, T. Gennet, J. L. Alleman, K. M. Jones, P. A. Parilla,
M. J. Heben: Proceedings of the 2000 U.S. DOE Hydrogen Program Review
(2000) nREL/CP-570-28890 23
[4] M. Hirscher, M. Becher: J. Nanosci. Nanotechnol. 3, 3 (2003) 23
[5] M. Hirscher, M. Becher, M. Haluska, F. von Zeppelin, X. Chen, U. Dettlaff-
Wegligowska, S. Roth: J. Alloys Compd. 356, 433 (2003) 23
[6] B. Panella, L. Kossykh, U. Dettlaff-Weglikowska, M. Hirscher, G. Zerbi,
S. Roth: Volumetric measurement of hydrogen storage in HCl-treated polyani-
line and polypyrrole, Synthetic Met. 151 (3), 208–210 (2005) 23
[7] A. Centrone, L. Brambilla, G. Zerbi: Phys. Rev. B 71, 245406 (2005) 23
[8] European Commission, Fifth Framework Programme, Research Project
“MAC-MES: Molecular Approach To Carbon Based Materials for Energy
Storage” (G5RD-CT2001-00571) 24
[9] T. Renouard, L. Gherghel, M. Wachtler, F. Bonino, B. Scrosati, R. Nuffer,
C. Mathis, K. Müllen: J. Power Sources 139, 242 (2005) 24
[10] A. Centrone, L. Brambilla, T. Renouard, L. Gherghel, C. Mathis, K. Müllen,
G. Zerbi: Carbon 43, 1593 (2005) 24
[11] F. Negri, C. Castiglioni, M. Tommasini, G. Zerbi: J. Phys. Chem. A 106,
3306 (2002) 24
[12] E. D. Donato, M. Tommasini, G. Fustella, L. Brambilla, C. Castiglioni,
G. Zerbi, C. D. Simpson, K. Müllen, F. Negri: Chem. Phys. 301, 81 (2004)
24
[13] C. Castiglioni, E. D. Donato, M. Tommasini, G. Zerbi: Synthetic Met. 139,

885 (2003) 24, 28
[14] F. Negri, E. D. Donato, M. Tommasini, C. Castiglioni, G. Zerbi, K. Müllen:
J. Chem. Phys. 120, 11889 (2004) 24
[15] M. D. Watson, A. Fechtenkötter, K. Müllen: Chem. Rev. 101, 1267 (2001)
24
[16] F. Schuetz, K. Sing, J. Weitkamp (Eds.): Handbook of Porous Solids, vol. 3
(Wiley-VCH, Weinheim) Chap. 4.8 25
[17] H. W. Kroto, J. E. Fischer, D. E. Cox: The Fullerenes (Pergamon 1993) 25
[18] M. Dresselhaus, G. Dresselhaus, P. Avouris (Eds.): Carbon Nanotubes – Syn-
thesis, Structure, Properties and Applications (Springer, Berlin, Heidelberg
2001) 25, 28
[19] E. B. Wilson, J. C. Decius, P. C. Cross: Molecular Vibrations (McGraw-Hill,
New York 1955) 25
[20] S. Califano: Vibrational States (Wiley, New York 1976) 25
[21] J. O. Hirschfelder, C. F. Curtiss, R. B. Bird: Molecular Theory of Gases and
Liquids (Wiley, New York 1964) 25
[22] A. I. Kitaigorodsky: Molecular Crystals and Molecules (Academic, London
1973) 25
[23] A. J. Pertsin, A. I. Kitaigorodsky: The Atom-Atom Potential Method
(Springer, Berlin, Heidelberg 1987) 25
[24] K. Müllen, G. Wegner (Eds.): Electronic Materials: The Oligomer Approach
(Wiley-VCH, Weinheim 1998) 25
[25] G. Zerbi, M. Veronelli, S. Martina, A. D. Schluter, G. Wegner: Adv. Mater.
6, 385 (1994) 25
[26] R. G. Snyder, J. H. Schachtschneider: Spectrochim. Acta 19, 85 (1963) 26,
27
[27] J. H. Schachtschneider, R. G. Snyder: Spectrochim. Acta 19, 117 (1963) 26,
27
[28] M. Prelas, G. Popovici, L. Bigelow (Eds.): Handbook of Industrial Diamond
and Diamond Films (Dekker, New York 1997) 26
[29] R. Tubino, L. Piseri, G. Zerbi: J. Chem. Phys. 56, 1022 (1972) 26, 27
[30] S. Flügge: Handbuch der Physik, vol. VII/1 (Springer, Berlin 1995) 26
[31] J. L. Warren, R. G. Wenzel, J. L. Yarnell: Inelastic Scattering Neutrons Proc.
Symp. 5th, vol. 1 (Vienna, Austria 1965 (1966)) p. 361 26
[32] L. Piseri, G. Zerbi: J. Chem. Phys. 48, 3561 (1968) 27
[33] G. Zerbi, L. Piseri: J. Chem. Phys. 49, 3840 (1968) 27
[34] S. Landa, V. Machacek: Collect. Czech. Chem. Commun. 5, 1 (1933) 27
[35] J. E. Dahl, S. G. Liu, R. M. K. Carlson: Science 299, 96 (2003) 27
[36] J. E. P. Dahl, J. M. Moldowan, T. M. Peakman, J. C. Clardy, E. Lobkovsky,
M. M. Olmstead, P. W. May, T. J. Davis, J. W. Steeds, K. E. Peters, A. Pep-
per, A. Ekuan, R. M. K. Carlson: Angew. Chem. Int. Edit. 42, 2040 (2003)
27
[37] J. O. Jensen: Spectrochim. Acta A 60, 1895 (2004) 27
[38] P. R. Wallace: Phys. Rev. 71, 622 (1947) 71, 622 (1947). 28
[39] P. R. Wallace: Phys. Rev. 72, 258 (1947) 28
[40] A. Coulson, R. Taylor: P. Phys. Soc. Lond. A 65, 815 (1952) 28
[41] C. Castiglioni, M. Tommasini, G. Zerbi: Philos. T. Roy. Soc. A 362, 2425

(2004) 28
[42] M. Tommasini, E. Di Donato, C. Castiglioni, G. Zerbi: Relaxing the graphite
lattice along critical directions: The effect of the electron–phonon coupling on
the pi electron band structure, Chem. Phys. Lett. 414 (1-3), 166–173 (2005)
28, 29
[43] S. Piscanec, M. Lazzeri, F. Mauri, A. C. Ferrari, J. Robertson: Phys. Rev.
Lett. 93, 185503 (2004) 28, 29
[44] Pócsik, M. Hunfhausen, M. Koós, L. Ley: J. Non-Cryst. Solids 227–230, 1083
(1998) 28
[45] M. J. Matthews, M. A. Pimenta, G. Dresselhaus, M. S. Dresselhaus, M. Endo:
Phys. Rev. B 59, 6585 (1999) 28
[46] J. Maultzsch, S. Reich, C. Thomsen: Phys. Rev. B 70, 155403 (2004) 28
[47] C. Thomsen, S. Reich: Phys. Rev. Lett. 85, 5214 (2000) 28
[48] T. Skotheim (Ed.): Handboook of Conducting Polymers, vol. 1 (Dekker, New
York 1986)
[49] T. Skotheim (Ed.): Handboook of Conducting Polymers, vol. 2 (Dekker, New
York 1986)
[50] T. Skotheim, R. Elsenbaumer, J. Reynolds (Eds.): Handbook of Conducting
Polymers, 2nd ed. (Dekker, New York 1988)
[51] G. Leising, R. Uitz, B. Ankele, W. Ottinger, F. Steltzer: Mol. Cryst. Liq.
Cryst. 117, 327 (1985)
[52] H. Naarman: Synthetic Met. 17, 223 (1987)
[53] G. Britton, S. Liaaen-Jensen, H. Pfander (Eds.): Carotenoids, vol. 1B
(Birkhäuser, Basel, Switzerland 1995)
[54] R. E. Peierls: Quantum Theory of Solids (Clarendon, Oxford 1955)
[55] G. Zerbi, M. D. Zoppo: Polimeri – aspetti chimico fisici, in Enciclopedia Ital-
iana Treccani, vol. XIII, Enciclopedia del 900 (2004) p. 285
[56] M. Gussoni, C. Castiglioni, G. Zerbi: Vibrational spectroscopy of polyconju-
gated materials: Polyacetylene and polyenes, in R. Clark, R. Hester (Eds.):
Spectroscopy of Advanced Materials, vol. 19 (Wiley, New York 1991) p. 25
[57] C. Castiglioni, J. T. L. Navarrete, G. Zerbi, M. Gussoni: Solid State Commun.
65, 625 (1988)
[58] E. Ehrenfreund, Z. Vardeny, O. Brafman, B. Horovitz: Phys. Rev. B 6, 1535
(1987)
[59] Z. Vardeny, E. Ehrenfreund, O. Brafman: Phys. Rev. Lett. 51, 2326 (1983)
[60] A. Bianco, M. D. Zoppo, G. Zerbi: J. Chem. Phys. 120, 1450 (2004)
[61] P. Chandrasekhar: Conducting Polymers and Fundamentals and Applications:
A Practical Approach (Kluwer, Boston 1999)
[62] M. C. Rumi, G. Zerbi: Chem. Phys. 242, 123 (1999)
[63] B. Tian, G. Zerbi, R. Schenk, K. Müllen: J. Chem. Phys. 95, 3191 (1991)
[64] B. Tian, G. Zerbi, K. Müllen: J. Chem. Phys. 95, 3198 (1991)
[65] J. T. L. Navarrete, G. Zerbi: J. Chem. Phys. 94, 957 (1991)
[66] J. T. L. Navarrete, G. Zerbi: J. Chem. Phys. 94, 965 (1991)
[67] M. C. Rumi, G. Zerbi: J. Chem. Phys. 108, 8662 (1998)
[68] J. E. Anthony, D. L. Eaton, S. R. Parkin: Organic Letters 4, 15 (2002)
[69] M. M. Payne, S. R. Parkin, J. E. Anthony: J. Am. Chem. Soc. 127, 8028
(2005)
[70] G. Zerbi, M. Gussoni, C. Castiglioni: Vibrational spectroscopy of polycon-

jugated aromatic materials with electrical and nonlinear optical properties,
in J. L. Brédas, R. Silbey (Eds.): Conjugated Polymers (Kluwer, New York
1991) p. 435
[71] M. Rumi, G. Zerbi: Chem. Phys. 242, 123 (1999)
[72] A. M. van de Craats, J. M. Warman, A. Fechtenkötter, J. D. Brand,
M. A. Harbison, K. Müllen: Adv. Mater. 11, 1469 (1999)
[73] M. Carminati, L. Brambilla, G. Zerbi, K. Müllen, J. Wu: Nanoscale archi-
tectures for molecular electronics: Vibrational spectroscopy and structure of
solid hexa-n-dodecyl-hexa-peri-hexabenzocoronene, J. Chem. Phys. 123 (14)
(2005)
[74] A. Milani, L. Brambilla, M. Del Zoppo, G. Zerbi: manuscript in preparation
[75] T. Luu, E. Elliott, A. D. Slepkov, S. Eisler, R. McDonald, F. A. Hegmann,
R. R. Tykwinski: Organic Letters 7, 51 (2005)
[76] S. Eisler, A. D. Slepkov, E. Elliott, T. Luu, R. McDonald, F. A. Hegmann,
R. R. Tykwinski: J. Am. Chem. Soc. 127, 2666 (2005)
[77] M. Saito, Y. Okamoto: Phys. Rev. B 60, 8939 (1999)
[78] F. Cataldo: Carbon 42, 129 (2004)
[79] F. Cataldo: Tetrahedron Lett. 45, 141 (2004)
[80] A. Lucotti, M. Tommasini, M. Del Zoppo, C. Castiglioni, G. Zerbi, F. Cataldo,
C. S. Casari, A. L. Bassi, V. Russo, M. Bogana, C. E. Bottani: Raman and
SERS investigation of isolated sp carbon chains, Chem. Phys. Lett. 417 (1-3)
(2006)
[81] G. Zerbi: Probing the real structure of chain molecules by vibrational spec-
troscopy. Some recent aspects, in Advances in Chemistry Series, vol. 203
(American Chemical Society 1983) p. 487
[82] G. Zerbi: Vibrational spectroscopy of very large molecules, in R. Clark,
R. Hester (Eds.): Advances in Infrared and Raman Spectroscopy, vol. 11
(Wiley-Heyden, New York 1984) p. 301
[83] G. Zerbi, M. D. Zoppo: Vibrational spectra as a probe of structural or-
der/disorder in chain molecules and polymers, in G. Zerbi (Ed.): Modern
Polymer Spectroscopy (Wiley-VCH, Weinheim 1998) p. 87
[84] M. Tasumi, G. Zerbi: J. Chem. Phys. 48, 3813 (1968)
[85] M. Gussoni: Infrared and Raman intensities from electrooptical parameters,
in R. Clark, R. Hester (Eds.): Advances in Infrared and Raman Spectroscopy,
vol. 6 (Heyden, London 1964) p. 96
[86] M. Gussoni, C. Castiglioni, G. Zerbi: Vibrational intensities: Interpretation
and use for diagnostic purposes, in J. Chalmers, P. Griffith (Eds.): Handbook
of Vibrational Spectroscopy, vol. 3 (Wiley, New York 2002) pp. 2040–2077
[87] A. A. Maradudin: Solid State Physics – Advances in Research and Applica-
tions, vol. 18 (Academic 1966)
[88] A. A. Maradudin: Solid State Physics – Advances in Research and Applica-
tions, vol. 19 (Academic 1967)
[89] J. T. L. Navarrete, G. Zerbi: Solid State Commun. 64, 1183 (1987)
[90] I. D. W. Samuel, I. Ledoux, C. Dhenaut, J. Zyss, H. H. Fox, R. R. Schrock,
R. J. Silbey: Science 265, 1070 (1994)
[91] I. Ledoux, I. D. W. Samuel, J. Zyss, S. N. Yaliraki, F. J. Schattenmann,
R. R. Schrock, R. J. Silbey: Chem. Phys. 245, 1 (1999)
[92] X. Rocquefelte, G. M. Rignanese, V. Meunier, H. Terrones, M. Terrones,

J. C. Charlier: Nano Lett. 4, 805 (2004)
[93] F. Valencia, A. H. Romero, E. Hernandez, M. Terrones, H. Terrones: New J.
Phys. 5, 123 (2003)
[94] H. Terrones, A. L. Mackay: Carbon 30, 1251 (1992)
[95] G. Zerbi, S. Sandroni: Spectrochim. Acta 24A, 483 (1968)
[96] G. Zerbi, S. Sandroni: Spectrochim. Acta 24A, 511 (1968)
[97] A. C. Ferrari, J. Robertson: Phys. Rev. B 61, 14095 (2000)
[99] C. Castiglioni, F. Negri, M. Rigolio, G. Zerbi: J. Chem. Phys. 115, 3769
(2001)
[100] M. Tommasini, N. Severin, E. D. Donato, C. Castiglioni, J. Rabe, G. Zerbi:
manuscript in preparation
[101] M. Profeta, F. Mauri: Phys. Rev. B 63, 245415 (2001)
[102] D. Donadio, L. Colombo, P. Milani, G. Benedek: Phys. Rev. Lett. 83, 776
(1999)
[104] H. Kuzmany, R. Pfeiffer, N. Salk, B. Gunther: Carbon 42, 911 (2004)
[105] C. S. Casari, A. L. Bassi, L. Ravagnan, F. Siviero, C. Lenardi, E. Barborini,
P. Piseri, P. Milani, C. E. Bottani: Carbon 42, 1103 (2004)
[106] F. Toffolo, M. Tommasini, M. D. Zoppo, G. Zerbi: Chem. Phys. Lett. 405,
108 (2005)
[107] N. W. Ockwig, O. Delgado-Friedrichs, M. O’Keeffe, O. M. Yaghi: Reticular
chemistry: Occurrence and taxonomy of nets and grammar for the design of
frameworks, Accounts Chem. Res. 38 (3), 176–182 (2005)
[108] A. Centrone, D. Y. Siberio-Perez, A. R. Millward, O. M. Yaghi, A. J. Matzger,
G. Zerbi: Raman spectra of hydrogen and deuterium adsorbed on a metal-
organic framework, Chem. Phys. Lett. 411 (4-6), 516–519 (2005)
Index
sp2 -bonded clusters, 38 defects, 27, 28, 33–35
density functional theory (DFT), 32
annealing, 41 diamond, 26, 41
diamond-like carbon (DLC), 25
bond length alternation (BLA), 31 dispersion
effective BLA (EBLA), 31, 34 phonon dispersion, 26
confinement, 28, 36, 38, 39 effective conjugation coordinate theory

conjugation length, 30 (ECCT), 38
effective conjugation length (ECL), electron–phonon coupling, 28, 30, 38
30
cumulene, 32 first principle calculations, 24, 28, 30,
32, 34
D band, 36–38 fullerene, 25, 41
D-band dipersion, 37
D-band, 38 G band, 37
graphite, 28, 41 Peierls distortion, 29, 32, 39

graphite-like carbon (GLC), 25 polyacetylene (PA), 28, 29, 32, 41
graphitization, 41 polyconjugated systems, 31
polycyclic aromatic hydrocarbons
HOMO–LUMO gap, 29, 30 (PAHs), 36–38, 42
hydrogen adsorption, 42, 44, 45 polymers, 29, 30, 33, 34
hydrogen desorption, 42 polyconjugated polymers, 30, 31, 34
hydrogen storage, 44 polymethylene, 25
polyyne, 32, 33, 41
infrared (IR) spectroscopy, 24, 25,
pyrolytic carbons, 37, 42
33–36, 42
Kohn anomaly, 28, 32 quantum chemistry, 38
lattice dynamics, 24, 27, 29, 33 Raman scattering, 35

Raman cross section, 29, 36
molecular approach, 24, 29, 38 Raman intensity, 34
molecular dynamics, 24, 27, 33, 35 resonant Raman scattering, 28, 30
multiwavelength Raman spectroscopy selection rules, 28
(MWRS), 37–40 Raman spectroscopy, 24–28, 32–36, 42
nanotubes, 25, 28, 41
surface enhanced Raman spectroscopy
oligomers, 25, 27, 28, 30–32 (SERS), 32
Pócsik diagram, 37, 39, 40 valence force field, 25–27, 30

Biocompatibility, Cytotoxicity and Bioactivity
of Amorphous Carbon Films
Sandra E. Rodil1 , René Olivares2 , Higinio Arzate2 , and Stephen Muhl1

1
Instituto de Investigaciones en Materiales, Universidad Nacional Autónoma de
México, Circuito Exterior s/n, CU, MEX-04510 México D. F.
2
Facultad de Odontologı́a, Universidad Nacional Autónoma de México, Circuito
Exterior s/n, CU, MEX-04510 México D. F.
ser38@zinalco.iimatercu.unam.mx
Abstract. In this work we investigate the possibility of using an amorphous car-

bon layer as a bioactive coating that promotes bone ingrowth for medical implants
in direct contact with bone. The initial step was to confirm the tissue compatibil-
ity of the amorphous carbon coatings by studying cellular adhesion, proliferation
and viability with human osteoblasts and fibroblasts, cultured on the carbon sur-
face. To investigate the bone-bonding capabilities of the sputtered a-C coatings,
“in vitro” mineralisation studies were performed. For the biomineralisation assays,
human osteoblasts were grown on a-C samples for periods up to 14 days, leading
to the formation of mineralised structures that were morphologically and biochem-
ically investigated by the scanning electron microscope and X-ray microanalysis,
respectively.
1 Introduction
There is an increasing interest in the development of novel coatings or sur-

face modification treatments to improve the bone-bonding ability of metal
implants [1]. An essential condition for an artificial material to bond directly
to living bone is the formation of bone-like apatite on its surface. Direct bone-
implant joint formation is important since if the ingrowing bone is separated
from the implant by an intervening soft tissue layer, excessive micromove-
ments in the interface may occur. These micromovements are known to pro-
duce negative reactions such as destruction of the surface oxide layer and wear
of the implant surface, and these can be accompanied by corrosion and wear
debris accumulation, causing the subsequent failure of the implant [2]. The
parameters that are important for long-term implant life include the biocom-
patibility of the chosen implant material, the surface chemistry and physics,
the macro- and microstructure, the surgical implant procedure, the time and
mode of implant loading and the implantation bed itself. Therefore, the bio-
compatibility of an implant is only one of parameters influencing the tissue
response to an implant material; other factors must be satisfied for a device
to be suitable for implantation. Particularly, for bone ingrowth many factors
have been shown to be important: (a) mechanical factors, such as micro-
motion and loading; (b) geometrical factors: the surface of ingrowth fixation
56 Sandra E. Rodil et al.
implants must have a geometry that accommodates the bone ingrowth, and
therefore noncemented porous coated implants have been proposed where
the pore size and the interface gap play major roles; and (c) materials fac-
tors: the composition of the surface layer is a very important determining
factor [3]. Porous coatings on titanium (Ti) and cobalt–chrome (Co–Cr) al-
loys are the most commonly used materials for odontological and orthopedic
implants. However, the failure rates today demonstrate that there is still
scope for improvement [4]. Other potential implant materials include mostly
ceramics, such as, hydroxyapatite [5], tricalcium phosphate [6] and carbon-
based ceramics (pyrolitic carbon, glassy carbon) [7]. All have been shown
to be conducive to bone ingrowth (osteoinductive), but they have not been
used for load-bearing implants due to strength concerns. Here is where the
surface modification comes into play; by modifying the surface chemistry of
an implant, it is possible to retain the mechanical and geometrical conditions
needed to assure the long-term stability of the implant, together with enhanc-
ing a particular biological response such as bone ingrowth [8]. Both pyrolitic
and glassy carbon have been shown to be osteoinductive, but they are not me-
chanically stable. Therefore carbon-based coatings having graphite-like prop-
erties deposited on mechanically stable substrates are interesting systems to
evaluate for their potential to enhance bone ingrowth. Amorphous carbon
(a-C) and diamond-like carbon (DLC) films are known as bioinert materials
with no toxic reactions with living organism [9, 10]. Furthermore, the combi-
nation of high hardness, low coefficient of friction, high wear and corrosion
resistance, with the bioinert character is a good reason for selecting carbon
films as a surface finish on biomedical implants. Previous studies have shown
that these films exhibited good compatibility with different cell types, includ-
ing macrophages, fibroblasts, human myeloblasts and human embryo kidney
cells [11, 12, 13, 14, 15, 16]. They have also shown excellent haemocompati-
bility, reduced platelet adhesion, activation and aggregation, which are good
indicators of reduced thrombus formation [17, 18, 19]. Another important fac-
tor in favor of the amorphous carbon films as a biocompatible surface is that
their properties (surface energy, electrical conductivity, tribological proper-
ties, etc.) can be tailored by the deposition conditions or doping/bonding
using additional elements to induce specific biological responses [20, 21]. To-
day, two main biomedical applications of carbon-based coatings can be seen:
those of DLC in blood-contacting devices (stents and heart valves); and the
use of DLC to reduce wear in load-bearing joints. Coronary artery stents
and heart valves are affected by platelet activation due to blood contact with
the biomaterial or the release of metallic ions. This activation is an impor-
tant trigger for thrombosis. To solve this situation, the use of coatings on
metallic stents and heart valves has been suggested. Of these coatings, car-
bon coating is the most frequently used. In vitro studies [22, 23] have shown
that DLC significantly reduces the release of metal ions and diminishes the
platelet activation, therefore lower rates of acute and subacute thrombosis
Biocompatibility, Cytotoxicity and Bioactivity 57
are expected. Although, recent clinical trials showed no statistical differences

between metal bare and coated stents [24, 25, 26]. While, for coated heart
valves, clinical trials have demonstrated that the most bio-haemocompatible
material is pyrolitic carbon, either as a solid or a coating [27, 28]. On the
other hand, the use of diamond-like carbon for orthopedic implants has been
motivated by the low coefficient of friction, high wear resistance and hardness
of the films. These properties are needed to improve the lifetime of total hip
or knee replacements. The standard design of a hip replacement comprises a
metal or sometimes a ceramic head running against an ultra-high-weight poly-
ethylene acetabular cup. This can last for up to 15 years, which is not enough
for young, active patients, and moreover, early failures are usually reported.
The failure is mainly a consequence of generation of polyethylene wear debris
resulting from the relative movement of the femoral head and the acetabular
cup. Thus, coatings to prevent such wear have been suggested and DLC is one
of the potential candidates [29, 30]. Minor applications of carbon films in the
biomedical field include coating of polymer-based medical products, such as
catheters, drainage tubes or polymer contact lenses [31]. Not much research
has focused on studying the bone-related cell responses of amorphous carbon
films. The studies of the bone-forming activity have been performed on cell
populations extracted from bone tissue: osteoblasts. Du et al. [32] studied
the morphological behavior of osteoblast on DLC and amorphous carbon ni-
tride deposited on silicon substrates, showing that cells were able to attach,
spread and proliferate on the surfaces without apparent impairment of cell
physiology. However, the ability of carbon surfaces to promote bone growth
or biomineralization has not been studied. The formation and maintenance of
viable bone in close proximity to the surface of biomaterials are essential for
the stability and clinical success of noncemented orthopedic/dental implants.
The osteoblast is the cell type responsible for the deposition of bone within
the interfacial zone between the implant and host tissue. A better knowledge
of the osteoblasts’ function could enhance current understanding of the cellu-
lar/molecular events that occur at the tissue–implant interface, and with this
information one could produce biomaterials engineered to elicit specific re-
sponses, such as enhanced osteoblastic mineral deposition. The bone-forming
activity of osteoblasts entails an initial phase of attachment, proliferation, dif-
ferentiation and extracellular matrix synthesis, and a second phase of bone
matrix mineralization. Mineralization in osteoblasts, cultures in vitro can be
demonstrated in two main ways: detection of calcium and phosphate deposits
in the cell layer or detection of bone-matrix proteins that are known to regu-
late “ex novo” bone formation. For example, bone sialoprotein is specifically
expressed by osteoblasts depositing bone ex novo [33]. Osteoblasts cells are
anchorage-dependant, and their development depends strongly on the ini-
tial cellular attachment [34]. In this paper we investigate cellular adhesion
and subsequent cellular functions: proliferation and deposition of a miner-
alized matrix. The cellular adhesion and proliferation in conjunction with
viability tests constitute the best combination of cytotoxicity–biocompatibil-

ity tests, while the analysis of the production of mineralized matrix on the
implant materials represents the most appropriate bioactivity test for bone-
bonding applications. The bioactivity was monitored by direct observation
of the morphology of the deposited mineralized matrix, as well as the ex-
pression of some biochemical parameters of osteoblastic phenotypes: alkaline
phosphatase (ALP) activity and bone sialoprotein (BSP).
2 Experimental Details
Two sets of experiments were realized to evaluate the biocompatibility and
bioactivity of the a-C coatings. In the first set, we studied the cytotoxicity us-
ing 1 cm2 stainless steel (AISI316L) squares as substrates. From profilometer
measurements the surface roughness was between 0.05–0.1 µm with randomly
distributed scratches. The second set, used to evaluate the bioactivity, was of
stainless steel discs of 15 mm in diameter: the same size as that of the wells
in the culture plates. The surface of the discs was sand-blasted with SiO2
particles to obtain a uniform average roughness of 2 µm. In an earlier study
we confirmed that human osteoblast attachment on amorphous carbon films
was enhanced on substrates with an average surface roughness greater than
1 µm [35].
2.1 Film Deposition
Prior to deposition, the substrates were ultrasonically cleaned in acetone

and isopropanol for 30 min, respectively, and then air-dried. A thin layer of
titanium was deposited as an interface layer to increase the adhesion between
the amorphous carbon film and the stainless steel (SS) substrate. The Ti layer
was deposited using a pulsed magnetron sputtering system using argon as the
precursor gas and a high-purity, 99.99%, Ti target. The a-C were deposited
using a high purity hollow cathode graphite target, 7 cm2 , in a DC magnetron
sputtering system and an argon plasma. The Ti/SS substrates were initially
cleaned by an argon plasma for 10 min. The base pressure prior deposition
was less than 2 · 10−4 Pa and the carbon films were deposited at 4 Pa using
0.4 A and argon flow rate of 20 sccm for 5 min, giving a film thickness of
150 nm. For the characterization of the film properties, a-C films under the
same conditions were deposited on silicon and quartz substrates. All layers
were deposited at room temperature.
2.2 Film Characterization
The films deposited on silicon substrates were used to obtain the thickness us-
ing a DEKTAK profilometer. The optical absorption in the ultraviolet–visible
range was measured for the samples deposited on quartz using an UV–VIS
UNICAM spectrometer in the range 1.5–3.5 eV. Spectroscopic ellipsometry
was also used to study the optical properties and thickness of the films.
The energy range was 1.5–5 eV, and the ellipsometric parameters were ob-
tained using a Jobyn Yvon photoelastic modulated ellipsometer. The optical
constants were obtained by fitting spectroscopic ellipsometry measurements
with a two-oscillator Tauc–Lorentz model [36]. Surface composition was in-
vestigated with X-ray photoelectron spectroscopy using a Thermo Scientific
Multilab and Al Kα-radiation.
2.3 Cell Preparation
Human alveolar bone-derived cells (HABDC) were obtained by a conventional

explant technique [37]. The cells were cultured in 75 cm2 flasks in a standard
culture medium composed of: Dulbecco’s Modified Eagle’s Medium (DMEM),
supplemented with 10% fetal bovine serum (FBS) and antibiotic solution
(Streptoycin 100 mg/ml and penicillin 100 U/ml, Sigma Chem Co.). The cells
were incubated in a 100% humidified environment at 37◦ C in an atmosphere
of 95% air and 5% CO2 .
2.4 Cytotoxicity Test
The cytotoxicity was evaluated on the a-C films, the SS substrate, the Ti coat-
ing and a plastic positive control. All samples were sterilized by autoclave.
The cytotoxicity study include three different tests: adhesion after 24 h, prolif-
eration and viability up to 7 d. The HABDCs were plated at an initial density
of 1 · 104 /well and left to adhere for 3 h. After this time, 600 µl of the culture
medium were added. For the adhesion tests the cells were kept for 24 h and for
the proliferation and viability assays the cells were left in the culture plates
for 1, 3 and 7 d. The experimental and control cultures were treated every
2 d with fresh media. To determine the number of attached cells, we followed
the same procedure in each case; that is, after incubation, the unattached
cells were removed with a phosphate buffered saline solution (PBS) and the
attached cells were fixed with 3.5% paraformaldehyde. The evaluation of the
number of attached cells was performed by staining the cellular membranes
and measuring the optical density at the corresponding color. To perform the
staining, the fixed cells were incubated with 0.1% toluidine blue dye for 3 h.
The dye was extracted with sodium dodecyl sulfate (SDS) and the optical ab-
sorption measured with a microplate enzyme-linked immune assay (ELISA)
reader at 600 nm. The number of cells was then determined by correlating the
absorbance of the experimental samples with a standard correlation curve.
For the proliferation test we followed the same procedure for each incubation
time. On the other hand, for the cell viability the conversion of the yellow
dye (3-[4,5-dimethylthiazolyl-2-y]-2,5-diphenyltetrazolium bromide, MTT) to
blue formazan was used as a marker of cell viability or cell enzymatic activ-
ity (mitochondrial). The cells were plated and incubated as described above.
After each term 10 µl of MTT were added and incubated for 3 h. Then,
the supernatant was removed and 600 µl of dimethyl sulfoxide (DMSO) were
added to each well. After 60 min of slow shaking the absorbance was read at
570 nm which gives a reading directly proportional to the number of viable
cells.
2.5 Bioactivity
Evaluation of the biomineralization consisted of the morphological examina-

tion of the mineralized extracellular matrix and the evaluation of a selection
of proteins involved in the bone-formation process.
2.5.1 Morphological Assay
The samples (a-C, Ti, SS and positive control) were cultured in the same
way as for the proliferation assay until they reached confluence. Then, we
added the mineralization medium: 50 µg/ml of ascorbic acid, 10 mM of
β-glycerophosphate and 100 nM of dexamethasone. These compounds are
known to accelerate the mineralization process. The media was changed every
2 d. Biomineralization was evaluated after 7, 14 and 21 d. Following culti-
vation the samples were prepared for observation in the scanning electron
microscope (SEM) in the following way: cell cultures were fixed with 4%
formaldehyde in 0.1 M phosphate buffer solution (pH 7.3), then dehydrated
in graded ethanol and finally sputter-coated with gold. The microscope was
a CAMBRIDGE-LEICA STEREOSCANN 440 SEM.
2.5.2 Protein Synthesis
Alkaline Phosphatase Activity
In this case, the cells were incubated in the culture medium, i.e., no min-
eralization medium was added, for 3, 7 and 14 d. The alkaline phosphatase
activity was determined in cell lysates, obtained by treatment of the cultures
with 0.1 Triton X100 in PBS. The lysates were analyzed by measuring the
release of p-nitrophenol from p-nitrophenyl phosphate in an alkaline buffer
solution, and the colorimetric determination of the product (p-nitrophenol).
Alkaline phosphatase catalyzes the cleavage of a phosphate group from a va-
riety of compounds, including p-nitrophenyl phosphate, which is colorless.
However, one product, p-nitrophenol, is yellow in basic solutions. The ap-
pearance and intensity of yellow color thus indicates the degree to which the
substrate has been acted upon by the enzyme. Normalizing the results to
the time and the total protein concentration gave us information about the
specific activity of the ALP in each of the surfaces. The total protein con-
tent of the lysates was determined using a commercially available kit (Macro
BCA, Pierce Chemical Co.). The ALP activity is reported as nanomoles of
p-nitrophenol produced per minute normalized to the milligrams of total
protein content (nmol min−1 mg protein−1 ). The samples were produced in
triplicate and the measurements repeated at least three times, giving a total
of nine data values for each incubation period.
Western Blot Analysis of Bone Sialoprotein
Similarly, the HABDCs were plated on surfaces in triplicate at an initial den-

sity of 1 · 104 /well in a culture medium in 24-well culture plates; control cells
were cultured directly on tissue culture polystyrene. The cells were cultured
for 7 and 14 d. Western blots allowed us to determine the molecular weight of
the protein and to measure the relative amounts of the protein present in the
different samples. The proteins were separated by gel electrophoresis. The
proteins were transferred to a sheet of special blotting paper called nitrocel-
lulose. The proteins retained the same pattern of separation they had on the
gel. The blot was incubated with a generic protein (such as milk proteins)
to bind to any remaining sticky places on the nitrocellulose. An antibody
was then added to the solution which is able to bind to its specific protein.
The antibody has an enzyme (e.g., alkaline phosphatase or horseradish per-
oxidase) or dye attached to it that cannot be seen at this time. The location
of the antibody is revealed by incubating it with a colorless substrate that
the attached enzyme converts to a colored product that can be seen and
photographed. The bandwidth is proportional to the quantity of the protein,
so the analysis gives us a qualitative assessment of the protein content. The
relative level of the protein was assessed by measuring the integrated density
of all pixels in each band, excluding the local background and normalizing to
the area of the reference. The results are expressed as intensity per mm2 .
Bone Sialoprotein Inmunofluorescence
For this test, the cells were culture for 14 and 21 d in the culture medium.
After the incubation period, cells were fixed using 4% paraformaldehyde for
15 min. After washing with PBS they were processed for immunofluorescence:
cells were permeabilized with 0.5% Triton X-100 in PBS for 15 min followed
by blocking with 1% ovalbumin in PBS for 30 min. A primary monoclonal
antibody to BSP was used, followed by corresponding Alexa Fluor 488 (green
fluorescence) conjugated goat secondary antibody. The samples were exam-
ined by epifluorescence under a conventional fluorescence microscope.
Fig. 1. (a) Absorption coefficient determined by three different techniques: analy-

sis of reflectance and transmittance spectra, photothermal deflexion spectroscopy
and ellipsometry spectroscopy. (b) Refractive index and extinction coefficient as a
function of energy
3 Results
3.1 Film Properties
Because of the diversity of properties that amorphous carbon films exhibit,

it is very important to perform as complete as possible a characterization of
the type of film used. The results of this evaluation for the present study are
described below. The average surface roughness remained nearly constant at
2 µm after coating the substrate. In a previous paper, we reported the energy
loss spectra of amorphous carbon films [38] deposited by magnetron sputter-
ing and demonstrated that the films are nearly 100% sp2 bonded, although
a quantitative sp2/sp3 fraction could not be obtained since the π peak was
Fig. 2. Visible Raman spectra. Results of the decomposition using two Gaussian
peaks for G and D
too wide, indicating that more than one configuration was present and thus
impeding the area calculation [39]. It is well known that the optical and elec-
trical properties of amorphous carbon depend not only on the quantity of
sp2 bonds but also on their distribution or configuration. The optical mea-
surements showed that the films were highly absorbing for energies above
1 eV, having a Tauc optical gap of about 1 eV. The absorption coefficient
as a function of the energy, as determined by three independent techniques,
is shown in Fig. 1a. The E04 falls below the measured range, and therefore
we only report the√Tauc gap that results from an extrapolation to zero of
the line fitting the α × energy vs. energy spectra in the high energy range.
Typical data of the refractive index and extinction coefficient are shown in
Fig. 1b. The existence of the small gap is in agreement with a small degree of
clustering of the sp2 phase detected by the analysis of the Raman spectrum,
shown in Fig. 2 together with the results of the decomposition of the Raman
spectra into two Gaussians: D and G peaks. The XPS analysis demonstrated
that the surface composition consisted predominantly of carbon, and oxygen
as a surface contaminant (Fig. 3). We checked for the presence of metal atoms
from the substrate or the Ti coating but no signal was found, indicating that
the a-C film completely and uniformly covers the substrate.
According to this characterization we can conclude that the films were uni-
formly deposited on the rough stainless steel substrates, copying the surface
topography. The physical and structural properties, optical gap, refractive
index and bonding characteristics of the films suggest that the films used for
this investigation are graphite-like amorphous carbon films.
Fig. 3. XPS spectra of the sample, general view
3.2 Cytotoxicity
In vitro methods provide a necessary and useful adjunct to “in vivo” studies
for selecting materials as candidates for implants or biomaterials. This study
includes two stages; the first, cytotoxicity tests, is carried out to detect any
toxic effects of the potential materials. The second stage allows a more thor-
ough evaluation of the biocompatibility of the material in relation to its end
use, in our case the biomineralization tests. The cytotoxicity of amorphous
carbon films has been tested with different cell lines, and in general it is cat-
alogued as a biocompatible material. However, considering the differences in
film properties and the importance to test the material using cells specific
for the proposed application of the medical device, we decided to evaluate
the cytotoxicity using primary human osteoblasts on the sputtered a-C films.
Toxicity in vitro is a negative or deleterious effect that induces changes in the
cellular response, such as cell death, reduced cell adhesion and proliferation,
altered cellular morphology, and inhibition of biosynthetic functions. All of
these processes were evaluated. Cell death was not observed during the pe-
riod of evaluation up to 7 d. The cellular adhesion, on the other hand, which
is very important for this specific type of cells, was actually improved by the
films in comparison to either the stainless steel substrate, the Ti coating and
the positive control. This is observed in Fig. 4, in which we plot the number
of attached cells after 24 h on each surface. The a-C had the highest number
of attached cells compared to all other materials (ANOVA p < 0.05). Cell
proliferation was evaluated for periods up to 7 d. Figure 5 shows that the
number of cells increased as a function of the incubation time for all sur-
faces. The number of cells on the a-C films doubled at three d of incubation,
while the doubling time for the Ti films was about 7 d. On the stainless
steel the number of cells remained constant at about 1.2 · 104 even after 7 d.
These results were further confirmed by the cell viability assay (MTT test),
which is very important since the combined proliferation/metabolic test is a
way of providing information about the cell growth and metabolic activity
of the cells. The results of the MTT assay (Fig. 6) are presented as the opti-
cal absorbance at 570 nm. We found high levels of MTT conversion (higher
absorbance, high metabolic activity), compare to the control, for Ti and a-
C films, and this increased with time. This increment reflects the increased
number of cells on the surfaces and therefore confirms that there are a greater
number of both dividing and metabolically active cells. In accordance to the
lower proliferation rate measured for the SS substrate, the enzymatic activity
of the osteoblasts cells on SS is lower. Finally, scanning electron microscopy
allowed the observation of the cell morphology. Osteoblasts with dorsal ruf-
fles close to each other connected by filopodia, attached and spread on the
substrate were observed during the different incubation times (not shown).
The results of the cytotoxicity tests suggested that the amorphous carbon
and titanium coatings have no negative affect on cell adhesion, viability and
proliferation. The cell number and absorbance values obtained for a-C and Ti
resulted significantly higher with respect to the positive control and the SS
substrate, indicating the absence of toxic effects. Morphologically, the cells
were in good condition and well attached to the surfaces.
3.3 Bioactivity
3.3.1 Morphological Analysis
Secondary electron image examination at 14 and 21 d of cultured osteoblasts
on the surfaces revealed a continuous cell multilayer and an extracellular
matrix that completely covered the surface. Figure 7 shows the images at
different magnifications for the a-C surfaces at 14 and 21 d. Figure 8 shows
corresponding images for SS and Ti coatings. The extracellular matrix was
composed of both noncollagenous and collagenous components, with small
(< 1µm) spherical structures or mineral deposits. These mineral deposits
were seen interspersed among the osteoblast cell layer or within the fibrillar
matrix. The images are similar to those reported in other works [40, 41, 42],
although the incubation periods used in our work are relatively shorter. By
comparing Figs. 7 and 8 we can observe that similar results were obtained
for the SS substrate, a-C and Ti coatings. Therefore, by simple morpholog-
ical analysis of the mineralization process it is impossible to determine any
substrate-dependent effect, emphasizing the importance of making quantita-
tive measurements, such as those obtained by the analysis of specific proteins.
3.3.2 Proteins
Chemical reactions in living organisms occur rapidly at moderate tempera-
tures and under mild conditions primarily because of the catalytic action of
Fig. 4. Number of attached cells after 24 h of incubation for the SS substrate, a-C
and Ti coatings and the plastic culture dish
a-C
24000
SS
Ti
Control
Number of Cells
20000
16000
12000
8000
0 1 2 3 4 5 6 7
Days
Fig. 5. Osteoblast proliferation for 1, 3 and 7 d of incubation. The initial plated
cells was 1 · 104 per surface
specialized proteins called enzymes. Each step in the chain of a biochemical

reaction is usually catalyzed by a specific enzyme. A large number of proteins
are involved in preparing the matrix for mineralization: phosphatases that
regulate extracellular phosphate concentrations, kinases that regulate matrix
protein phosphorylation and metalloproteinases that degrade the extracellu-
lar matrix. In order to mineralize, osteoblasts express these compounds into
the extracellular matrix, specifically alkaline phosphatase, bone sialoprotein,
osteocalcin and osteopontin. The rate of an enzyme-catalyzed reaction may
be measured by (a) the disappearance of substrate, which is the substance
acted upon by the enzyme and changed to the product, or (b) the appear-
2.0
a-C
1.8 SS
Ti
Absorbance at 570 nm
1.6 Control
1.4
1.2
1.0
0.8
0.6
3 4 5 6 7
Days
Fig. 6. Cell viability (MTT test) expressed as the absorbance at 570 nm
ance of the product. The ALP activity was measured by the product quan-
tity as described above. The levels of this enzyme in the various surfaces
are shown in Fig. 9, as a function of the incubation time. ALP plays a vital
but yet undefined role in bone mineralization. It is thought to regulate phos-
phate transport. The enzyme is glycosylated and attached to the cytoplasmic
membrane on its external surface, where it interacts with extracellular ma-
trix proteins. The results showed that all the cultures presented a significant
induction of the ALP activity, which is in some way expected since we use os-
teoblasts, which produce high ALP concentrations. However, the activity was
significantly higher for the amorphous carbon coatings compared to the other
surfaces (ANOVA p < 0.05). An increase in the activity and expression of
ALP is a strong marker of both the osteoblast phenotype and mineralization.
The other bone-related protein investigated in this work was bone sialo-
protein (BSP), a phosphorylated glycoprotein. BSP is a major structural pro-
tein of the bone matrix that is specifically expressed by fully differentiated
osteoblasts. The expression of BSP is normally restricted to mineralized con-
nective tissue of bones and teeth, where it has been associated with mineral
crystal formation and is considered a potent nucleator of hydroxyapatite.
Western blot analysis was performed to determine the expression of bone
sialoprotein on the surfaces. Semiquantitative information about the level
of BSP expression was obtained by densitometric studies, and both images
are shown in Fig. 10. BSP was expressed on all surfaces after 7 and 14 d
of culture. For a-C, SS and Ti we observed a progressive increase in BSP
expression with time, while in the tissue plastic control, the level decreased.
Remarkably, the level of BSP expression was much higher on a-C than in any
other surface.
Because of the difficulty in analyzing the samples to obtain the compo-
sition of the mineralized matrix, particularly of the small spherical nodules,
Fig. 7. SEM images of the mineralized matrix in the amorphous carbon surfaces
after A, B 14 d. C and D 21 d. Bar: A 2µm, B 10µm, C 2µm, D 30µm
Fig. 8. SEM images of the mineralized matrix in the SS substrate after A 14 d, B

21 d. SEM images of the mineralized matrix in the Ti coatings after C 14 d and D
21 d. Bar: A, 10 µm; B, 2 µm; C, 30 µm; D, 2 µm
Fig. 9. Alkaline phosphatase (ALP) activity
Fig. 10. Western blot analysis of bone sialoprotein (top) and densitometric analysis
of the bands, showing the intensity per µm2 (bottom)
we determined the distribution of bone sialoprotein on the surfaces by im-

munofluorescence labeling. This technique allowed us to confirm that the
small nodules shown in the SEM images corresponded to mineral deposits
containing high concentrations of BSP. Figure 11 shows the different images
of the BSP distribution on the a-C coating, and similar images were obtained
for the other surfaces (not shown). An important proportion of the cells were
immunoreactive to BSP for all the time periods (14 and 21 d). BSP was
present in both the cell layer and the extracellular matrix. Active osteoblasts
contain BSP in a large juxtanuclear mass reminescent of the Golgi region, so
the underlying cellular layer was clearly seen. However, fluorescence intensity
was stronger in the spherical nodules that appeared at 14 d of incubation
and grew into big mineral deposits, forming large mineralized layers, as those
observed in Figs. 9d and 7a by SEM.
4 Discussion
The present study describes an in vitro model to study the response of human
osteoblasts to well-characterized a-C coatings in comparison with titanium
coatings and the stainless steel substrates. We performed both short- and
medium-term studies to determine the cytotoxicity and the biomineraliza-
tion of the different surfaces. We observed an enhanced cellular adhesion and
proliferation/viability on the a-C coatings, demonstrating the biocompatibil-
ity of the material. Moreover, we demonstrated the formation of a calcified
matrix with bone sialoprotein and alkaline phosphatase being expressed.
By comparison between the different surfaces, we showed that osteoblast
cells respond differently to biomaterial in both short- and medium-term cul-
tures. The SEM examination of the osteoblast cultures in the presence of
β-glycerophosphate, dexamethasone and ascorbate revealed the formation of
a stratified matrix mixed with cell layers and the mineralized nodules. The
ALP activity and Western blot were performed to determine the time de-
velopment and levels of two important proteins, ALP and BSP, associated
with the mineralization process. Our results showed different responses to the
biomaterial, and in both cases the amount of protein expressed on the a-C
surfaces was larger, suggesting that the film properties promote the differen-
tiation and mineralization of human osteoblast.
A good correlation between proliferation and mineralization results was
obtained. It is well established that proliferation stops when the mineraliza-
tion process is initiated [43] and that mineralization initiates once the cells
have formed a certain amount of extracellular matrix. This is clearly reflected
in our results. Figure 4 shows that cell proliferation attained a saturation level
at about 7 d, after which the ALP activity increased dramatically (Fig. 9).
This was important to establish the moment at which the mineralization
medium should be added to the culture, since once the culture medium is
Fig. 11. Distribution of the bone sialoprotein on the amorphous carbon surfaces
after A, B 14 d. C, D 14 d. A, C and B 10X; D 20X
rich in phosphate ions, mineralization will occur, but without the appropri-
ate matrix, the mineralization obtained is dystrophic and does not necessarily
resemble the bone formation in vivo.
Human osteoblast cultures may be regarded as a potential in vitro model
to study biomaterial/bone tissue interactions. However, in order to deter-
mine the differences between biomaterial-induced responses, morphological
analysis of the cell layers grown on the surface is not sufficient. If possible,
structural characterization of the mineral deposits must be performed, but
standard immunofluorescence or staining techniques are more appropriate to
quantitatively compare the response of the osteogenic cells to the different
materials.
In conclusion, we have shown that graphite-like amorphous carbon is a
potential candidate for further research as a bioactive coating to induce bone
formation. Many issues must still be investigated, for example, film proper-
ties, and more important, film–substrate adhesion must be improve to the
point that no failure is possible in long-term applications. Mechanical tests
associated with the torque resistance that the coating should support during
the insertion of the implant into the bone must be performed. Similarly, other
biological tests, such as genotoxicity and in vitro bone growth, need to be
carried out.
References
[1] X. Liu, P. K. Chu, C. Ding: Mat. Sci. and Eng. R 47, 49 (2004) 55
[2] J. M. Hollis, C. Flahiff: Factors affecting bone ingrowth, in D. Wise, D. Tran-
tolo, D. Altobelli, M. Yaszemski, J. Gresser, E. Schwartz (Eds.): Encyclopedic
Handbook of Biomaterials and Bioengineering (Marcel Dekker, New York 1995)
pp. 799–821 55
[3] B. Ratner: J. Biomed. Mater. Res. 7, 837 (1993) 56
[4] A. Buford, T. Goswami: Mater. and Design. 25, 385 (2004) 56
[5] W. R. Lacefield: Bioceramics: Materials Characterization versus In Vivo Be-
haviour (New York Academy of Science, New York 1988) p. 72 56
[6] D. B. Burr, S. Mori, R. D. Boyd, T. C. Sun, L. Lane, J. D. Blaha, J. Parr:
Trans. Orthop. Res. Soc. (USA, Japan, Canada) p. 8 (1991) 56
[7] S. D. Cook, L. P. Patron, S. L. Salkeld, J. J. Klawitter, R. D. Beckenbaug:
Pyrolitic carbon as an implant for small joint replacement, in D. Wise (Ed.):
Biomaterials and Bioengineering Handbook (Marcel Dekker, New York 2000)
p. 383 56
[8] A. Bolz: Applications of thin-film technology in biomedical engineering, in
D. Wise, D. Trantolo, D. Altobelli, M. Yaszemski, J. Gresser, E. Schwartz
(Eds.): Encyclopedic Handbook of Biomaterials and Engineering (Marcel
Drekker, New York 1995) pp. 1287–1330 56
[9] F. Z. Cui, D. J. Li: Surf. Coat. Technol. 131, 481 (2000) 56
[10] R. Hauert: Diam. Relat. Mater. 14, 78 (2003) 56
[11] S. Linder, W. Pinkowski, M. Aepfelbacher: Biomaterials 23, 767 (2002) 56
[12] L. A. Thomson, F. C. Law, N. Rushton, J. Franks: Biomaterials 12, 37 (1991)

56
[13] R. Butter, M. Allen, L. Chandra, A. H. Lettingt, N. Rushton: Diam. Relat.
Mater. 4, 857 (1995) 56
[14] J. Wang, H. Huang, P. Yang, Y. X. Leng, H. Sun, et al.: Biomaterials 25, 3163
(2004) 56
[15] H. J. Steffen, J. Schmidt, A. Gonzalez-Elipe: Surf. Interf. Analysis 29, 386
(2000) 56
[16] S. E. Rodil, R. Olivares, H. Arzate, S. Muhl: Diam. Relat. Mater. 12, 931
(2003) 56
[17] M. I. Jones, I. R. Coll, D. M. Grant, K. G. Parker, T. L. Parker: Diam. Relat.
Mater. 8, 457 (1999) 56
[18] N. Nurdin, P. Francois, Y. Mugnier, J. Krumeich, M. Moret, B. O. Aronsson,
P. Descouts: European Cells and Mater. 5, 17 (2003) 56
[19] J. Y. Chen, L. P. Wang, K. Y. Fu, Y. Leng, Y. X. Leng, et al.: Surf. Coat.
Technol. 156, 289 (2002) 56
[20] A. Dorner-Reisel, C. Schurer, G. Irmer, F. Simon, C. Nishan, E. Muller: Anal.
Bioanal. Chem. 374, 753 (2002) 56
[21] S. C. H. Kwok, J. Wang, P. K. Chu: Diam. Relat. Mater. 14, 78 (2005) 56
[22] K. Gutensohn, C. Beythien, J. Bau, T. Fenner, P. Grewe, et al.: Throm. Reser.
99, 577 (2000) 56
[23] P. D. Maguire, J. A. McLaughlin, T. I. I. Okpalugo, P. Lemoine, P. Papakon-
stantantinou, et al.: Diam. Relat. Mater. 14, 1277 56
[24] F. Airoldi, A. Colombo, D. Tavano, G. Stankovic, S. Klugman, et al.: The
American J. Cardio. 93, 474 (2004) 57
[25] O. Schaefer, C. Lohrmann, J. Winterer, E. Kotter, M. Langer: Clin. Radiol.
59, 1128 (2004) 57
[26] Y. H. Kim, C. W. Lee, M. K. Hong, S. W. Park, S. J. Tahk, et al.: Am. Heart
J. 149, 337 (2005) 57
[27] A. Milano, U. Bortolotti, A. Mazzucco, E. Mossuto, L. Testolin, et al.: J.
Thorac. Cardiovasc. Surg. 103, 267 (1992) 57
[28] J. B. Borman, C. de Riberolles: E. J. Cardio-Thor. Surg. 23, 86 (2003) 57
[29] A. Dorner-Riesel, C. Schurer, E. Muller: Diam. Relat. Mater. 13, 823 (2004)
57
[30] S. Affatato, M. Frigo, A. Toni: J. Biomed. Mat. Res. 3, 221 (2000) 57
[31] Y. Ohgoe, K. H. Hirakuri, K. Tsuchimoto, G. Friedbacher, O. Miyashita: Surf.
Coat. Technol. 184, 263 (2004) 57
[32] C. Du, X. W. Su, F. Z. Cui, X. D. Zhu: Biomaterials 19, 651 (1998) 57
[33] P. Bianco, L. W. Fisher, M. F. Young, J. D. Termine, P. G. Robey: Calc. Tiss.
Inter. 49, 421 (1991) 57
[34] K. Anselme: Biomaterials 21, 687 (2000) 57
[35] S. E. Rodil, C. Ramirez, R. Olivares, H. Arzate: Diam. Relat. Mater. in press
58
[36] G. E. Jellison, F. A. Modine, P. Doshi, A. Rohatgi: Thin Solid Films 313-314,
193 (1998) 59
[37] H. Arzate, M. A. Alvarez-Perez, M. E. A. Mendoza: J. Peridontal Res. 33, 249
(1998) 59
[38] S. E. Rodil, S. Muhl, S. Maca, A. C. Ferrari: Thin Solid Films 433, 119 (2003)
62
[39] A. C. Ferrari, A. Libassi, B. K. Tanner, V. Stolojan, J. Yuan, L. M. Brown,
S. E. Rodil, B. Kleinsorge, J. Robertson: Phys. Rev. B 62, 11089 (2000) 63
[40] E. Parker, A. Shiga, J. E. Davies: in J. Davies (Ed.): Bone Engineering (em
squared, Toronto 1999) pp. 63–77 65
[41] M. J. Coelho, M. H. Fernandes: Biomaterials 21, 1095 (2000) 65
[42] M. W. Squire, J. L. Ricci, R. Bizios: Biomaterials 17, 725 (1996) 65
[43] G. S. Stein, J. B. Lian: End. Rev. 4, 290 (1995) 71
Index
a-C, 55 G band, 63
sp3 /sp2 bonding ratio, 62
haemocompatibility, 56, 57
alkaline phosphatase (ALP), 58–61, 66, hardness, 56, 57
67, 70, 71
mineralization, 57, 60, 64–67, 71
bioactivity, 55, 58, 60, 65, 73 MTT test, 65, 67
biocompatibility, 55–58, 64, 71
biomaterials, 57, 64, 73 optical properties, 59
biomedical implants, 55–57, 64, 73
bone sialoprotein (BSP), 57, 58, 61, 66, proliferation, 57, 59, 64, 65, 71
67, 70–72
Raman spectroscopy, 63
coatings, 56–58, 67, 71–73
refractive index, 63
bioactive coatings, 55
biomedical coatings, 57
scanning electron microscopy (SEM),
cytotoxicity, 58, 59, 64, 65, 71
60, 65, 68, 69, 71
D band, 63
diamond-like carbon (DLC), 56, 57 tribological properties, 56
electrical properties, 63 viability, 58, 59, 65, 71
friction coefficient, 56, 57 wear resistance, 56, 57

Characterisation of the Growth Mechanism
during PECVD of Multiwalled Carbon
Nanotubes
Martin S. Bell1 , Rodrigo G. Lacerda2, Kenneth B.K. Teo1 , and

William I. Milne1
1
Engineering Department, University of Cambridge, Cambridge CB2 1PZ, UK
msb39@eng.cam.ac.uk
2
Universidade Federal de Minas Gerais, Departamento de Fı́sica, BR-30123-970
Belo Horizonte, MG, Brasil
Abstract. The growth mechanism of multiwalled carbon nanotubes has been a

subject of considerable research interest. Published results show that nanotubes
may be formed by a variety of methods using different combinations of carbon-
rich gases and etchant gases. We present an overview of different forms of carbon,
showing how nanotubes relate to other carbon structures and how they may be
produced. The use of plasma-enhanced chemical vapour deposition (PECVD) for
growing vertically aligned nanotubes for electronic device applications is reported,
including an analysis of the species present in the plasma during PECVD. It is
shown that the presence of a reactive species such as ammonia suppresses decom-
position of carbon feedstock gas and favours the formation of carbon nanotubes. It
is further shown that gas-phase reactions remove excess carbon, allowing the pro-
duction of nanotubes without unwanted amorphous carbon deposits, which is es-
sential for electronic applications. Plasma analysis is used to determine an optimum
carbon-source to etchant gas ratio, which is verified by postdeposition analysis.
1 Introduction
Carbon is a fascinating material due to the ability of carbon atoms to bond
with each other in different ways, giving rise to materials as varied as dia-
mond and graphite. Carbon has six electrons, arranged as (1s)2 (2s)2 (2p)2 .
It forms covalent bonds with other carbon atoms, using the outer four elec-
trons. To form these bonds, one of the 2s electrons is promoted to a 2p level.
The hybridisation of these electron orbitals allows carbon to form different
structures.
In diamond, sp3 hybridisation produces four identical and very strong
bonds. Each atom bonds to four others, giving rise to a tetrahedral structure,
symmetric in three dimensions.
In graphite, the remaining 2s electron hybridises with two of the 2p elec-
trons, giving three sp2 orbitals, which are located in an (x, y) plane, separated
by 120◦ . The remaining 2p electron orbital is perpendicular to this plane,
hence it is known as a pz orbital. The sp2 orbitals form strong σ-bonds be-
78 Martin S. Bell et al.
A
B
Fig. 1. AB stacking in graphite
Fig. 2. C60 molecule
tween the carbon atoms in the plane. Graphite consists of a series of these
parallel “graphene” planes.
The pz (or π) orbital perpendicular to the plane provides a weak van der
Waals bond that binds the planes together. The weakness of these π-bonds
allows the graphene planes to be easily separated. In high-quality “Bernal”
graphite, the graphene sheets are stacked ABAB. . . (Fig. 1).
These graphene sheets have an interlayer spacing of approximately
0.334 nm. The presence of defects in the graphene sheets gives rise to a larger
interlayer spacing (≈ 0.35 nm), and to a random rotation of the graphene
layers with respect to one another. This less-crystalline form is known as
turbostratic graphite.
A third structure observed in carbon is C60 , also known as “buckmin-
sterfullerene” after the architect Buckminster Fuller, who is famous for his
geodesic dome designs. The C60 molecule is an icosahedral structure made
up of 20 hexagons and 12 pentagons (Fig. 2). Each carbon atom is joined to
Growth Mechanism During PECVD of Multiwalled Carbon Nanotubes 79
a2
a1
Fig. 3. Unit vectors in the graphene plane
three others, so it is predominantly σ-bonded, like graphite. There are similar

molecules containing more carbon atoms (e.g., C70 ).
The discovery of C60 in 1985 [1] may be identified as the first key step
along the path which led to the synthesis of carbon nanotubes. Carbon nano-
tubes themselves were identified in 1991 [2].
A single-walled nanotube is simply a graphene sheet rolled up into a
cylinder. The dangling bonds at the extremities of the sheet join together,
completing the cylinder. The size of the nanotube and its electrical properties
are determined by the “chiral vector”, which links the two points in the
graphene sheet that join up to complete the nanotube. The chiral vector is
defined as c = ma1 + na2 , where a1 and a2 are the unit vectors in the
graphene plane (Fig. 3).
The nanotube, of course, requires not just sides but a cap at the end
to eliminate the dangling bonds. The cap is an icosahedral structure made
up of pentagons and hexagons, like a C60 molecule. A nanotube is classified
according to its chiral vector, and referred to as an (m, n) tube. There are
two special cases that exhibit particularly high symmetry. These are known
as “armchair” and “zig-zag” nanotubes. For an armchair nanotube, n = m,
whilst for a zig-zag tube, n = 0. In these cases, the chiral vector is perpen-
dicular to the nanotube axis (Fig. 4). For the two specific nanotubes shown,
the cap in each case is half of a C60 molecule.
More generally, if n = 0, and n = m, a “chiral” nanotube is formed, where
the hexagons are arranged helically around the nanotube. Single-walled nano-
tubes typically have diameters in the range 1–5 nm. They may be metallic
(0 eV band gap) or semiconducting (typically 0.4–0.7 eV band gap) depend-
ing upon the chiral vector. The band gap is also inversely proportional to the
diameter [3, 4, 5].
A multiwalled nanotube consists, in essence, of a series of concentric single
walled nanotubes. As discussed above, the interlayer spacing in well-ordered
ABAB graphite is 0.334 nm. For the ABAB structure to be replicated in a
multiwalled nanotube would require successive nanotube circumferences to
Fig. 4. “Armchair (5,5)” and “zig-zag (9,0)” nanotubes (top and bottom, respec-
tively)
differ by an integer multiple of (2π × 0.334 nm), as well as requiring well-

ordered individual nanotubes.
Measurements have shown that the intershell spacing in multiwalled nano-
tubes can range from 0.34 to 0.39 nm, with the spacing decreasing as nano-
tube diameter increases [3, 6]. The geometrical constraints in forming the
seamless graphene cylinders cause the layers to be uncorrelated with respect
to one another. Multiwalled nanotubes therefore resemble the less-ordered
turbostratic form of graphite rather than the higher-quality Bernal form.
Multiwalled nanotubes typically have outer diameters in the range 50–
100 nm. Electrical conduction can take place within each of the individual
shells of a multiwalled nanotube. However, if a multiwalled nanotube is con-
tacted on the outside, the electric current is conducted through its outermost
shell only [7]. Where conduction occurs through this outer shell, the band
gap approaches 0 eV due to the relatively large diameter and the nanotube
acts as a metallic conductor.
Before we go any further, it is worth mentioning a related structure,
namely carbon nanofibres. Carbon nanofibres are another kind of graphitic fil-
ament, which differ from nanotubes in the orientation of the graphene planes.
The graphene planes in a carbon nanotube are parallel to the nanotube axis;
in a nanofibre, the graphene layers are not parallel to the axis, but are rather
arranged in a stacked form (layers perpendicular to the fibre axis) or herring-
bone form (layers at an angle to the axis). These nanofibres are illustrated
in Fig. 5, with the nanofibre axis indicated by an arrow in each case.
Nanofibres have many useful properties in common with carbon nano-
tubes and share some potential applications. They do not, however, share all
of the beneficial properties of nanotubes that derive from their structure.
When carbon nanotubes were discovered, it was realised that because
the graphene planes are parallel to the filament axis, they would inherit
Fig. 5. “Stacked” and “herringbone” nanofibres (top and bottom, respectively)
several important properties of graphite. In particular, a nanotube exhibits

high electrical conductivity, thermal conductivity and mechanical strength
along its axis. As there are very few open edges and dangling bonds in the
structure, nanotubes are also very inert, and species tend to be physically
adsorbed onto the graphene walls rather than react with them. As carbon
nanotubes are completely covalently bonded, they can be excellent electrical
conductors which do not suffer from electromigration, a limiting factor for
metals.
Carbon nanotubes have been investigated for a wide range of applications.
A nanotube can behave as a high-aspect-ratio electrical conductor with mi-
cron-scale length and nanometre-scale diameter. Structures of this type are
highly desirable as field emission tips for applications such as field emission
displays [8, 9], X-ray tubes [10], electron sources for microscopy and lithogra-
phy [11], gas discharge tubes [12] and vacuum microwave amplifiers. The high
aspect ratio and small diameter of the nanotube is also useful for scanning
probe tips [13, 14].
Semiconducting single-wall carbon nanotubes have been investigated for
use as transistors or logic elements [15, 16]. The electronic properties of the
carbon nanotubes are highly sensitive to adsorbed species [17, 18], which is a
challenge for logic circuits, requiring the nanotubes to be suitably encapsu-
lated. However, this sensitivity may be utilized in chemical or biological sen-
sors for gas detection. The coherent nature of electron transport in crystalline
nanotubes makes these structures potentially suitable for spin-electronic de-
vices [19].
Carbon nanotubes may also be used as electromechanical sensors as their
electrical characteristics change in response to mechanical deformation of
their structure [20]. Conversely, nanotubes can mechanically deflect under
electric stimulation (e.g., due to charge induced on the nanotubes) making
them useful as cantilevers or actuators [21, 22].
Aside from applications which exploit the properties of individual nano-
tubes, there are a further range of applications which utilise the large surface
area to volume ratio of carbon nanotubes and nanofibres when manufactured

in bulk. One interesting application for these structures is as electrodes in
electrochemical supercapacitors [23, 24]. It has been shown that such struc-
tures could lead to higher charge storage than conventional capacitors and
batteries [21, 25]. The high electrical conductivity and relative inertness of
nanotubes also make them potentially useful as electrodes in electrochemical
reactions [26, 27]. Nanotubes can support reactant particles in catalytic con-
version reactions [28, 29]. It has been proposed that hydrogen could be stored
amongst and inside nanotubes/nanofibres for fuel cell applications [29, 30],
although results show that the amount of hydrogen stored is not as high as
originally anticipated [31]. The use of nanotubes and nanofibres as filters or
membranes for molecular transport has also been proposed [32].
Nanotubes and nanofibres can be added to polymers to improve their
strength and stiffness, and can also add electrical conductivity to polymer-
based composite systems [33, 34], which has proved useful in the automobile
industry for the electrostatic painting of components. Carbon nanotubes are
ideal reinforcing fibres for composites due to their high aspect ratio and
axial strength [35]. Single-wall carbon nanotubes are preferred to multiwall
nanotubes because the inner layers of the multiwall nanotubes contribute
little under structural loading, and thus would reduce the stiffness for a given
volume of nanotubes [35, 36].
2 Production of Carbon Nanotubes

There are a number of known methods for producing carbon nanotubes.
Between them it is possible to produce nanotubes with differing properties
and in different forms. The most common methods used for the production
of nanotubes are arc discharge [37, 38], laser vaporisation [39] and chemical
vapour deposition.
The first nanotubes were produced in an arc discharge arrangement [2].
In a typical arc discharge apparatus, two graphite electrodes are held a short
distance apart inside a vacuum chamber. The chamber is held at high vac-
uum and a flow of helium is supplied. A DC voltage is applied across the two
electrodes, and a large arc current flows, removing carbon atoms from the
anode in the process. The position of the anode is adjustable so that as the
anode is consumed, the gap between the two electrodes can be held constant.
A carbon deposit forms on the cathode; this deposit contains nanotubes as
well as larger carbon nanoparticles. In a high-quality arc discharge, the de-
posit can contain 70% nanotubes. These nanotubes typically exhibit a good
crystalline structure and are very straight, due to the very high temperatures
(3000◦ C) developed during the arc discharge. However, in order to make use
of these nanotubes in an electronic device, they must be separated from the
cathode, purified and manipulated onto a substrate.
Like arc discharge, laser vaporisation is a short-duration, high-temper-

ature technique for nanotube production. A graphite target is heated to
around 1200◦C in a horizontal furnace whilst an inert gas flows across it.
A high-power laser is then used to vaporise the graphite target. The vapor-
ised material is carried by the gas onto a cooled collector, where it condenses.
As with arc discharge, the material contains a mixture of nanotubes and
nanoparticles, and once again the material must be separated, purified and
manipulated in order to fabricate devices. Laser vaporisation can produce
up to 90% nanotubes, however, unlike arc discharge, its use is limited to
small-scale production.
In both of these cases, a block of solid graphitic carbon is heated to
a very high temperature and carbon atoms are separated from the block,
reassembling either on the cathode in the case of arc discharge or on a cooled
collector in the case of laser vaporisation. It is during this reassembly that
the highly ordered nanotubes are formed.
Catalytic chemical vapour deposition is very different. Instead of begin-
ning with a block of carbon, the feedstock is a hydrocarbon gas, which dissoci-
ates either thermally (thermal CVD) or in the presence of a plasma (plasma-
enhanced CVD). Once again, the dissociated carbon atoms self-assemble into
highly ordered nanotubes; however, in this case the nanotubes form on a
substrate, which may be made of any material which can withstand the de-
position temperature (e.g., silicon or glass). The self-assembly is facilitated
by catalyst nanoparticles deposited on the substrate surface, which seed the
nanotube growth. These catalyst nanoparticles determine both the location
and diameter of the nanotubes formed. This ability to deposit nanotubes
selectively based upon the location of catalyst particles opens the door to
lithography and a range of possibilities for direct deposition of nanotube de-
vices.
Thermal CVD requires high temperatures (800–1000◦C) and has two dis-
advantages. First, the nanotubes produced are not just randomly oriented,
but also they are not straight. Second, the high temperature rules out the
use of many desirable substrate materials (e.g., glass).
Plasma-enhanced CVD (‘PECVD’) is a technique used extensively in the
semiconductor industry to allow CVD at lower temperatures. In this case, en-
ergy in the plasma replaces some of the heat energy, allowing gas dissociation
and nanotube formation to take place at lower temperatures (600–700◦C).
PECVD has a further advantage: the electric field aligns the nanotubes dur-
ing growth [40, 41]. There are a number of different techniques available for
creating the plasma. These include rf-PECVD [42], microwave PECVD [43],
inductively coupled PECVD [26] and DC glow discharge PECVD [44].
For an electronic device, it is essential to be able to make good electrical
contact to the nanotubes. Silicon is often selected as an appropriate substrate.
Also, for an electronic device, it is important that the synthesis of nanotubes
is performed without the deposition of amorphous carbon (“a-C”), which
prevents the formation of nanotubes by poisoning the growth catalyst and

can cause short circuits on the substrate surface. One method that has been
shown to prevent the deposition of a-C is to combine the carbon source (a
hydrocarbon gas) with a hydrogen-rich gas (typically NH3 or H2 ), which
produces reactive species in the plasma that remove any excess carbon.
3 Plasma Composition during PECVD

We have seen that we can form carbon nanotubes using several different
techniques. In the case of arc discharge and laser vaporisation, carbon is
liberated from a graphite source before reassembling in nanotube form. In
the case of CVD, it is not clear if carbon itself is present in the chemical
vapour, or if carbon is extracted from hydrocarbon molecules in a reaction
at the surface of the metal catalyst.
Different conditions for nanotube synthesis yield different results: Under
certain conditions we get well-defined nanotubes; at other conditions we get
amorphous carbon. What is the mechanism that determines this? In order to
answer these questions, we conducted an analysis of the plasma composition
during PECVD of carbon nanotubes [45].
Multiwalled carbon nanotubes were grown using a DC glow discharge
PECVD setup with the substrate located on a resistively heated graphite
stage. It has been reported that C2 H2 is the most efficient carbon feedstock
for the growth of carbon filaments [46], and we selected C2 H2 to be the carbon
feedstock gas for our carbon nanotube growth. NH3 was added to generate
reactive species, which have been shown to help prevent formation of a-C.
The gases were fed into the chamber through a metal pipe that acted as an
anode for the plasma discharge. A DC plasma was generated between this
pipe and the tungsten heaters within the graphite stage. The chamber was
maintained at vacuum using a rotary pump.
Si substrates coated with a thin catalyst film were placed on the graphite
stage and heated to 520◦ C in NH3 . At this temperature, the thin film cat-
alyst agglomerates into particles suitable for seeding nanotube growth. The
DC plasma was then initiated and C2 H2 added into the gas flow as the carbon
feedstock. The temperature was maintained at 650◦C throughout the deposi-
tion, and the chamber pressure was maintained at 5.3 mbar. The plasma DC
bias was maintained at 600 V, with current typically around 100 mA. The
volume flow-rate proportion of C2 H2 in the NH3 :C2 H2 plasma was varied
between 0% and 70% whilst maintaining other parameters at their standard
settings.
Mass spectrometry was performed on species extracted 15 mm from
the graphite stage. Neutral molecules were extracted from the plasma and
analysed using residual gas analysis (RGA). Positive ions were extracted and
analysed using secondary-ion mass spectrometry. An RGA mass spectrum
Fig. 6. RGA and positive ion mass spectra (top and bottom, respectively)
for neutral species and a mass spectrum for positive ion species at 23% C2 H2
are shown in Fig. 6.
The species in the RGA spectrum are consistent with the cracking pat-
terns of C2 H2 (base peak at amu 26) and NH3 (base peak at 17 amu) and also
indicate the presence of H2 , HCN, H2 O, N2 and CO2 [47]. This is consistent
with data reported by other authors [48]. The dominant ions detected were
NH+ + + + +
3 and C2 H2 . Other detected species were NH2 , NH4 , HCN and C2 H .
+
4 Characterisation of the Growth Mechanism

The results obtained by growing carbon nanotubes in this particular config-
uration have previously been reported [49]. It was demonstrated that well-
aligned nanotubes were grown for C2 H2 concentrations between 4–20%, that
at 29% C2 H2 the nanotubes became more obelisk-like, and that by 38% there
are no longer any tubes at all. These results are shown in Fig. 7. It was also
reported that the nanotube growth rate peaked at around 20% C2 H2 content.
This growth condition gave a clean, a-C-free substrate [50].
We first investigated the nature of the carbon precursor for the formation
of nanotubes. We were unable to detect C2 , CH4 or other higher carbon
species even after acquiring data for a prolonged period, and thus concluded
that C2 H2 is the dominant precursor for nanotube formation. This conclusion
is supported by other authors [51], who have detected the presence of C2 H2
Fig. 7. Scanning electron microscope images of carbon nanotubes grown at different

gas ratios
catalyst C xH y
particle C
C C
C C C xH y catalyst
Cx Hy
particle
barrier layer barrier layer
(a) weak interaction (b) strong interaction
Fig. 8. Growth models
in a CH4 /H2 /NH3 plasma yielding nanotubes. For this to be the case, our
growth model must include a reaction at the catalyst surface to extract carbon
from C2 H2 .
The catalyst takes the form of a transition metal (Fe, Ni, Co or Mo), which
may be applied chemically from a solution containing the catalyst [52] or
directly by using techniques such as thermal evaporation, ion beam sputter-
ing [53] or magnetron sputtering [54]. For our experiment we used magnetron
sputtering to give us a well-controlled Ni film.
When heat is applied to the substrate bearing the film, the increased sur-
face mobility of the catalyst atoms causes the film to coalesce into nanoclus-
ters [55]. The thickness of the catalyst film, temperature and time determine
the size of these nanoclusters [49, 56, 57].
There is a further consideration with respect to the catalyst layer. This
is the chemical interaction between the catalyst and substrate when heat is
applied. If there is a reaction, the catalyst material will dissipate, ending its
usefulness for seeding nanotube growth. This is a problem when using Si as
a substrate with transition metals, as they would diffuse into the substrate
at the temperatures used. In order to eliminate this problem, a thin diffusion

barrier of SiO2 was sputtered onto the Si substrate before the metal catalyst
layer was deposited.
Depending upon the strength of the interaction between the catalyst metal
and the barrier layer upon which it is deposited, two different growth models
have been proposed [58]. These two models are illustrated in Fig. 8. In each
case, carbon is extracted from the hydrocarbon feedstock gas and diffuses
through the catalyst particle before taking its place in the forming nanotube.
The growth model when there is weak interaction between the catalyst par-
ticle and the barrier layer is known as “tip growth”, as the catalyst particle
stays at the tip of the nanotube. The model when there is strong interaction
is known as “base growth”, as the catalyst particle remains anchored to the
barrier layer at the base of the nanotube. For nickel on SiO2 , interaction is
weak, indicating tip growth. Indeed, bright catalyst particles can clearly be
seen at the tips of the nanotubes in Fig. 7.
These growth models are widely accepted in the nanotube community,
but there is a puzzle concerning the diffusion of carbon through the catalyst
particle. The temperatures achieved during PECVD are well below the melt-
ing point of the catalyst metals, yet it is thought that the catalyst particle
must be in the liquid phase for this diffusion to take place. An explanation for
this is that the saturation of the metal with carbon atoms causes a significant
decrease in the melting point for the catalyst particles, allowing the particle
to melt and the diffusion to take place [59].
Figure 7 shows that varying the gas ratio produces different structures,
with carbon nanotubes deposited at low C2 H2 ratios and a-C deposited at
high C2 H2 ratios. In order to understand why this is the case, we must ex-
amine the role of NH3 in the plasma. The production of a-C-free nanotubes
requires a controlled deposition of carbon, which can then self-assemble into
an energetically favoured nanotube form. This controlled deposition rate is
achieved through the combination of two reactions: the dissociation of a
carbon-rich gas (in our case C2 H2 ) and the removal of excess carbon that
would otherwise lead to amorphous carbon deposits.
It has been widely reported that atomic hydrogen is the active species
for the removal of excess carbon. Both H2 and NH3 have been reported to
act as sources for atomic hydrogen through electron impact dissociation in
plasma [51, 60]. Generation of H from an NH3 :C2 H2 DC plasma has also
been shown in simulations [61]. It has further been shown that NH3 is a
more effective generator of atomic hydrogen than H2 [51]. NH3 therefore has
a key role in removing any excess carbon through the generation of reactive
species which combine with and carry away carbon atoms.
The efficacy of NH3 in nanotube production has been further attributed
to the preferential decomposition of NH3 suppressing the decomposition of
C2 H2 . This is because the chemical bonds that hold the NH3 molecule to-
gether are weaker than those which hold C2 H2 together. It has also been
Fig. 9. H2 RGA signal for varying gas ratio
Fig. 10. HCN RGA signal for varying gas ratio
suggested that nitrogen is incorporated into catalyst particles, forming ni-

trides and changing the composition of the catalyst surface [62]. This may
also help to explain the diffusion puzzle. We can clearly see the preferential
decomposition of NH3 in our RGA data. A graph of the RGA signal for H2
for varying gas ratio is shown in Fig. 9.
H2 is generated by the decomposition of both NH3 and C2 H2 . To the
left of the figure, where the plasma is predominantly NH3 , the amount of H2
increases as NH3 increases. In this region, the H2 is derived from decompo-
sition of NH3 . To the right of the figure, where the plasma is predominantly
C2 H2 , the amount of H2 increases as C2 H2 increases. In this region, the H2

is derived from the decomposition of C2 H2 .
At high NH3 ratios, NH3 decomposes preferentially over C2 H2 . This
causes the C2 H2 to decompose slowly, generating the small amounts of car-
bon necessary for nanotubes to self-assemble. At high C2 H2 ratios, there
is insufficient NH3 to effectively suppress C2 H2 decomposition, resulting in
high levels of carbon generation and deposition of excess carbon as a-C. NH3
therefore has two key roles in the formation of carbon nanotubes: not only
does it generate atomic hydrogen species to remove any excess carbon, but
it also suppresses the decomposition of C2 H2 , limiting the amount of carbon
generated in the first place.
Our observed minimum in H2 is close to our previously reported peak
in clean nanotube growth rate at around 20% C2 H2 . At this ratio, carbon-
removing species are at a minimum and C2 H2 decomposition is low, giving
rise to controlled deposition of clean nanotubes. Interestingly, ion intensities
peak at around 40% C2 H2 , where a-C is deposited. The optimum condition
for nanotube growth is therefore not the condition of maximum ion inten-
sity. This contrasts with most thin film growth, where the condition with
maximum ionisation is preferred.
Further work [63] has shown that the plasma during the nanotube depo-
sition is sufficiently energetic for endothermic ion-molecule reactions to take
place. In many applications, these reactions can and have been ignored as
they require significant pressures and ion energies. However, in this case the
high pressure and DC bias present the perfect condition for these reactions.
The process for the removal of excess carbon by atomic hydrogen species
should give rise to products including both carbon and hydrogen. The princi-
pal reaction product we detected in our RGA data was HCN. A graph of the
RGA signal for HCN with varying gas ratio is shown in Fig. 10. This shows
that HCN is generated where both C2 H2 and NH3 are in abundance.
The Figure also shows that a purely NH3 plasma does not generate signif-
icant HCN. This confirms that the reactions that remove excess carbon are
predominantly taking place in the gas phase. If there was a strong surface
reaction, we would see HCN arising from reactions between the active species
and the graphite stage, even in the absence of C2 H2 .
5 Conclusion
We have shown that carbon nanotubes have many exciting features and po-
tential applications. Nanotubes can be made using several different tech-
niques, and PECVD is particularly suitable for the fabrication of electronic
devices.
During PECVD, carbon is extracted from C2 H2 in a gas-phase reaction
at the catalyst surface, where it self-assembles into nanotube form. We have
explained why varying the gas ratio during deposition yields different struc-
tures and shown that there is an optimum ratio for clean nanotube growth.
This enables the fabrication of a-C-free devices for electronics applications,
which we believe opens up an enormous range of potential applications.
A vast amount of research effort has been conducted into carbon nano-
tubes, and we have an increasingly clear understanding of their properties
and how these can be manipulated to give desirable devices. Whilst nano-
tubes have made only limited forays into commercial products to date, we
believe that over the next decade the remarkable promise they have shown
in the research arena will be realised in the wider world.
References
[1] H. W. Kroto, J. R. Heath, S. C. O’Brien, R. F. Curl, R. E. Smalley: Nature
318, 162 (1985) 79
[2] S. Ijima: Nature (London) 354, 56 (1991) 79, 82
[3] M. Endo, K. Takeuchi, T. Hiraoka, T. Furuta, T. Kasai, X. Sun, C. H. Kiang,
M. S. Dresselhaus: J. Phys. Chem. Solids 58, 1707 (1997) 79, 80
[4] T. W. Odom, J. L. Huang, P. Kim, C. M. Lieber: Nature 391, 62 (1998) 79
[5] J. W. G. Wildoer, L. C. Venema, A. G. Rinzler, R. E. Smalley, C. Dekker:
Nature 391, 59 (1998) 79
[6] C. H. Kiang, M. Endo, P. M. Ajayan, G. Dresselhaus, M. S. Dresselhaus: Phys.
Rev. Lett. 81, 1869 (1998) 80
[7] A. Bachtold, C. Strunk, J. P. Salvetat, J. M. Bonard, L. Forro, T. Nussbaumer,
C. Schonenberger: Nature 397, 673 (1999) 80
[8] W. B. Choi, D. S. Chung, J. H. Kang, H. Y. Kim, Y. W. Jin, I. T. Han,
Y. H. Lee, J. E. Jung, N. S. Lee, G. S. Park, J. M. Kim: Appl. Phys. Lett. 75,
3129 (1999) 81
[9] D. S. Chung, S. H. Park, H. W. Lee: Appl. Phys. Lett. 80, 4045 (2002) 81
[10] G. Z. Yue, Q. Qiu, B. Gao, Y. Cheng, J. Zhang, H. Shimoda, S. Chang, J. P. Lu,
O. Zhou: Appl. Phys. Lett. 81, 355 (2002) 81
[11] N. deJonge, Y. Lanny, K. Schoots, T. H. Ooesterkamp: Nature 420, 393 (2002)
81
[12] R. Rosen, W. Simendinger, C. Debbault, H. Shimoda, L. Fleming, B. Stoner,
O. Zhou: Appl. Phys. Lett. 76, 1668 (2000) 81
[13] H. J. Dai, J. H. Hafner, A. G. Rinzler, D. T. Colbert, R. E. Smalley: Nature
384, 147 (1996) 81
[14] H. Nishijima, S. Kamo, S. Akita, Y. Nakayama, K. I. Hohmura,
S. H. Yoshimura, K. Takeyasu: Appl. Phys. Lett. 74, 4061 (1999) 81
[15] V. Derycke, R. Martel, J. Appenzeller, P. Avouris: Nano Lett. 1, 453 (2001)
81
[16] A. Bachtold, P. Hadley, T. Nakanishi, C. Dekker: Science 294, 1317 (2001) 81
[17] J. Kong, N. R. Franklin, C. W. Zhou, M. G. Chapline, S. Peng, K. J. Cho,
H. J. Dai: Science 287, 622 (2000) 81
[18] P. G. Collins, K. Bradley, M. Ishigami, A. Zettl: Science 287, 1801 (2000) 81
[19] K. Tsukagoshi, B. W. Alphenaar, H. Ago: Nature 401, 572 (1999) 81
[20] T. W. Tombler, C. W. Zhou, L. Alexseyev, J. Kong, H. J. Dai, L. Lei,

C. S. Jayanthi, M. J. Tang, S. Y. Wu: Nature 405, 769 (2000) 81
[21] R. H. Baughman, A. A. Zakhidov, W. A. de Heer: Science 297, 787 (2002)
81, 82
[22] R. H. Baughman, C. X. Cui, A. A. Zakhidov: Science 284, 1340 (1999) 81
[23] K. H. An, W. S. Kim, Y. S. Park, Y. C. Choi, S. M. Lee, D. C. Chung, D. J. Bae,
S. C. Lim, Y. H. Lee: Adv. Mater. 13, 497 (2001) 82
[24] M. Hughes, M. S. P. Shaffer, A. C. Renouf, C. Singh, G. Z. Chen, J. Fray,
A. H. Windle: Adv. Mater. 14, 382 (2002) 82
[25] E. Frackowiak, F. Beguin: Carbon 40, 1775 (2002) 82
[26] J. Li, R. Stevens, L. Delzeit, H. T. Ng, A. Cassell, J. Han, M. Meyyappan:
Appl. Phys. Lett. 81, 910 (2002) 82, 83
[27] M. A. Guillorn, T. E. McKnight, A. Melechko, V. I. Merkulov, P. F. Britt,
D. W. Austin, D. H. Lowndes, M. L. Simpson: J. Appl. Phys. 91, 3824 (2002)
82
[28] G. L. Che, B. B. Lakshmi, E. R. Fisher, C. R. Martin: Nature 393, 346 (1998)
82
[29] E. S. Steigerwalt, G. A. Deluga, D. E. Cliffel, C. M. Lukehart: J. Phys. Chem.
B 105, 8097 (2001) 82
[30] C. A. Bessel, K. Laubernds, N. M. Rodriguez, R. T. K. Baker: J. Phys. Chem.
B 105, 1115 (2001) 82
[31] M. Hirscher, M. Becher, M. Haluska, U. Dettlaff-Weglikowska, A. Quintel,
G. S. Duesberg, Y. M. Choi, P. Downes, M. Hulman, S. Roth, I. Stepanek,
P. Bernier: Appl. Phys. A-Mater. Sci. Process. 72, 129 (2001) 82
[32] L. Zhang, A. V. Melechko, V. I. Merkulov, M. A. Guillorn, M. L. Simpson,
D. H. Lowndes, M. J. Doktycz: Appl. Phys. Lett. 81, 135 (2002) 82
[33] J. Sandler, P. Werner, M. S. P. Shaffer, V. Demchuk, V. Altstadt, A. H. Windle:
Compos. Pt. A-Appl. Sci. Manuf. 33, 1033 (2002) 82
[34] E. Kymakis, G. A. J. Amaratunga: Appl. Phys. Lett. 80, 112 (2002) 82
[35] P. Calvert: Nature 399, 210 (1999) 82
[36] M. F. Yu, O. Lourie, M. J. Dyer, K. Moloni, T. F. Kelly, R. S. Ruoff: Science
287, 637 (2000) 82
[37] T. W. Ebbesen, P. M. Ajayan: Nature 358, 220 (1992) 82
[38] C. Journet, W. K. Maser, P. Bernier, A. Loiseau, M. L. delaChapelle,
S. Lefrant, P. Deniard, R. Lee, J. E. Fischer: Nature 388, 756 (1997) 82
[39] A. Thess, R. Lee, P. Nikolaev, H. J. Dai, P. Petit, J. Robert, C. H. Xu,
Y. H. Lee, S. G. Kim, A. G. Rinzler, D. T. Colbert, G. E. Scuseria, D. Tomanek,
J. E. Fischer, R. E. Smalley: Science 273, 483 (1996) 82
[40] K. B. T. Teo, M. Chhowalla, G. A. J. Amaratunga, W. I. Milne, G. Pirio,
P. Legagneux, F. Wyczisk, J. Olivier, D. Pribat: J. Vac. Sci. Technol. 20, 116
(2002) 83
[41] L. Delzeit, I. McAninch, B. A. Cruden, D. Hash, B. Chen, J. Han, M. Meyyap-
pan: J. Appl. Phys. 91, 6027 (2002) 83
[42] B. O. Boskovic, V. Stolojan, R. U. A. Khan, S. Haq, S. R. P. Silva: Nature
Mater. 27 Oct 2002 (2002) 83
[43] C. Bower, W. Zhu, S. Jin, O. Zhou: Appl. Phys. Lett. 77, 830 (2000) 83
[44] V. I. Merkulov, D. H. Lowndes, Y. Y. Wei, G. Eres, E. Voelkl: Appl. Phys.
Lett. 76, 3555 (2000) 83
[45] M. S. Bell, R. G. Lacerda, K. B. K. Teo, N. L. Rupesinghe, M. Chhowalla,

G. A. J. Amaratunga, W. I. Milne: Appl. Phys. Lett. 85, 1137 (2004) 84
[46] R. T. K. Baker, P. S. Harris: Chemistry and Physics of Carbon, vol. 14 (Dekker,
New York 1978) p. 83 84
[47] NIST Chemistry WebBook
URL http://webbook.nist.gov/ 85
[48] B. A. Cruden, A. M. Cassell, Q. Ye, M. Meyyappan: J. Appl. Phys. 94, 4070
(2003) 85
[49] M. Chhowalla, K. B. K. Teo, C. Ducati, N. L. Rupesinghe, G. A. J. Ama-
ratunga, A. C. Ferrari, D. Roy, J. Robertson, W. I. Milne: J. Appl. Phys. 90,
5308 (2001) 85, 86
[50] K. B. T. Teo, M. Chhowalla, G. A. J. Amaratunga, W. I. Milne, D. G. Hasko,
G. Pirio, P. Legagneux, F. Wyczisk, D. Pribat: Appl. Phys. Lett. 79, 1534
(2001) 85
[51] Y. S. Woo, D. K. Jeon, I. T. Han, N. S. Lee, J. E. Jung, J. M. Kim: Diam.
and Rel. Mats. 11, 59 (2002) 85, 87
[52] A. M. Cassell, S. Verma, L. Delzeit, M. Meyyappan: Langmuir 17, 266 (2001)
86
[53] L. Delzeit, C. V. Nguyen, R. M. Stevens, J. Han, M. Meyyappan: Nanotech-
nology 13, 280 (2002) 86
[54] Y. C. Choi, Y. H. Lee, B. S. Lee, G. Park, W. B. Choi, S. Lee, J. M. Kim: J.
Vac. Sci. Technol. A 18, 1864 (2000) 86
[55] V. I. Merkulov, D. H. Lowndes, Y. Y. Wei, G. Eres, E. Voelkl: Appl. Phys.
Lett. 75, 1086 (1999) 86
[56] Y. C. Choi, Y. M. Shin, S. C. Lim, D. J. Bae, Y. H. Lee, B. S. Lee, D. Chung:
J. Appl. Phys. 88, 4898 (2000) 86
[57] C. Bower, O. Zhou, W. Zhu, D. J. Werder, S. H. Jin: Appl. Phys. Lett. 77,
2767 (2000) 86
[58] R. T. Baker: Carbon 27, 315 (1989) 87
[59] A. R. Harutyunyan, T. Tokune, E. Mora: Appl. Phys. Lett. 86, 153113 (2005)
87
[60] J. B. O. Caughman, L. R. Baylor, M. A. Guillorn, V. I. Merkulov, D. H. Lown-
des, L. F. Allard: Appl. Phys. Lett. 83, 1207 (2003) 87
[61] D. Hash, D. Bose, T. R. Govindan, M. Meyyappan: J. Appl. Phys. 93, 6284
(2003) 87
[62] M. Jung, K. Y. Eun, J.-K. Lee, Y.-J. Baik, K.-R. Lee, J. W. Park: Diam. and
Rel. Mats. 10, 1235 (2001) 88
[63] D. B. Hash, M. S. Bell, K. B. K. Teo, B. A. Cruden, W. I. Milne, M. Meyyap-
pan: Nanotechnology 16, 925 (2005) 89
Index
a-C, 83 C60 , 78, 79
C70 , 79
arc discharge, 82
catalytic chemical vapour deposition
buckminsterfullerene, 78, 79 (CVD), 83
chemical vapour deposition (CVD), 83 chiral vector, 79

chiral vector, 79 growth mechanism, 85
multiwalled nanotubes (MWNTs),
glow discharge, 84 77, 79, 82, 84
graphene, 78–80 single-walled nanotubes (SWNTs),
79, 81, 82
laser vaporisation, 82
zig-zag nanotubes, 79, 80
mass spectrometry, 84, 85
plasma, 83
nanofibres, 80, 82 plasma-enhanced chemical vapour
herringbone nanofibres, 81 deposition (PECVD), 77, 83, 84,
stacked nanofibres, 81 87, 89
nanotubes, 77–85, 87–89
armchair nanotubes, 79, 80 residual gas analysis (RGA), 84, 85, 88,
chiral nanotubes, 79 89
Correlation Between Local Structure and Film
Properties in Amorphous Carbon Materials
Giovanni Fanchini1,2 and Alberto Tagliaferro2

1
Present address: Ceramics and Material Engineering Dept., Rutgers University,
607 Taylor Road, Piscataway, NJ-08854, USA
2
Dipartimento di Fisica and LAQ Intese, Politecnico di Torino, Via Duca degli
Abruzzi 24, I-10129 Torino, Italy
alberto.tagliaferro@polito.it
Abstract. A discussion of the various amorphous carbon and amorphous hydro-

genated carbon films is presented. A link between sp2 , sp3 and H contents and the
mechanical and optoelectronic properties is established using some characteristic
lengths of the structure defined in a previous paper. These lengths describe the
average size of the sp2 agglomerates, the distance between such agglomerates, the
decay length of π states and the accommodation length of sp2 clusters in the sp3
network.
1 Introduction
Several of the Group IV elements have a crucial role in our life. If silicon
and its semiconductor relatives are at the basis of the modern life, carbon
is the basis of life itself. Moreover, carbon in its various forms and alloys
has a relevant role in the various aspects of modern life, too. Diamond is
coveted because of its appealing appearance and emotional value. Graphite
is used in mechanical applications. Carbon-based fuels are still running the
world, and plastics and polymers are everywhere around us. Even the more
“humble” amorphous forms of carbon have crucial importance in modern
life. For instance, amorphous carbon is a protective layer in the majority of
computer hard disks. Amorphous carbon can be grown at room temperatures
in thin film form, even on curved layers. It can be alloyed with other elements
(H, N, Si) to obtain materials having different properties. The present work
aims at providing a better understanding of the role and perspectives of
the various types of amorphous carbon and hydrogenated amorphous carbon
films. This work is based on the correlation between local structure and film
properties.
2 Electronic States in Carbon

The peculiarity of the carbon atom that determines its multiform impact on
reality is its ability to form either σ- or π-bonds. It is customary to describe
96 Giovanni Fanchini and Alberto Tagliaferro
molecular orbitals as a superposition of atomic orbitals pertaining to neigh-

bouring ions [1]. In this light, σ orbitals arise from the hybridisation of 2s
and 2p carbon atomic orbitals. If the p states (px , py , pz ) of the neighbouring
ions mix with their s states, we have four bonding and four antibonding σ
orbitals and the so-called sp3 hybridisation. This type of hybridisation can
be found, for instance, in diamond crystals.
If the s states mixes with two corresponding p states for each atom we
have three bonding and three antibonding σ orbitals. The remaining p states
hybridise to form a π state. This is the sp2 hybridisation, of which graphite
is a typical example.
Finally, if the s states mix with one p state for each neighbouring atom, we
have two bonding and two antibonding σ orbitals and two additional π states.
This is the sp1 hybridisation and is found, for instance, in acetylene.
The σ and π orbitals have very differently shaped electron clouds. σ or-
bitals have the highest density of the electron cloud along the axis connecting
the two neighbouring nuclei and form a strong, almost always covalent, bond
between the ions. Each π orbital has its electron cloud formed by two lobes
symmetric with respect to a plane containing the axis connecting the two
nuclei and absent on such plane. The π-bonds formed between two nuclei are
generally quite weak. Hence the energy spacing between π-bonding and anti-
bonding states is much lower than that between σ-bonding and antibonding
states. Although this is the situation for a pair of carbon ions, things can
be quite heavily modified if the collection of ions is enlarged. In such a case,
several additional effects come into play due to the delocalisation of π or-
bitals. For instance, if we analyse a benzene ring, we found that all π orbitals
arising form the six carbon ions have merged to form a delocalised orbital
containing all π electrons. This orbital now forms a strong additional bond
between all atoms of the ring. It becomes very costly in term of energy to
break a single π-bond, as it is fused with the others so that the structure is
overall much more stable. π orbitals have other peculiar properties that are
of relevance in organic chemistry and in life itself. For instance, the resonance
plays a relevant role in determining the properties of aromatic and aliphatic
structures. However, we will focus on this and other peculiarities only if they
will become relevant to our discussion.
3 Hybridisation and Local Structure

In the previous paragraph we reviewed the various type of hybridisations.
Here we analyse how the hybridisation affects the local and the overall struc-
tures of carbon materials. We will start our discussion from the viewpoint lo-
cal structure, i.e., of the distribution of neighbouring carbon ions. Obviously,
the spatial orientation of the various bonds is determined by the minimisa-
tion of the free energy of the system. Hence, different medium- and long-range
structures can be found, even if the local structure remains almost the same.
Correlation Between Local Structure and Film Properties 97
sp3 hybridisation leads to a tetrahedral local structure, in which the ref-

erence carbon ion is placed in the centre of the tetrahedron and the four
bonds connect it with the vertexes of the tetrahedron where the neighbour-
ing ions sit. The resulting structure is tridimensional, and by the repetition of
this block the diamond structure is built. The distance between neighbouring
atoms in diamond is 0.154 nm.
sp2 hybridisation leads to three σ-bonds lying in the same plane and form-
ing 120◦ angles between any two of them. The repetition of this block leads
to a bidimensional planar structure based on hexagonal blocks: the graphene
sheet. The distance between neighbouring atoms is 0.142 nm, shorter than
in diamond. This is due to the presence of π orbitals that strengthen the
bond between neighbouring ions. In-plane bonding is hence determined by a
concurrence of σ and π orbitals. We would like to focus on the point that
undistorted sp2 hybridised structures tend to be bidimensional. The out-of-
plane bonding between graphene sheets in graphite is due to weak Van der
Waals forces, and this justifies the prompt delamination of graphite.
sp1 hybridisation leads to two σ bonds lying on a line, which can be
straight or kinked depending on the case [1]. The presence of foreign atoms
can modify this picture. For instance, monovalent H or F become bond ter-
minators and can lead to linear structures (such as polymeric chains), even
for sp2 hybridised ions.
4 The Various Forms of Amorphous Carbon

In the previous paragraphs we have very briefly discussed crystalline struc-
tures and polymers. Now we move on to the field of our interest, i.e., amor-
phous structures, in the attempt to clarify a few items. The misuse of the term
diamond-like carbon is quite common in literature. So we will spend a few
lines to detail and justify the categorisation we make use of. As it is custom-
ary to enlarge the phrase amorphous carbon to encompass hydrogenated films
too, we will extend our classification to such films. In amorphous carbons the
various hybridisations (sp3 , sp2 , sp1 ) are present in different amounts and
in different spatial arrangements. It is customary [2] to disregard the role
of sp1 states. In this general discussion we will stick to this point of view,
but it should be remembered that each sp1 hybridised carbon ion provides
the same amount of π and σ states and hence even limited quantities of sp1
hybridised atoms might have an impact on the properties. An interesting
point on which we will elaborate later is that at a given sp3 /sp2 ratio (i.e.,
a given percentage of sp3 hybridised atoms) we can obtain films having dif-
ferent properties since the spatial arrangement of the sp2 and sp3 hybridised
ions can be different. Although one may naively think that a statistical dis-
tribution in space of hybridisations is possible among the various atoms of a
structure, it is by far not energetically favourable [3]. It was shown quite a
while ago that the system is stabilised if sp2 hybridised ions are clustered [3].
The form of these clusters can be different depending upon the case, ranging
from conglomerates of benzene-like rings to linear chains [2].
Most of the amorphous carbon materials that we will discuss below can
be considered as multiphase systems, in which sp3 - and sp2 -coordinated re-
gions coexist. The fact that sp2 regions are basically bidimensional while
sp3 ones are tridimensional has interesting consequences on the structure
of amorphous carbons. In fact, the sp3 fraction needed to reach percolation
threshold in the sp3 phase (∼ 0.3) is much lower than that needed for per-
colation of the sp2 phase (∼ 0.8 [4]). Hence, in most cases, the sp3 regions
are interconnected in an overconstrained network that hosts sp2 coordinated
clusters. The sp3 network will constitute the (more or less rigid, following
H incorporation in it) backbone of the material, strongly affecting its me-
chanical properties. Only in very special cases (glassy carbon) the material
backbone is made by sp2 network in which sp3 regions are dispersed.
Nonhydrogenated films can be labelled as follows [5]:
– Glassy carbon (GC). They have such a high sp2 content that the various
sp2 regions are made of entangled and interconnected benzene-like rings.
The spare sp3 regions are diluted in the network and have a negligible
effect on the material properties. The fact that sp2 regions are entangled
leads to mechanical properties (hardness, . . . ) different from those of
graphite [2].
– Tetrahedral amorphous carbon (ta-C). ta-C films are highly con-
strained sp3 -rich films having mechanical properties close to those of di-
amond. sp2 regions are mainly formed by sp2 paired up or diluted in the
sp3 matrix.
– Amorphous carbon (a-C). This broad class contains all films that have
properties and characteristics intermediate between the previous ones.
The inclusion of hydrogen affects the structure and the properties of the
films in many ways. Generally speaking, it decreases the local coordination
number [6] and hence makes the structure more “floppy”. The floppiness
or rigidity of the material obviously correlates with the amount of hydrogen
and its spatial distribution. As an sp3 -coordinated region has a higher average
coordination number than an sp2 -coordinated one, a higher hydrogen content
is needed to bring it to floppiness. A role is played by the number and spatial
distribution of the sp2 sites. Benzene-like rings are rigid and can be distorted
to a high cost in energy as a collection of bonds has to be distorted [7].
Basically, all hydrogenated amorphous carbon films can be assigned to
one of the following categories:
– Tetrahedral hydrogenated amorphous carbon (ta-C:H). ta-C:H
have a limited hydrogen (20–25 at %) and high (> 2) sp3 /sp2 ratio
coupled with a small amount of sp2 inclusions [2], which are probably
organised more in chains than in rings. This structure leads to mechan-
ical properties closer to those of diamond than for all other types of
hydrogenated amorphous carbon.
– Diamond-like hydrogenated amorphous carbon (DLHC). DLHC
originally bore the name of diamond-like carbon (DLC), although some
authors still apply this term to all type of amorphous hydrogenated car-
bon, generating quite a deal of confusion in the reader. Compared to
ta-C:H films, these have lower sp3 /sp2 ratios (∼ 1) and similar hydrogen
contents. The larger sp2 content is organised in larger clusters, which
require larger distortions of the sp3 network to be accommodated.
– Graphite-like hydrogenated amorphous carbon (GLHC). Although
the hydrogen content is similar to that of the previous two categories, the
sp3 /sp2 ratio is much lower (< 0.5). The average coordination number is
reduced close to the floppiness limit [5].
– Polymer-like hydrogenated amorphous carbon (PLHC). These
films are hydrogen rich (> 40 at %), and their sp2 phase is mainly or-
ganised in chains. These features lead to a floppy system.
The careful reader has surely noticed that some combination of H, sp3 ,
2
sp values are intermediate between the various classes. Our categorisation
is aimed at usefulness and to have a limited number of classes. Hence it can-
not be exhaustive and does actually leave some shadowed regions. Materials
belonging to those regions have properties and characteristics bridging those
of neighbouring classes.
5 Phase Matching and Its Effects on Materials

If we attempted to compare the medium- and long-range structures of the
amorphous carbon films grown around the world, we would probably find that
all of them are more or less different. This might suggest that the quest for
general rules to help in understanding and maybe predicting film properties
might be without hope. Nevertheless, it has been shown that by developing
a few simple concepts [5] a reasonable understanding of film properties can
be achieved.
Let us review the basics of that model and later expand its consequences
to encompass mechanical properties too. As most of the optoelectronic prop-
erties of amorphous carbon films are determined by π states [2], it is crucial
to focus on those aspects of the structure that affect the spatial and energy
distribution of these states. In particular, it is crucial to analyse the role of
“phase matching”, i.e., the distortions induced in the local bonding at the
boundaries between sp3 and sp2 regions. In order to describe these effects
the following parameters were defined (see Fig. 1) [5]:
Lc is the diameter of a given sp2 cluster, assumed for sake of simplicity to
be planar and circular in shape. However, the general features of the
discussion will hold for chain-like clusters too.
Fig. 1. Characteristic lengths in amorphous carbon films [5]
Lc−c is the distance between the borders of two neighbouring clusters.

Lπ is the distance from the cluster border at which the π wavefunction
becomes negligible. We can roughly estimate this parameter to be of
the order of a few tenths of nanometres [2].
Before the definition of the last parameter, let us hypothesise that we insert an
sp2 cluster in a fully sp3 -coordinated network. The structure of the sp3 region
near the inclusion needs to be modified in order to accommodate the cluster.
If we go far enough, however, no effect of the inclusion on the sp3 structure
will be detected:
Lac is the “accommodation length”, i.e., the maximum distance from the
cluster at which the modification induced in the sp3 network is present.
In most cases these modifications consist in distortions of the bonds,
which can be labelled as structural disorder [2]. For a given cluster size
the actual value of Lac depends on the rigidity of the sp3 network, which
is correlated to the local coordination number [1].
The detailed discussion of the cases that can arise by varying the relative
values of the various lengths is reported in [5]. Here we will briefly discuss
the various form of amorphous carbon in the light of these parameters. As
said in the previous paragraphs, in most cases the backbone of the material
is constituted by the (more or less hydrogenated) interconnected sp3 phase.
The various possibilities are described in Fig. 2:
1. Noninteracting clusters (Fig. 2a). When the sum 2Lac of the accommo-
dation distances of two neighbouring clusters is lower than their distances
Lc−c, the two clusters have neither direct nor indirect interaction. The
sp2 clusters behave almost like an ensemble of independent molecules em-
bedded in a medium instead of dispersed in vacuum. π states are localised
in the clusters, and π and π ∗ bands are made of localised states [5].
2. Overlapping accommodation regions (Fig. 2b). The interaction between
clusters is indirect since the distortions due to phase matching affect the
energetic distribution of π states [8]. π and π ∗ bands are still made of
localised states, but their shape is distorted [5].
Fig. 2. The various local structures (see text) [5]
3. Overlapping π states (Fig. 2d). When the sp2 content approaches the
value for sp2 phase percolation, we reach a condition in which the π states,
whose wavefunction decays outside the clusters, overlap. Part of the π and
π ∗ bands is made of localised states, and part is made of extended (better:
overlapping) ones [5].
4. Percolating sp2 phase (Fig. 2c). When the sp2 regions percolate they
become the interconnected backbone of the material. The sp3 inclusions
will not even affect the mechanical properties. π and π ∗ bands are made
of extended states, and their differences from graphite ones are mainly
due to disorder [5].
In order to better understand the implication of what we have just dis-
cussed, we note that the presence of hydrogen in the sp3 network reduces
its local average coordination number and the backbone rigidity [9]. This
should help in understanding why we attribute PLHC to case 1 above. The
high hydrogen content of the sp3 phase strongly reduces Lac and the distor-
tions related to phase matching at the boundaries between sp2 clusters and
hydrogenated sp3 backbone. For GLHC a similar discussion can be made, al-
though the decrease in the average coordination number has to be attributed
more to the high sp2 amount [10] than to the relatively low H content. How-
ever, the higher sp2 content leads to larger cluster sizes and this affects the
optoelectronic properties (see below). In DLHC films, because of the low
hydrogen content and the relatively high sp3 content, the accommodation
regions highly overlap (case 2). The low sp2 content prevents case 3 from
occurring. In ta-C:H we have a similar situation, although sp2 clusters have
a more likely aliphatic rather than aromatic character [2].
In nonhydrogenated films, the absence of hydrogen increases, at a given
sp3 /sp2 ratio, the local coordination number and Lac with respect to hy-
drogenated films. Both ta-C and a-C can then be fitted in case 3. How-
ever, the large difference in the sp3 content leads to a large difference in the
2Lac /Lc−c ratio and the local coordination number. As a consequence, the sp3
phase of ta-C is much more distorted. Finally, GC can be attributed to case
4 because of its very high sp2 content, which leads to sp2 phase percolation.
6 Optoelectronic and Mechanical Properties
We have already noticed that in amorphous carbon films the optoelectronic

properties are mainly determined by the π states and the mechanical ones
by the backbone (usually the sp3 phase). We will now analyse more in detail
how the peculiarities of the structure of each type of film will reflect on its
properties.
GC. The percolation of the sp2 phase would suggest that the mechan-
ical properties will be close to that of graphite. The fact that the
graphene strips [11] are entangled prevents delamination and gives
a certain, although limited, hardness [2] to the material. The op-
toelectronic properties differ from those of graphite because the
long-range order of graphene sheets and the consequent semimetal-
lic behaviour are absent. Nevertheless, the graphene strips have
macroscopic size so that their optical gap is in the far infrared,
making the material optically black. Because of the presence of π
and π ∗ extended states close in energy [5], the electrical conduc-
tivity is high and is weakly dependent on temperature.
ta-C. The high sp3 content and the absence of hydrogen give rise to a
rigid backbone so that the mechanical properties of this material
are close to those of diamond [12, 13]. Still the optical properties in
the visible range are determined by the highly distorted sp2 inclu-
sions, and the absorption coefficient has a rather slow increasing
trend as photon energy is increased [14]. The high level of distor-
tions also leads to a high spin density [14]. Although the optical
gap can reach high values because of the spread of energy levels
of sp2 inclusions due to distortions, the electrical conductivity is
much higher than in other materials having similar gaps and no
π states.
a-C. The presence of a rich sp2 phase with medium-sized (up to a
few nanometres) sp2 inclusions reduces the distortions of the pre-
vious cases and leads to intermediate properties. The fact that
sp2 clusters are smaller than in GC opens the optical gap [2] and
makes electronic conductivity lower and with a higher activation
energy [15]. The lack of long-range entanglement makes the mate-
rial softer with mechanical properties of no relevance [11, 12]. The
high cluster density leads to a high ESR signal, even in absence of
relevant distortions.
ta-C:H. The presence of H makes the sp3 backbone less overconstrained

(i.e., rigid) than in ta-C. This decreases distortions and degrades
the mechanical properties with respect to ta-C, although they re-
main the highest in the a-C:H class [13, 16]. Less distortions in the
sp2 clusters lead to steeper absorption coefficient trends in the vis-
ible region, even if a higher sp2 content is present [17]. Electronic
properties are improved, and in some cases photoconductivity is
detected [18]. The ESR signal is reduced with respect to ta-C and
DLHC as a consequence of the less distorted sp2 clusters [5].
DLHC. Despite the presence of hydrogen these materials are highly dis-
torted because of the size of the clusters and rigidity of the
sp3 backbone. This leads to the better mechanical properties
among “non-ta-C:H” hydrogenated amorphous carbon films [15].
Distortions also have a crucial impact on optical properties, reduc-
ing the optical gap, and ESR signal, giving rise to a very large spin
density [5]. Electrical conductivity is higher than in ta-C:H [15, 18]
because of the higher sp2 content, which leads to a reduced distance
between sp2 regions, and the greater size and higher distortions of
the sp2 phase that reduce the gap [5, 19].
GLHC. The larger amount of sp2 content compared to the previous case
deteriorates mechanical properties [15]. Despite such sp2 increases,
the optical gap increases because distortions are reduced [2]. The
reduction in distortions greatly reduces the spin density too [5, 20].
PLHC. The relevant hydrogen content leads to floppy materials with me-
chanical properties of no interest [15]. The transparency of the hy-
drogenated sp3 phase and the limited size of sp2 inclusions, mainly
arranged in chains, highly increases the optical gap [2]. These ma-
terials become almost insulating [2, 19] but highly photolumines-
cent [2] because the absence of distortions prevents nonradiative
recombination paths to proliferate [2]. The spin density is orders
of magnitude lower than in the other cases [5].
7 Conclusion
We have discussed the various types of amorphous carbon and the role of the
main structural parameters in determining their properties. The overall pic-
ture was obtained by using a few simple parameters (cluster size, distance be-
tween neighbouring clusters, decay length of π states, accommodation length
of sp2 clusters) previously introduced in [5]. We have established a quali-
tative link between structure and mechanical and optoelectronic properties.
The present paper will, of course, be more useful if a link between deposition
methods and parameters and local structure could be established. However,
as such link is thoroughly discussed in [2], we address the reader to that paper
to find information on this aspect.
References
[1] R. McWeeny: Coulson’s Valence (Oxford University Press, Oxford, UK 1979)
96, 97, 100
[2] J. Robertson: Mater. Sci. Eng. R 37, 129 (2002) 97, 98, 99, 100, 101, 102, 103
[3] J. Robertson, E. P. O’Reilly: Phys. Rev. B 35, 2946 (1987) 97
[4] J. Robertson: Phil. Mag. B 76, 335 (1997) 98
[5] G. Fanchini, S. C. Ray, A. Tagliaferro: Diam. Rel. Mat. 12, 891 (2003) 98,
99, 100, 101, 102, 103
[6] J. C. Philips: J. Non-Cryst. Solids 34, 153 (1979) 98
[7] J. Robertson: Diam. Rel. Mat. 4, 298 (1995) 98
[8] G. Fanchini, A. Tagliaferro: Appl. Phys. Lett. 85, 730 (2004) 100
[9] F. Jansen, J. C. Angus: J. Vac. Sci. Technol. A 6, 1778 (1988) 101
[10] J. Robertson: Phys. Rev. Lett. 68, 220 (1992) 101
[11] J. Robertson: Adv. Phys. 35, 17 (1986) 102
[12] G. M. Pharr, D. L. Callahan, S. D. McAdams, et al.: Appl. Phys. Lett. 68,
779 (1996) 102
[13] A. C. Ferrari, J. Robertson, M. G. Beghi, C. E. Bottani, et al.: Appl. Phys.
Lett. 75, 1893 (1999) 102, 103
[14] K. B. K. Teo, A. C. Ferrari, G. Fanchini, S. E. Rodil, et al.: Diam. Rel. Mat.
11, 1086 (2002) 102
[15] P. Koidl, C. Wagner, B. Dischler, J. Wagner, et al.: Mat. Sci. Forum 52, 41
(1990) 102, 103
[16] M. Weiler, S. Sattel, K. Jung, H. Ehrhardt, et al.: Appl. Phys. Lett. 64, 2797
(1994) 103
[17] S. Sattel, J. Robertson, H. Ehrhardt: J. Appl. Phys. 82, 4566 (1997) 103
[18] A. Ilie: Diam. Rel. Mat. 10, 208 (2001) 103
[19] D. Dasgupta, F. Demichelis, A. Tagliaferro: Phil. Mag. B 63, 1255 (1991) 103
[20] F. Giorgis, A. Tagliaferro, M. Fanciulli: Amorphous Carbon: State of the Art
(World Scientific, Singapore 1998) p. 143 ff 103
Index
π states, 95–97, 99, 100, 102, 103 ta-C:H, 98
σ states, 95–97
a-C, 98 accommodation length, 100, 101, 103
sp2 -bonded clusters, 95, 98–103 amorphous carbon, 95, 98
cluster size, 103 characteristic lengths, 99, 100
distorted clusters, 102, 103
carbon coordination, 98–101
undistorted clusters, 97
cluster–cluster interaction, 100
sp1 hybridisation, 96, 97 clustering, 97
sp2 hybridisation, 96, 97
sp3 backbone, 98–103 diamond, 96, 97
sp3 hybridisation, 96, 97 diamond-like carbon (DLC), 97, 99
sp3 /sp2 bonding ratio, 97–99, 101 disorder, 100, 101
ta-C, 98 distortions, 98–103
electrical conductivity, 102, 103 mechanical properties, 98, 99, 101–103

electron spin resonance (ESR)
ESR signal, 102, 103 optical gap, 102, 103
electronic properties, 95 optical properties, 102
optoelectronic properties, 99, 101–103
floppiness, 98
percolation, 98, 101, 102
glassy carbon, 98 percolation threshold, 98
graphene, 97, 102 photoconductivity, 103
graphite, 96, 97
rigidity, 98, 100
hydrogenated diamond-like carbon
(DLC:H), 99 spin density, 102, 103
hydrogenated graphite-like carbon
(GLC:H), 99 tetrahedral amorphous carbon, 98
hydrogenated polymer-like carbon tetrahedral hydrogenated amorphous
(PLC:H), 99 carbon, 98
Defects in CVD Diamond Films from Their
Response as Nuclear Detectors
Marco Marinelli, Enrico Milani, Aldo Tucciarone, and Gianluca Verona

Rinati
Dipartimento di Ingegneria Meccanica, Università di Roma “Tor Vergata”,

Via del Politecnico 1, I-00133 Roma, Italy
milani@ing.uniroma2.it
Abstract. CVD diamond films can be used to realize nuclear detectors with out-
standing working capability in harsh environments. Since efficient particle detection
requires high drift lengths of the carriers produced by the ionizing particle, the pres-
ence of defects severely limits the performance of these detectors. This is a major
issue because the fabrication technology of CVD diamond is much less advanced
than that of more conventional materials like silicon. The different kinds of defects
in CVD diamond and their influence on the detector response are discussed. The
connections between the microscopic structure of CVD diamond and the priming
(or pumping) effect, which is widely used to increase CVD diamond detector per-
formance, are elucidated. The analysis of the response of CVD diamond-based de-
tectors is used to extract qualitative and quantitative information on the properties
of defects limiting the free movement of charge carriers in the detector (e.g., carrier
type for which the traps are active, activation energies, geometrical distribution in
the film, etc.).
1 Introduction
Diamond’s extreme properties [1] extend over different areas (electrical, me-
chanical, thermal, etc.). As a consequence, applications in many fields could
in principle benefit from the utilization of diamond as the primary material
for device realization. The use of natural samples is almost always not feasi-
ble, because of their cost and lack of standard properties. Diamond crystals
grown by high-pressure, high-temperature methods (HPHT), on the other
hand, generally have small dimensions and contain significant amounts of
impurities. For this reason, the demonstration in 1986 of the possibility to
grow diamond films by chemical vapor deposition (CVD) by exciting a plasma
in a hydrogen/hydrocarbon gas mixture at low pressure (about 100 mbar)
and relatively low temperatures (about 750◦ C) using silicon wafers as the
growth substrate created very high expectations. As is often the case, the ini-
tial enthusiasm cooled down when it was realized that the new technique had
some limitations, especially in terms of reproducibility and quality. Therefore,
while in some areas CVD diamond devices are now commercially available, for
many applications the growth technique is still not adequate. In particular,
electronic applications of CVD diamond are severely limited by the polycrys-
108 Marco Marinelli et al.
tallinity of the grown films and by their impurity content, so that they are
particularly demanding in terms of sample quality (the so-called electronic
grade diamond). The presence of grain boundaries and of traps limits the
mean free path of charge carriers, thus strongly affecting the amplitude and
the homogeneity of the response of CVD diamond-based electronic devices.
We will review here some experimental techniques that can be used to
investigate the type and distribution of electrically active defects in CVD
diamond films, using the response of nuclear particle detectors built using
these films. The information provided by these techniques is reported for
high-quality polycrystalline CVD diamond films grown in our laboratories.
Clearly, other samples will give different results, but the approach that we
present here can be used on any sample.
Special attention will be devoted to the so-called pumping or priming
effect, i.e., the increase in the detector response after preirradiation with ion-
izing radiation, a phenomenon which is characteristic of diamond detectors.
Although the pumping process is reversibile (by annealing or by exposure to
strong visible light sources), the pumped state remains stable over at least
several weeks at room temperature, provided that the sample is not exposed
to strong light sources. Exposure to normal ambient light results in no mea-
surable depumping on this time scale. The proposed experiments will disclose
the physical grounds of the pumping process and discriminate its effect on
electron and hole motion in the detection process. Besides their scientific in-
terest, these techniques could be of practical interest since the knowledge of
defect properties and of the defect distribution could be very useful in under-
standing how changes in the growth parameters (gas composition, substrate
temperature, pregrowth nucleation process, etc.) affect the structure of the
films, thus in principle allowing researchers to obtain more efficient devices.
2 CVD Diamond Nuclear Detectors:

Realization and Physics
A detailed description of the various techniques to grow CVD diamond films
is beyond the scope of this contribution. For short reviews the reader may
refer to refs. [2, 3, 4, 5]. In the following we will give only the basic information
necessary to understand the content of the rest of the article.
In low-temperature CVD growth, carbon is fed in the growth reactor,
generally in the form of hydrocarbons (e.g., CH4 ) strongly diluted in H2
atmosphere. Some sort of energy is then transmitted to the gas mixture
to create a plasma. Under appropriate conditions (gas composition, plasma
temperature and pressure) if a substrate (e.g., a silicon or molybdenum slab)
is exposed to the plasma, a thin film of diamond is obtained. A pretreatment
(such as scratching) of the substrate surface is required to promote diamond
nucleation.
Defects in CVD Diamond Films 109
Fig. 1. SEM image of a cross section of a CVD diamond film showing the columnar
nature of the growth
The standard deposition conditions used for the films grown in the Roma
Tor Vergata University laboratories (to which we will often refer in the follow-
ing) are as follows. The reactor is a NIRIM-type microwave tubular reactor,
suitably modified to increase film quality [6]. The gas mixture is 1% CH4
in H2 at a pressure of about 130 mbar, with a substrate temperature of
about 700–750◦C. The substrates are 0.5 mm thick silicon slabs, pretreated
by scratching. The growth rate is about 0.7 µm/h. The high crystalline qual-
ity of these films is demonstrated by their excellent Raman spectra and very
low luminescence background [7].
The heteroepitaxial nature of CVD diamond growth on silicon substrates,
and the fact that nucleation takes places in the form of randomly oriented
seeds, leads to the growth of polycrystalline films. A peculiar characteristic
of these films is their so-called columnar growth, well-evidenced visually in
Fig. 1, which shows a scanning electron microscope view of a cross section of
a CVD diamond film. As the thickness increases during the growth, some of
the grains stop growing, while others become bigger and bigger, so that grain
size increase from the diamond–substrate interface to the growth surface. On
the growth surface of a CVD diamond film therefore, the grain dimension
is roughly proportional to the film thickness. For films in the 10–200 mm
thickness range surface grain size is typically about 1/4 to 1/3 of film thick-
ness. Obviously all film properties which depend on the presence of grain
boundaries are not uniform along the film cross section.
The realization of a nuclear particle detector from a CVD diamond film

is very simple: it is sufficient to deposit on both sides of the film a metal
contact [8] and apply to these contacts an appropriate voltage so as to create
an electric field inside the film perpendicular to the film surface. Since the
detector has the geometry of a thin slab, a so-called parallel plate detector is
obtained. In the following, “positive” bias (or polarity) will refer to the case
of the contact on the film growth surface being at a higher voltage than the
contact on the Si substrate, the opposite being referred to as “negative” bias.
Typically, a 1 V/µm electric field is applied.
The working principle of the device is as follows. The nuclear particle
creates in the detector a (generally) great number of electron–hole pairs, and
the electric field drives the generated carriers towards the two electrodes,
according to charge sign. In the presence of trapping centers, the electron
and hole will not necessarily reach the electrodes, and we will call lc the total
distance that the electron and hole move apart (clearly, lc ≤ D, D being
the detector thickness). Using Ramo’s theorem [9, 10], each electron–hole
pair created by an ionizing particle induces in the measuring circuit a charge
qc = elc /D, e being the electronic charge. The amplitude of the voltage pulse
measured by the external circuit is proportional to the total collected charge
Qc (i.e., the sum of qc over all electron–hole pairs produced by the ionizing
particle). The ratio η = Qc /Q0 between Qc and the total charge Q0 generated
by the ionizing particle in the detector is the detector’s efficiency, one of the
most widely used parameters to identify the detector performance.
Introducing Le and Lh as the average of the distances le and lh (with
le + lh = lc ) that charge carriers move apart from the generation point, we
obtain:
Le + Lh
η= . (1)
D
The experimentally accessible information is η, since the detector output is
proportional to Qc , but like Le and Lh it depends on the system geometry
through the sample thickness (i.e., interelectrode spacing) and the penetra-
tion depth of the ionizing particles. The physically relevant parameters, re-
lated to the defect distribution, are the mean free distances λe = µe τe E and
λh = µh τh E, where µ is the carrier mobility, τ is the time before the carri-
ers being trapped and E is the applied electric field. The charge collection
distance (CCD) is defined as the sum of the mean free distance of charge
carriers within the material:
δ = λe + λh . (2)
A direct relationship between the experimentally accessible information (η)
and the intrinsic material properties (λe and λh ) can be obtained through
the Hecht theory [11]:
Le + Lh [λ2 e−D/λh (eG/λh − 1) + λ2e e−D/λe (eG/λe − 1)]
η= − h , (3)
D DG
where G is the penetration depth of the particles to be detected. Assuming

λe = λh = δ/2, (3) becomes
δ δ 2 (1 − e−2G/δ )(1 − e2(G−D)/δ )

η= − . (4)
D 4DG
Equations (3) and (4) or their approximate (for λe , λh D) counterpart
η δ/D are often used to get an estimate of δ from efficiency measurement.
However, they are valid only in the case of uniform distribution of the charge
created by the ionizing particle along the penetration path, and homogeneous
trap distribution, which is not the case of polycrystalline CVD diamond. We
will see in a later section how to extend this formula to the general case.
Obviously, the Hecht formula tends to saturate, so that η → 1 for δ → ∞.
A peculiar feature of CVD diamond particle detectors is the increase in
their efficiency after irradiation with ionizing radiation, a process called prim-
ing or pumping. The increased efficiency is believed to be due to a saturation
of deep traps [12], leading to increased carrier mean free paths. Typically,
irradiation with X-rays or β-particles is used to reach the pumped state.
In addition to the increase in efficiency, pumping greatly extends the tem-
perature region in which the detector can be operated without significant
performance degradation [13, 14]. Ion Beam Induced Current measurements
showed that in CVD diamond samples the collection efficiency is not homo-
geneously distributed, and pumping increases the size of the high collection
efficiency regions, making the response spatially more homogeneous [15].
In CVD diamond particle detectors a crucial role is played by their thick-
ness. Because of the columnar nature of CVD diamond growth, the crystal
quality is much worse close to the substrate interface than at the growth
surface. In particular, the CCD has been found to increase with film thick-
ness [8, 16]. CCDs up to about 300 µm have been reported [17] in millimeter-
thick synthetic diamond after removing about half of the sample thickness
from the poor-quality substrate side.
3 Analysis of the Charge Collection Spectrum
3.1 Qualitative Analysis
As reported in the previous section, each particle impinging on the detector

produces a signal connected to the CCD of the detector. Because of inho-
mogeneities in the density of defects, electric field distribution, etc., different
particles will give signals of different amplitudes. The statistical distribution
of the amplitudes of the pulses generated by the incoming radiation is re-
ferred to as the charge collection spectrum (see Fig. 2 for a typical spectrum
of our detectors in the as-grown and pumped states), easily generated using
a multichannel analyzer. Its average value gives the the average amplitude of
Fig. 2. Typical collection spectra of a high-quality CVD diamond detector in the

as-grown (left peak ) and pumped (right peak ) states. Solid lines are Monte Carlo
simulations
the response (that is, the induced charge and therefore the detector’s average
efficiency and CCD), while its width gives the inhomogeneity of the response
(i.e., the energy resolution). The spectrum is reported as a function either of
the channel number in the multichannel analyzer or (when the multichannel
scale has been calibrated through the response of a silicon detector, which
has 100% efficiency) of the efficiency or of the “collected energy”, simply
defined as Ec = ηEd , Ed being the energy deposited by each particle in the
detector. (For example, 241 Am α-particles deposit all their 5.5 MeV energy in
the detector provided the detector thickness exceeds their penetration depth
G 14 µm.) From Fig. 2 it is clear that even very high quality polycrys-
talline CVD diamond detectors cannot be compared to Si-based ones, either
in terms of efficiency (100% for a silicon detector) or especially in terms of
energy resolution (in the case of silicon detectors the FWHM of the charge
collection spectrum is well below 1% of the peak position).
The spectra reported in Fig. 2 show that in this case priming changes not
only the position of the peak, but also its shape, from a right-asymmetric
(positive skewness) to a left-asymmetric (negative skewness) one, with a sort
of cut-off at high efficiency values. This behavior can be related to the dif-
ferent distribution of the two factors limiting the response of CVD diamond
particle detectors, namely in-grain defects and grain boundaries. Because of
the columnar growth, the latter strongly increase in concentration towards
the substrate side of the sample, while the former ones can be considered to be
homogeneously distributed. If we assume that priming only acts on in-grain
defects, we have a qualitative explanation of the spectra [18]. In the as-grown
state the response is limited by the homogeneously distributed in-grain de-
fects (hence the positive skewness), since the low average efficiency implies
that the generated charges do not reach the grain boundary region close to
the substrate. Indeed, the detector used in Fig. 2 is D = 115 µm thick, while
the charges are generated within the penetration depth G 14 µm of the
5.5 MeV α-particles used. From from η δ/D we have therefore δ 6 µm,
so that G + δ D. After priming, the increase in efficiency is due to the
deactivation by priming of the in-grain defects, creating a “defect-free” layer
before the grain boundary region located close to the diamond–silicon inter-
face, which produces the high efficiency cut-off.
A Monte Carlo simulation quantitatively support this view [18], since
both the primed and unprimed charge collection spectra can be perfectly
reproduced considering that only in-grain defects can be deactivated by
priming. In-grain defects are modelled by a homogeneous distribution of
random trapping centers, and grain boundaries by a density distribution
D(x) = A exp(−x/b), where x is the distance from the diamond–substrate
interface normalized to the thickness of the sample, and A, b are parameters.
The MC simulated response curves shown in Fig. 2 are obtained including
in the simulation both in-grain defects and grain boundaries in the as-grown
case, and only grain boundaries, with the same distribution utilized in the
former curve, in the primed case. It is not possible to reproduce the left-
asymmetric primed curve without introducing a grain boundary distribution
strongly peaked at the substrate side as required by the columnar growth.
The negative skewness appears, therefore, to be characteristic of high crys-
talline quality samples where in-grain defects are substantially inactivated by
pumping, leading to high values of η. In the following, the average efficiency
calculated from the charge collection spectrum will be referred to simply as
the efficiency. To effectively separate the contribution of in-grain traps and
grain boundaries, the film thickness must be substantially higher than both
the extension of the grain boundary region (otherwise the grain boundary
defects dominate the in-grain traps even in the as-grown state) and the pene-
tration depth of the radiation (otherwise the carriers would be generated over
the whole sample thickness, and even in the as-grown state would “sample”
in their motion the grain boundary region).
More evidence can indeed be collected varying the detector thickness. For
thin samples, where the grain size is comparable to (or smaller than) the mean
free path due to in-grain defects, only a moderate effect is to be expected from
the pumping process, since grain boundaries are the main obstacle to high
CCD values even in the as-grown state. As the film thickness (i.e., grain size)
increases, however, substantial improvements in the efficiency and CCD of
the detectors can be anticipated on the basis of the model because of the
saturation of in-grain defects.
Figure 3 shows the charge collection efficiency under positive bias of sev-
eral films grown in the same conditions (i.e., same crystal quality) with thick-
ness from about 20 µm (i.e., just above the penetration depth of 5.5 MeV
Fig. 3. Experimental efficiency data in the as-grown and pumped state for positive
bias polarization. The solid lines are the best fits obtained with the extended Hecht
model (see Sect. 3.3)
Fig. 4. Collection spectra for positive and negative bias in the pumped state and
in the as-grown state (inset)
α-particles) to about 160 µm [19]. The curves relative to the as-grown and
pumped states show a markedly different behavior. The initial efficiency in-
crease with thickness in as-grown samples is followed for D > 40 µm by a de-
crease roughly following a 1/D law, while in the pumped state the efficiency
steadily increases up to a thickness of 100 µm, and then slightly decreases.
A quantitative explanation of the data in Fig. 3 will be presented in

next section, but we can understand them qualitatively [19, 20]. For very low
thickness values (D < 40 µm) the microgranularity of the substrate interface
layer prevents significant drift lengths of the carriers, so that the efficiency
is very low. Since priming does not act on grain boundary traps, the as-
grown and pumped η(D) curves are superimposed. As the film thickness
increases, so does the grain dimension, and the free path of carriers sharply
increases until eventually (D > 40 µm) the grain size becomes larger than
the mean free path due to in-grain defects. Now grain boundaries play a
minor role and the effect of pumping, which partly saturates in-grain traps,
becomes increasingly evident. In the normal state the CCD is limited by the
substantially homogeneous distribution of in-grain defects, as demonstrated
by the 1/D decrease law of η: the approximation η δ/D is indeed valid
(δ D), correctly leading to a δ value independent from thickness, with a
value close to 10 µm. In the pumped state, when grain boundaries still play
a role because of the partial deactivation of in-grain traps, the increase in
grain size (i.e., δ) compensates thickness variation in (4), leading to a more
pronounced increase of η. Eventually, of course, the efficiency must reach a
limiting value.
Another procedure that allows us to extract valuable information from
the collection spectrum is to make the measurement both in positive and in
negative polarity. Indeed, the geometry of the detection process may not be
symmetric between the two contacts, e.g., because the ionization produced
in the detector is not uniform, or because of the columnar structure of the
film. As a consequence, the contribution of the two type of carriers (flowing
towards opposite electrodes) to the signal will not be the same. Compari-
son of the collection spectra measured under both positive and negative bias
could give information on the contribution of each type of carrier. Consider
the idealized situation in which λh is comparable to film thickness, λe λh ,
and the penetration depth of the ionizing particles is small (G D). Practi-
cally no signal would be observed under negative bias, since neither electrons
(having negligible mean free path) nor holes (blocked by the negative contact)
would move in the detector, while a strong signal due to holes would be de-
tected under positive bias. Figure 4 reports typical detection spectra obtained
irradiating our CVD diamond detectors with 5.5 MeV α-particles, both in
the as-grown and pumped states for each polarization. In the as-grown state
the efficiency does not change substantially with bias polarity, while in the
pumped state the efficiency is much higher in positive polarity. As discussed
in [21, 22], this implies that in the as-grown state λe ≈ λh , and that λe does
not change with the pumping procedure, while λh strongly increases.
3.2 Quantitative Analysis: The General Model
We pointed out that (3) is only valid when the distribution of the charge
created by the ionizing particle along the penetration path and the trap
distribution are uniform. Because of the columnar growth of polycrystalline

CVD diamond, and the Bragg distribution of the energy released by a ionizing
particle, both these requirements are not fulfilled in our case.
Following [19], we now extend the Hecht theory to take into account
inhomogeneous CVD diamond samples and nonuniform ionization profiles.
We will then apply this model to extract information on the inhomogeneous
defect distribution and on the defect properties from the detector response.
The shape of the electron and hole trap distribution, as a function of the
distance x from the growth surface, is taken as:
1
we,h (x ) = + b · e(x −D)/c , (5)
λe,h
where the first term on the right refers to the in-grain defects (i.e., the in-
trinsic properties of diamond crystals), while the second terms models grain
boundaries (i.e., the columnar nature of the film).
The mean distances covered by charge carriers from the generation points,
lh and le , depend on the distance x of the generation point from the top
electrode and can be defined by considering the probability P that a charge
carrier moves at least through a distance s. In a positive bias configuration:
s s
− w (x +s ) ds − w (x −s ) ds
Ph (x, s) = e 0 h Pe (x , s) = e 0 e (6)
D−x x
lh (x) = Ph (x, s) ds le (x) = Pe (x, s) ds (7)
0 0
in which the integration limits take into account that s cannot exceed the
distance of the charge generation point from the corresponding collection
electrode.
The average drift distances Le and Lh of all the charge carriers generated
by an ionizing particle are calculated by averaging le (x) and lh (x) over the
distribution function u(x) of the generation points:
Lh = lh u(x) , Le = le u(x) . (8)
For analytical treatment, the Bragg ionization distribution u(x) can usu-
ally be very well approximated by the exponential function u(x) = A+B ex/C
with a cut-off at the penetration depth G. In the case of 5.5 MeV α-particles,
for example, the above formula works fine using as parameter values (obtained
by the nuclear physics simulation program SRIM [23]) A = 267.7 keV/µm,
B = 11.17 keV/µm, C = 3.45 µm and G = 13.54µm. After calculating Lh
and Le using (5)–(8), the efficiency η can be finally derived from (1). Obvi-
ously, in the case of a negatively polarized detector, the role of electrons and
holes must be exchanged.
Combining (1), (5)–(8), therefore, the quantitative relationship between

the intrinsic properties λe and λh and the experimentally accessible quantity η
can be derived. Clearly, η depends on a number of intrinsic properties of the
film (λe , λh , b and c) in a very complicated way. Direct inversion of the
equations to extract these parameters from the measured η values is not
possible. Rather, it is possible to measure η as a function of other quantities
such as the penetration depth G of the radiation, or the sample thickness D
(from which λe , λh , b and c do not depend) and to fit the obtained curves
with the above proposed formulae.
3.3 The Use of Detector Thickness
As an example, we will now apply this model to the analysis of the η(D)
curves reported in Fig. 3. Since CVD diamond properties are different in
the as-grown an pumped state, in principle we should consider four fitting
parameters for each of the two η(D) curves. However, we have seen that (at
least for the films used in [18, 19, 20, 21, 22]) in the as-grown state λe ≈ λh ,
and only λh significantly increases after pumping. Moreover, the pumping
procedure does not influence the grain-boundary defects (described through
the b and c parameters in (5)) [18, 19]. Therefore the trap distribution in
CVD diamond can be described with good approximation by four parameters
only. These are b, c, λe whose values are the same both in the as-grown and
pumped state, and the λh value in the pumped state (in the as-grown state
λh is substantially equal to λe ). A reduced number of parameters obviously
puts more constraints on the fit, therefore increasing its reliability.
The simultaneous fit of both the η(D) data sets in Fig. 3 with the theory is
represented by the continuous lines, and the corresponding fitting parameters
are reported in Table 1. The goodness of the fit is an indirect proof that the
assumptions made are substantially correct, so that the charge collection
distance δ = λe + λh increase after pumping from 14 µm to 62 µm is really
due only to the large variation of λh , λe being by hypothesis unaffected by
priming. In the case of negative bias the η(D) values are very similar to
those measured under positive bias in the as-grown state, irrespective of the
sample being in the normal or primed state. In this case, in fact, the increase
in λh with pumping does not strongly affect η because holes can only move
an average distance G/2 (roughly equal to λe ) before being stopped by the
negative contact.
Table 1. Best-fit parameters for the efficiency vs. thickness curves in Fig. 5
λh λe b c
µm µm µm−1 µm−1
As-grown 7 7 1.1 10
Pumped 55 7 1.1 10
Fig. 5. Charge collection distances for the as-grown and pumped state obtained
with the homogeneous Hecht model (solid lines) and with the extended Hecht model
(dashed lines)
An interesting result is the identification of the strongly defective region

near the Si–D interface, due to the columnar growth. We can divide CVD
diamond film in two regions, a “homogeneous” one where the action of uni-
formly distributed in-grain defects predominates on grain boundaries, and an
“inhomogeneous” one, closer to the substrate, where the opposite is true. The
borderline between them depends on the values of the b and c parameters,
and, from (5), corresponds to a distance yB = D − xB from the substrate
interface, where
1 1
+ = 2b · e(xB −D)/c = 2b · e−yB /c . (9)
λh λe
A value yB = 26 µm was obtained from the data reported in Table 1, in the
pumped state. When D = yB grain boundaries are always dominant over in-
grain defects, and no effect of pumping is expected at all. This is confirmed
by measurements performed on films of thickness lower than about 30 µm.
In literature the CCD values for CVD diamond films are often derived
from efficiency measurements using (5), i.e., the homogeneous Hecht model.
Despite being widely used, this procedure is clearly incorrect. In the general
model the δ value is correctly treated as an intrinsic property of the material,
and as such is independent of film thickness. As a consequence, the plot of δ
as a function of the film thickness D is given by a constant value (continuous
lines in Fig. 5). The dependence of η on D is due to the different relevance of
in-grain and grain boundary defects for films of different thickness, according
to (5). The analysis of the η(D) curves by the homogeneous model is totally
different. Since in (4) there would be no adjustable parameters, the η(D)
curves simply cannot be fitted. The only way to obtain δ would require a
point-to-point efficiency data conversion by means of (4), however, obtain-
ing a nonphysical, variable δ value (dashed line in Fig. 5). This value tends
to the correct one, obtained by the inhomogeneous model, only for D high
enough to make the influence of grain-boundaries negligible with respect to
in-grain defects. As in the pumped state in-grain defects are partially filled,
the asymptotic value of the CCD is higher for the pumped state than for the
as-grown one.
3.4 The Use of Penetration Depth
Another way to highlight the different action of in-grain and grain boundary
defects, besides varying the film thickness, would be, of course, to change
the penetration depth G of the detected particles for a given film, and use
the method outlined above to fit the η(G) curves. For such an experiment
to be meaningful, G should be varied in a wide region (for example, 0.1<
G/D < 0.8) so that a variable energy beam source would be necessary (Van
De Graaf generator, tandem accelerator, etc.). Some results can be found
in [24, 25], but to limit ourselves to techniques which only require normal
laboratory equipment, we will here discuss a simpler use of the penetration
depth, which requires a more limited range of values for G and can still
provide useful information. We will no longer be able to reconstruct the trap
distribution, but could quantitatively evaluate the mean free distances λe and
λh .
When the film thickness is significantly higher than both the penetration
depth and the carrier mean free path, the grain boundaries will have no
effect at all on the detector response, which only depends on G, λe and
λh . Varying G, the contribution of one type of carrier (depending on bias
polarity) gradually decreases as the penetration depth of the incident particles
is reduced, so that the η(G) curves for the two polarities depend on the mean
free path of each carrier type in a different way, allowing us to separately
extract the desired information.
To control the α-particle penetration depth in the CVD diamond, two
different approaches have been proposed [22]. The obvious way is to vary the
energy of the incident particles (variable energy method, VEM) by using an
absorbing layer of appropriate thickness. Air can be used as the absorber
for 5.5 MeV α-particles, since their energy is halved by a 2.5 cm layer, and
the source-to-sample distance can therefore be easily controlled. As an alter-
native, the incidence angle θ of the α-particles with respect to the sample
surface can be varied, the system being kept in vacuum to leave the incident
energy of the particles at 5.5 MeV (variable incidence method, VIM). In this
case, G is reduced by a factor cos θ with respect to the normal incidence
value G0 = 13.54 µm. A particle incidence angle in the 0◦ –80◦ range can be
controlled with a beam collimator, corresponding to 0.17 < cos θ < 1. The
beam divergence prevents reaching significantly higher incidence angles.
Fig. 6. Efficiency curves as a function of the penetration depth G for air measure-
ments in positive (square) and negative (circle) bias polarity; (a) in the pumped
state, (b) in the as-grown state. Solid lines are the fit curves
In Figs. 6b and 7b the η(G) curves are reported when the sample is
in the as-grown state using VEM and VIM, respectively. The same curves
are plotted for the pumped state in Figs. 6a and 7a. Although VEM and
VIM techniques lead to qualitatively similar behaviors, at lower penetration
depths (i.e., thicker absorbing layer) the energy released in the sample by the
ionizing radiation is reduced in the VEM case, and the spectrum is shifted to
the lower energy side, partly overlapping with the noise shoulder. The signal-
to-noise ratio decreases, eventually making measurements impossible. In the
VIM measurements the total energy released by the impinging particles is
constant for all the θ values, so that both the shape and the peak position of
the spectra are not greatly affected. Lower penetration depths can be reached
with approximately constant error bars on the efficiency measurements using
VIM.
In the analysis of the curves, (5)–(8) can be simplified because grain
boundaries play no role. Equation (5) becomes we,h = 1/λe,h and (6)–(8) can
Fig. 7. Efficiency curves as a function of the penetration depth G for tilt measure-
ments in positive (square) and negative (circle) bias polarity (a) in the pumped
state, (b) in the as-grown state. Solid lines are the fit curves
Table 2. Transformations of the parameters A, B, C and G used in the fits of

Figs. 6 and 7
Normal incidence VEM VIM

A (keV/µm) A0 = 267.7 A0 A0 G0 /G
B (keV/µm) B0 = 11.17 B0 e[(G0 −G)/C0 ] B0 G0 /G
C (µm) C0 = 3.45 C0 C0 G/G0
G (µm) G0 = 13.54 G (from SRIM) G (from SRIM)
be analytically calculated. For positive bias [22] we have Ph (s) = exp(−s/lh ),

Pe (s) = exp(−s/le ) and
lh (x) = λh (1 − e−(D−x )/λh ), le (x ) = λe (1 − e−x /λe ). (10)
Averaging over all the electron–hole pairs, using the energy density u(x) =
A + B exp(x/C) (with a cut-off at the penetration depth G), the efficiency η
for positive polarization one finds:
Le + Lh λe + λh 1
η= = −
D D D[AG − BC(eG/C − 1)]

e(1/C −1 /λe )/G − 1 −D/λh e
(1/C +1/λh )/G
−1
× BC λe + λh e
1 − C/λe 1 + C /λh

+ A[−λ2e (e−G/λe − 1) + λ2h e−D/λh (eG/λh − 1)] , (11)
while for negative polarization the role of electrons and holes is to be ex-
changed. In the standard case (normal incidence in vacuum) the parameters
A = A0 , B = B0 , C = C0 and G = G0 (Table 2) were calculated by fitting the
SRIM-simulated Bragg energy distribution with the function u(x) described
in the previous section. When the α-particle penetration depth changes, the
function u(x) is modified, leading to suitable changes of the parameters A,
B, C and G as reported in [22] (see Table 2).
The η(G) curves relative to both positive and negative polarization can
be simultaneously fitted by (9) using the proper transformations of A, B,
C and G (Figs. 6 and 7) and the very same values of λe and λh for all
four curves (Table 3) to increase the fitting constraints. Consistent values
of the two fitting parameters are obtained from the two methods, the lower
errors in VIM confirming the superior accuracy of this method as compared
with VEM. The pumping process is much more effective on hole conduction,
since λh becomes much greater than λe after pumping, in contrast with the
situation before pumping, when λe λh .
Table 3. Best-fit parameters for the efficiency vs. penetration depth curves in
Figs. 6 and 7
VEM VEM VIM VIM

(as-grown) (pumped) (as-grown) (pumped)
λe (µm) 8.6 ± 1.0 8.3 ± 1.3 8.6 ± 0.6 8.4 ± 1.0
λh (µm) 8.1 ± 1.0 36.1 ± 1.5 6.8 ± 0.6 37.6 ± 1.1
4 Time-Domain Analysis
4.1 The Background
As a direct consequence of Ramo’s theorem, the charge induced by a single

generated carrier in the measuring circuit in a time dt is dq = edx/D =
(ev/D)dt, where v is the drift velocity of the carrier. Therefore the time
evolution of the detector signal reflects the dynamics of carriers generated by
the ionizing radiation in the detector. Carrier dynamics are in turn related
to the nature and concentration of the traps present in the film. That is why
a deep insight on the dynamics of charges generated in the detector and the
identity of defects existing in CVD diamond can be obtained from the time
development of single pulses.
Zanio et al. [26] and Martini and McMath [27] showed how this time
behavior changes when detrapping and/or trapping events occur in the de-
tector, and applied this analysis to conventional Si detectors. This analysis
was then extended [21, 28] to describe the more complex behavior of CVD
diamond detectors.
In the original model of Martini and McMath, a single kind of defect is
considered, i.e., the situation found in Si detectors, and a single carrier type
(the ionizing radiation was a focused γ-ray beam directed close and parallel
to one of the electrodes, so that only one type of carrier was free to move
towards the opposite electrode, contributing to the output signal of the de-
tector according to Ramo’s theorem). In the absence of trapping defects the
total distance traveled by a single electron–hole pair (before being collected
at the electrodes) amounts exactly to D. In this case the detector’s efficiency
is 100%. If deep traps are present in the detector, which capture a fraction
of the carriers and do not allow their thermal release (detrapping), the col-
lected charge and the efficiency are accordingly reduced, but the collection
process is still completed within the transit time. When shallow traps are
involved, 100% efficiency may still be achieved, but the time response is now
slower because carriers experiencing trapping–detrapping have their motion
delayed by a time on the order of the detrapping time constant, so that the
charge collection process extends beyond TR . A so-called slow component
appears, as opposed to the fast one due to the ballistic motion of untrapped
carriers. For a 100 µm thick diamond detector working under an external
field E = 104 V/cm, TR = D/v ≈ D/mE ≈ 1 ns, so that in most cases
the fast component is not experimentally accessible because of the response
time of the instrument used to trace pulse shapes. Charge collection for the
slow component is ruled by the detrapping time constant, which can be much
longer than a few nanoseconds, so that it can be experimentally measured.
For t < TR no charge reaches the electrode, and an analytical solution for
the rate equation exists [27]. Unfortunately, the time evolution of the slow
component cannot be calculated analytically even in this simple case involv-
ing a single type of trapping centers and a single type of carriers generated
Fig. 8. Time evolution of the pulses measured in the as-grown state for positive
and negative polarities
Fig. 9. Time evolution of the pulses measured in the pumped state for positive
and negative polarities
at one electrode. For diamond detectors under α- or β-particle irradiation,

the released charge Q0 is not concentrated at one electrode but is distrib-
uted along the particle penetration depth, ranging from several microns to
the whole sample thickness. Consequently, both electrons and holes can in
principle contribute to charge collection. In addition, we will show in the
following that more than a single type of defect must be introduced to ex-
plain experimental results in the case of CVD diamond detectors. A computer
simulation can be used to quantitatively describe trapping–detrapping effects
once a qualitative picture of the trapping centers present in the detectors is
available.
4.2 Qualitative Analysis
The experimental time evolution of a diamond detector output before pump-

ing, i.e., in the as-grown state, is reported in Fig. 8 for positive and negative
field polarity [21]. The pulses have different amplitudes, reflecting the width
of the charge collection spectrum, but similar amplitudes are observed for
positive and negative polarity. In all cases the saturation value is immedi-
ately reached, the collected charge being constant after the 10 ns rise time
of the digital oscilloscope. This implies that once trapped by a defect, elec-
trons and holes are not detrapped, so that the defects limiting the detector’s
response in the as-grown state can be identified as deep ones. No significant
difference is observed between positive and negative polarity because CCD
(and therefore both electron and hole mean free paths before trapping) is
lower (see below) than the penetration depth G 15 µm of 5.5 MeV α-
particles in diamond. As a result, the growth surface boundary plays a very
limited role. Further, since the film thickness D d, the substrate interface
boundary plays no role at all. Therefore, even if electrons and holes should
have different mean free paths, no difference would be observed between neg-
ative and positive polarity. In principle only one kind of traps for electrons
and holes is necessary to explain the behavior in the as-grown state.
This picture changes when examining the pulse shapes for the same sam-
ple in the pumped state (Fig. 9). No significant effect of pumping is found
in the case of negative polarity. A dramatic change, on the other hand, can
be seen for positive field polarity. Not only is the amplitude of the pulses
greatly enhanced, reflecting the large increase of efficiency due to pumping,
but a significant slow component develops, breaking the symmetry between
positive and negative polarity.
It appears, therefore, that pumping saturates most of the deep defects
responsible for hole trapping without detrapping, leading to a greatly en-
hanced mean free path of holes in the pumped state, while not significantly
affecting electron traps [21, 29]. Thus, shallower defects existing in a lower
concentration than saturated, deeper ones, now become important. Since
they allow detrapping, they do not limit the overall amplitude of the pulse,
but slow down the charge collection process. For negative polarity, the lim-
ited changes in the pulse amplitude and shape with respect to the as-grown
state show that no significant saturation of electron traps occurs. Holes are
in principle substantially free to move, but since 5.5 MeV α-particles only
ionize within their penetration depth G 15 µm (much less than the hole
mean free path), holes can move only a few microns before being collected
at the surface electrode. They therefore cannot give a great contribution to
charge collection according to Ramo’s theorem, nor can they be trapped and
detrapped since their mean free path is now much higher than the real path
to the electrode. Therefore the slow component substantially does not show
up, and at the same time the fast one does not increase significantly with
respect to the as-grown case.
4.3 Quantitative Analysis: Computer Simulation

A computer simulation can be developed [21] to calculate, according to this
model, the pulse shape Q(t) in a discrete time approximation. For each carrier
type, the detector thickness D is divided into N laminae, the transit time
of a carrier in each lamina being ∆t = D/N v = TR /N , where v is the drift
velocity. For each carrier type (electrons and holes) deep (type-A) and shallow
(type-B) defects are taken into account, having trapping time constants τA+ ,
τB+ and detrapping time constants τAD , τBD , respectively. Carriers can be free,
ni (tj ) being the number of free carrier in the i-th lamina at time tj , or
trapped in each of the two trapping centers, nAi (tj ), nBi (tj ) being their
corresponding numbers. In the time interval ∆t = tj − tj−1 , free carriers
either move to the next layer contributing by ev∆t/d = e/N to the induced
charge, or are trapped in either defect. In the same time interval, trapped
carriers can be released and move to the next layer or remain trapped. We
have therefore at first order approximation in ∆t:

1 1
ni (tj ) = ni−1 (tj−1 ) − ni−1 (tj−1 )∆t + (12)
τA+ τB+
∆t ∆t
+ nA(i−1 ) (tj−1 ) D + nB(i−1 ) (tj−1 ) D ,
τA τB
∆t ∆t
nAi (tj ) = nAi (tj−1 ) − nAi (tj−1 ) D + ni (tj−1 ) + , (13)
τA τA
∆t ∆t
nBi (tj ) = nBi (tj−1 ) − nBi (tj−1 ) D + ni (tj−1 ) + , (14)
τB τB
(release and trapping of a carrier in a single time interval ∆t and other
second-order effects can be neglected for sufficiently large N values). The
total charge induced by each type of carrier is then
e
Q(tj ) = Q(tj−1 ) + + ni (ti ). (15)
N i
This procedure is repeated twice (for electrons and holes, respectively), and
the results are summed. Direct recombination effects have been neglected
since electrons and holes rapidly separate under the external electric field, and
the concentration of charges generated along the α-particle track is estimated
to be lower than that of trapping centers. The simulation parameters are the
+ + D D + + D
carriers’ velocity v and the time constants τeA , τeB , τeA , τeB and τhA , τhB , τhA ,
D
τhB for electrons and holes, respectively. However v only defines the value of
∆t, i.e., the time scale of the process. In fact, all time constants only appear
in (12)–(14) through their ratio with ∆t, so that their values are relative to
∆t. A reasonable value v = 5 · 106 cm/s [8] has been chosen, resulting in
TR = d/v = 2 ns.
According to the qualitative discussion reported in the previous section,
+
there are no type-B defects for electrons (i.e., τeB = ∞). Also, type-A defects
Fig. 10. Comparison of measured pulses with the simulation based on the trap-
ping–detrapping model in the as-grown and pumped states
do not allow detrapping (τeA D D

, τhA = ∞). Only τeA + +
, τhA +
and τhB D
, τhB must
therefore be determined. Since the CCD is more easily visualized in terms
of carrier mean free path, the trapping time constants are transformed into
+ + +
the mean free paths λeA = vτeA , λhA = vτhA , λhB = vτhB due to type-A or
type-B defects for electrons and holes, respectively. In contrast to the time
constants, the mean free path values do not depend on the arbitrary choice of
∆t (which cancels out in the product with the corresponding time constant),
and are therefore real ones.
In Fig. 10 simulated pulses are compared with experimental ones for posi-
tive polarity (the negative polarity case is a trivial one, since the slow compo-
nent is always very small). Because of the scaling property of the pulses [21],
a single pulse for the as-grown and pumped states has been selected, cho-
sen in the middle of the efficiency distribution determined by the α-particle
collection spectra (about 8% in the as-grown state and 40% in the pumped
state).
To find the best agreement with the measured pulses, the as-grown, pos-
itive polarity curve alone is simulated first. As discussed above, the lack of a
slow component means that the density of deep (A) traps for holes is much
higher than that of shallow (B) traps, so that the only parameters to be
determined are in this case λeA and λhA , the value of λhB being almost ir-
relevant. When, as in the present case, the efficiency is so low that the total
CCD is lower than the penetration depth of α-particles, the only accessible
information is the sum of electron and hole mean free paths and not their
separate values. Setting λeA = λhA , the as-grown state pulse is simulated
using λeA = λhA 4.5 µm (δ = 9 µm), corresponding to τeA + +
= τhA = 0.09 ns
with the chosen value for v. The pulse measured in the pumped state has
then been simulated, introducing also shallow (detrapping) hole traps (i.e.,
λhB ) and, as already argued in the preceding section, a substantially lower
Table 4. Parameter values used in the simulation of Fig. 10
As-grown Pumped
λeA (µm) 4.5 4.5
λhA (µm) 4.5 50
λhB (µm) 35 35
D
τhB (ns) 200 200
concentration of effective deep hole traps (i.e., a higher value for λhA ) as a
consequence of pumping. The concentrations of deep electron traps (i.e., λeA )
is kept unaltered. In order to reproduce the experimentally observed ampli-
tude ratio between fast and slow components, it turns out that λhA = 50 µm
+
(i.e., τhA = 1 ns), so that the pumping process reduces the density of effective
hole deep trapping centers by one order of magnitude, while λhB = 35µm (i.e.,
+
τhB = 0.7 ns), much higher than the 4.5 µm as-grown value of λhA , which
confirms the validity of the initial assignment λeA = λhA 4.5 µm based
upon neglecting hole type B traps in the simulation of the as-grown pulse.
The parameter values used are summarized in Table 4.
In agreement with the model, in thin films (D < 35 µm) a significant slow
component is not detected even after pumping, since holes are trapped at the
highly defective region close to the film–substrate interface before travelling
distances much in excess of 10 µm.
D
The time evolution of the slow component is well described when τhB
= 200 ns. A rough estimate of the activation energy ED of shallow defects
D
responsible for hole trapping and detrapping can be made since 1/τhB =
12 −1
s exp(−ED /kT ). Using s ≈ 10 s for the attempt frequency s (to which
ED is anyway relatively insensitive), we obtain ED 0.3 eV.
A more quantitative evaluation of ED can be obtained by measuring the
thermally activated detrapping time constant at various temperatures [28].
The slope of the linear dependence of ln(τD ) on 1/T gives then the activation
energy ED . Using the estimated value ED 0.3 eV, we see that a 20% de-
crease in T is expected to increase τ D by a factor of 15. Temperatures ranging
from about −40◦C to about 20◦ C should therefore correspond to values of τ D
in the range 200–3000 ns, well accessible experimentally. This temperature
range is easily covered using Peltier elements. In Fig. 11 the pulse shapes of
the detector in the pumped state and under positive bias are reported, mea-
sured at various temperatures and normalised to unit amplitude for better
comparison. The systematic speed-up of the response with temperature con-
firms that the slow component is due to thermally activated detrapping from
relatively shallow defects. Similar curves are obtained under negative bias.
In the normal state the slow component is strongly reduced, as expected, its
residual being attributed to an incomplete “depumping” process.
In Table 5 the values of ED calculated by fitting the τ D (T ) curves with (1)
are reported in all four cases (positive and negative polarity, as-grown and
Fig. 11. Pulses measured in the pumped state for positive polarity at temperatures
−40◦ C, −30◦ C, −15◦ C, 0◦ C, 20◦ C (from slowest to fastest). The amplitudes are
normalised to unity
Table 5. Values of the activation energy of shallow traps calculated from the
temperature dependence of the detrapping time constant
As-grown+ As-grown− Pumped+ Pumped−

EA (meV) 280 370 340 337
pumped state). A value ED = (0.35 ± 0.02) eV is found, with the exception of

the case of positive polarity in the normal state, probably as a consequence
of an imperfect temperature control.
5 Temperature Effects: Depumping
The method described in the previous section is not suitable to measure the
activation energy of deep traps, since temperatures should be raised beyond
the working range of the detector to have measurable detrapping times. To
circumvent this problem, the α-particle response of a diamond film was re-
cently measured [30,31] after successive annealing steps performed at different
temperatures in the 180–230◦C range. The measurement is therefore carried
out at room temperature, and only the annealing treatment is performed at
temperatures beyond those at which the detector works properly.
If before each annealing curve at a given temperature the sample is driven
to its fully pumped state, the effect of thermal annealing is to promote
thermal detrapping of carriers from occupied trapping centers, i.e., a par-
tial depumping. The increase in the number of active (i.e., not filled) traps
lowers the carrier mean free path before trapping, and the detector efficiency.
Table 6. Detrapping time constants τ for the various annealing temperatures, as

extracted from the fit of curves reported in Figs. 12 and 13
T (◦ C) τ (h)
Deeper trap 180 373
(Fig. 13) 190 159
206 71.1
220 14.7
228 6.9
Shallower trap 125 140
(Fig. 14) 138 48.7
165 4.5
With reference to type-A and type-B traps for holes discussed above, let nA
and nB be their concentrations, respectively. The total number per unit vol-
ume n∗A of active (not filled) type-A defects after an annealing step depends
on the annealing time t as n∗A (t) = nA (1 − e−t/τ ), where τ is the detrapping
time constant.
The collection distance of holes depends therefore on nA and nB as
1
λh (t) = , (16)
σA nA (1 − e−t/τ ) + σB nB
where σA and σB are the capture cross sections of type-A and type-B defects,
respectively. Introducing λh0 and λh∞ as the hole charge collection distances
in the fully pumped and completely depumped state, respectively, we have
λh0 = λh (t = 0) = 1/σB nB , and λh∞ = λh (t = ∞) = 1/(σA nA + σB nB ).
Substituting back in (16), we obtain
λh0
λh (t) = . (17)
1 + (λh0 /λh∞ − 1) · (1 − e−t/τ )
It is therefore possible to calculate the activation energy EA by extracting
τ from the λh (t) dependence and using 1/τ = s exp(−ED /kT ), provided
we calculate λh from the detector efficiency η using, e.g., the VIM or VEM
techniques discussed above. Since we are interested in λh only (pumping
only acts on hole traps), a simplified procedure can be used, consisting in
performing the α-particle irradiation at a single high incidence angle (80◦ )
with respect to the detector’s surface normal [30, 31].
The λh (t) curves were obtained at five different temperatures ranging
from 180◦ C to 228◦ C, as reported in Fig. 12. Equation (17) fits these curves
very well, exception made for the first point in each curve. The fast decrease
of about 5 µm in the value of λh observed in each first annealing step can be
attributed to the existence of other traps with lower activation energy, i.e.,
shorter time constants, that are not resolved in this temperature range and
time scale. To isolate the contribution of the first (deeper) trapping center, the
Fig. 12. Time evolution of the thermal depumping for the deeper defect. Annealing
temperatures are 180◦ C (), 190◦ C (), 206◦ C (•), 220◦ C () and 228◦ C ()
Fig. 13. Time evolution of the thermal depumping for the shallower defect. An-
nealing temperatures are 125◦ C (), 138◦ C (•) and 165◦ C ()
experimental data were first fitted with (17) considering only data with t >
1 h, when the initial drop is completed. The second (shallower) defect clearly
requires lower annealing temperatures to be investigated (see below). The
best-fit time constants for each temperature, cycle are reported in Table 6.
Plotting (Fig. 14) the detrapping time constants in logarithmic scale as a
function of the inverse of the annealing temperature a good linear trend
is observed, confirming the validity of the adopted model. The activation
energy of the defect is Ea1 = 1.62 ± 0.15 eV with an attempt frequency
s01 = 3.5 · 1011 Hz.
A similar procedure was employed to investigate the lower energy defects,
using a 125–165◦C temperature region. More careful experimental procedures
Fig. 14. Detrapping time constant τ as a function of the inverse of the annealing
temperature. The activation energies are calculated through the best fits (solid
lines) with 1/τ = s exp(−ED /kT )
Fig. 15. Decay of the detector efficiency after two hours annealing time vs. the
annealing temperature
should be adopted because of the smallness of the effect observed [31]. The
results are reported in Fig. 13 and the best-fit time constants reported in
Table 6 were extracted. At these temperatures the detrapping time constant
of the deeper traps would be, as calculated from the τ values in Table 6,
on the order of months, so that there is no “mixing” of the two effects. The
activation energy of these shallower traps is (Fig. 14) Ea2 = 1.33 ± 0.06 eV,
s02 = 5.7 · 1010 Hz.
The existence of two traps involved in the pumping–depumping mecha-

nism is evidenced, plotting (Fig. 15) the ratio λh (t = 2h)/λh (t = 0) as a
function of the annealing temperature. The double step shows that a first
kind of trap produces a moderate detrapping in a 2 hour time scale at tem-
peratures around 130◦ C, while another, deeper one comes into play in the
same time scale only at a temperature of about 210◦ C. A plateau exists in
between, when the first trap has been fully depumped after two hours while
detrapping from the other one is not yet effective on this time scale.
6 Conclusion
There is a great variety of experimental techniques that allow us to extract
information on the properties of defects present in CVD diamond films. We
reviewed here a number of them, characterized by the fact of being based
on the analysis of the response of nuclear detectors built from the samples
to be analyzed to irradiation with ionizing particles. Both the realization
of nuclear detectors from the films and the setup of the experiments are
relatively simple. Since basically all these techniques make use of the response
of a solid state nuclear detector, which depends on carrier dynamics, they
can only give information on electrically active defects, i.e., defects acting as
traps for electrons and holes. Within this field, however, they are a simple and
powerful tool to obtain information ranging from the type and distribution
of defects in the film to their activation energy, the type of carriers on which
they act, the trapping and detrapping times, and the mean free paths of the
carriers.
References
[1] J. Field (Ed.): The Properties of Diamond (Academic, London 1979) 107
[2] C. P. Klages: Appl. Phys. A 56, 513 (1993) 108
[3] B. Dischler, C. Wild: Low-Pressure Synthetic Diamond. Manufacturing and
Applications (Spinger, Berlin, Heidelberg 1998) 108
[4] S. Matsumoto: Thin Solid Films 368, 231 (2000) 108
[5] M. Werner, R. Locher: Rep. Prog. Phys. 61, 1665 (1998) 108
[6] M. Marinelli, E. Milani, E. Pace, A. Paoletti, M. Santoro, S. Sciortino, A. Tuc-
ciarone, G. V. Rinati: Photoresponse of CVD diamond films in the 100–300
nm spectral range, J. Electr. Soc. Proc. 97–32, 556 (1998) 109
[7] M. G. Donato, G. Faggio, M. Marinelli, G. Messina, E. Milani, A. Paoletti,
S. Santangelo, A. Tucciarone, G. V. Rinati: Europ. Phys. J. B 20, 133 (2001)
109
[8] S. Zhao: Characterization IF the Electrical Properties of Polycrystalline Dia-
mond Films, Phd thesis, Ohio State University (1994) 110, 111, 126
[9] W. Shockley: Journ. Appl. Phys. 9, 635 (1938) 110
[10] S. Ramo: Proc. of the I.R.E. 27, 584 (1939) 110
[11] K. Hecht: Z. Phys. 77, 235 (1932) 110

[12] T. Benke, A. Oh, A. Wagner, W. Zeuner, A. Bluhm, C. P. Klages, M. Paul,
L. Shaefer: Diam. Relat. Mater. 7, 1553 (1998) 111
[13] Y. Tanimura: Nucl. Instr. Meth. in Phys. Res. A 443, 325 (2000) 111
[14] M. Marinelli, E. Milani, A. Paoletti, A. Tucciarone, G. V. Rinati, M. Angelone,
M. Pillon: Nucl. Instr. Meth. in Phys. Res. A 476, 701 (2002) 111
[15] C. Manfredotti, F. Fizzotti, P. Polesello, E. Vittone: Nucl. Instr. Meth. in
Phys. Res. A 426, 327 (1999) 111
[16] D. Kania: Diamond radiation detectors, in A. Paoletti, A. Tucciarone (Eds.):
The Physics of Diamond (IOS, Amsterdam 1997) 111
[17] W. Adam, et al.: Nucl. Instr. Meth. in Phys. Res. A 511, 124 (2003) 111
M. Pillon: Appl. Phys. Lett. 75, 3216 (1999) 112, 113, 117
[19] A. Balducci, M. Marinelli, E. Milani, M. E. Morgada, G. Pucella, G. Rodriguez,
A. Tucciarone, G. V. Rinati, M. Angelone, M. Pillon: Appl. Phys. Lett. 86,
22108 (2005) 114, 115, 116, 117
M. Pillon: J. Appl. Phys. 89, 1430 (2001) 115, 117
M. Pillon: Phys. Rev. B 64, 195205 (2001) 115, 117, 123, 125, 126, 127
[22] M. Marinelli, E. Milani, G. Pucella, A. Tucciarone, G. V. Rinati, M. Angelone,
M. Pillon: Appl. Phys. Lett. 82, 4723 (2003) 115, 117, 119, 122
[23] J. F. Ziegler, J. P. Biersack, U. Littmark: The Stopping and Range of Ions in
Solids (Pergamon, New York 1985) 116
[24] R. Potenza, C. Tuvè: Measurements of defect density inside cvd diamond films
through nuclear particle penetration, in G. Messina, S. Santangelo (Eds.): Car-
bon, The Future Matrial for Advanced Technology Applications, vol. 100, Topics
Appl. Phys. (Springer, Berlin, Heidelberg 2006) 119
[25] C. Tuvè, V. Bellini, R. Potenza, C. Randieri, C. Sutera, M. Marinelli, E. Milani,
A. Paoletti, G. Pucella, A. Tucciarone, G. V. Rinati: Diam. Relat. Mater. 12,
499 (2003) 119
[26] K. Zanio, W. Akutagawa, J. W. Mayer: Appl. Phys. Lett. 11, 5 (1967) 123
[27] M. Martini, T. A. McMath: Nucl. Instr. Meth. in Phys. Res. 79, 259 (1970)
123
[28] M. Marinelli, E. Milani, A. Paoletti, G. Pucella, A. Tucciarone, G. V. Rinati,
M. Angelone, M. Pillon: Diam. Relat. Mater. 12, 1733 (2003) 123, 128
[29] C. Manfredotti, E. Vittone, A. L. Giudice, C. Paolini: Diam. Relat. Mater. 11,
446 (2002) 125
[30] M. Marinelli, E. Milani, M. E. Morgada, G. Pucella, G. Rodriguez, A. Tuccia-
rone, G. V. Rinati, M. Angelone, M. Pillon: Appl. Phys. Lett. 83, 3707 (2003)
129, 130
[31] A. Balducci, M. Marinelli, E. Milani, M. E. Morgada, G. Pucella, G. Rodriguez,
A. Tucciarone, G. V. Rinati, M. Angelone, M. Pillon: Phys. Stat. Sol. (a) 201,
2542 (2004) 129, 130, 132
Index
activation energy, 128–132 fast component, 123, 128
carrier mobility, 110, 116 Hecht theory, 110, 111, 114, 116, 118
charge collection distance (CCD), 110, high-pressure high-temperature
117, 118, 130 (HPHT), 107
charge collection spectrum, 111–113,
125 Monte Carlo simulation, 113
chemical vapour deposition (CVD),
penetration depth, 112
107–112
polycrystalline diamond, 108, 109, 112,
columnar growth, 109, 111–113, 115,
116
116, 118
Ramo’s theorem, 110, 123–125
defects, 108
defects distribution, 108, 110–113, slow component, 123–125, 127–129
115–117, 119, 133
grain boundaries, 108, 109, 112, 113, time constant, 123, 126, 128–132
115, 116, 118–120 trap, 108
grain boundaries distribution, 112, depumping, 108, 128, 129
113, 116, 118 priming, 108, 111–113, 115, 117
in-grain defects, 112, 113, 115, 116, pumping, 108, 111, 113, 115, 117,
118, 119 122, 124, 125, 128–130
in-grain defects distribution, 112, trapping, 110, 113, 123–131
113, 115, 118
detector efficiency, 110 variable energy method (VEM),
detectors, 108 119–122, 130
detrapping, 123–129, 131 variable incidence method (VIM),
diamond, 107 119–122, 130
Effects of Nanoscale Clustering in Amorphous
Carbon
J. David Carey and S. Ravi P. Silva
Nanoelectronics Centre, Advanced Technology Institute, School of Electronics and

Physical Sciences, University of Surrey, Guildford, GU2 7XH, UK
David.Carey@surrey.ac.uk
Abstract. In this review the effects of clustering associated with the sp2 and sp3
phases of amorphous carbon thin films are examined. We highlight that many of
the optical and electronic properties of these films can be explained by considera-
tion of disorder in the sp2 phase. Within the context of topological and structural
disorder, we explain the variation of the visible Raman line width, Raman shift,
Tauc gap and Urbach energy as a function of deposition conditions. We further
go on to describe how intra-sp2 cluster interactions are responsible for the narrow-
ing of the electron paramagnetic resonance line width with increasing spin density
and how this intracluster interaction can be extended to the intercluster transport
properties, in particular, for electron field emission from the films. We also examine
how the mechanical properties of carbon films are affected by clustering which can
be enhanced by thermal annealing.
1 Introduction and Bonding in Carbon

Crystalline carbon is unique amongst the elements of the periodic table, be-
ing able to form one of the hardest naturally occurring materials, diamond,
and also one of the softest, graphite. From an electronic point of view these
two materials possess, respectively, a 5.5 eV energy band gap and a zero band
gap, making the electronic properties vary between those found in an insula-
tor and those of a semimetal. This very different behaviour can be traced to
the bonding and the bond hybridization that are present (Fig. 1). In the case
of sp3 hybridization, four σ bonds, arranged at 109.5◦ to each other, allow
for the tetrahedral bonding of diamond. These four bonds are responsible
for the high hardness and large energy band gap. By contrast, in the case
of sp2 -hybridised carbon there are three σ bonds bonding in-plane to three
other carbon atoms at 120◦ to each other. In addition, there is one pz or-
bital perpendicular to the plane of the threefold bonding. This anisotropy in
the bonding distinguishes the in-plane electronic properties of graphite from
the properties perpendicular to the basal plane. Whilst there is a strong in-
plane interaction, the weaker out-of-plane interactions are responsible for the
lubricant properties of graphite.
These bond hybridizations are not restricted to the crystalline forms of
carbon and are available to form amorphous carbon films (Fig. 1), which have
138 J. David Carey and S. Ravi P. Silva
Fig. 1. A schematic representation of different types of hybridised bonding and

the resultant crystalline and model amorphous structures
a wide range of properties, some of which are presented in Table 1. It is im-

portant at the outset to distinguish between the different types of amorphous
carbon. For example, from Table 1 the physical and electronic properties of
tetrahedral amorphous carbon (ta-C) are very different from hydrogenated
diamond-like carbon (DLC:H), even though they both have a similar sp3
fraction. In the case of the former, the sp3 content is mainly made up of
C–C bonds, which results in a high-hardness thin film, whereas in the latter
the sp3 content also contains C–H bonding in addition to C–C bonding. In
this review the terms amorphous carbon, a-C, and hydrogenated amorphous
carbon, a-C:H, will be used as generic terms for disordered carbon, either H-
free or H-containing. For specific types of films, terms such as diamond-like
carbon or polymer-like carbon will be used [1].
One of the main driving forces for the technological use of amorphous
carbon thin films is the ability to grow thin films over large area, at low tem-
peratures with excellent uniformity. The low-temperature deposition, often
performed using plasma-enhanced chemical vapour deposition (PECVD), al-
lows the growth onto nonconventional substrates such as organics or glass
substrates. The excellent uniformity of the physical and electronic properties
is also matched by the mirror-smooth quality of the films. Extensive atomic-
force microscopic studies of a-C and a-C:H films have shown that they typ-
Effects of Nanoscale Clustering in Amorphous Carbon 139
Table 1. Physical properties of different forms of a-C thin films
Category sp3 Optical bandgap H Density Hardness

% eV at. % g cm−3 GPa
Polymer-like a-C PAC 60–80 2.0–5.0 40–65 0.6–1.2 Soft
Graphitic-like a-C GAC 0–30 0.0–0.6 0–40 1.2–2.0 Soft
Diamond-like a-C DLC 40–60 0.8–4.0 20–40 1.5–3.0 20–40
Tetrahedral a-C ta-C 65–90 1.6–2.6 0–30 2.5–3.5 40–65
ically have a rms roughness of less than 1 nm. This makes a-C films ideal
for microelectromechanical systems and for hard disk coatings. In addition,
a-C is believed to exhibit excellent biocompatibility and a-C-coated surgical
implants have been demonstrated. Less success has been forthcoming in the
development of active electronic applications of a-C and a-C:H films [2], and
as a consequence it is necessary to examine the electronic properties and the
effects of disorder and localization at a nanometer level. To that end, this ar-
ticle first deals with the physics of the disorder associated with the sp2 phase
and then extends the discussion to the intercluster transport properties of
the films. Finally, field emission from the carbon system is discussed.
2 Disorder in Amorphous Carbon

Amorphous carbon can be considered as a disordered mixed-phase material
consisting of a conductive sp2 phase, embedded in a less conductive sp3 ma-
trix. The sp2 -hybridised carbon atoms are usually in the form of nanometer-
sized clusters, mainly in the form of rings or olephinic chains. These sp2
clusters give rise to occupied π and unoccupied π ∗ bands with the separation
between the bands being related to the size of the cluster. The sp3 com-
ponent tends to be in the form of aliphatic chains of C atoms with C–C
and C–H bonding being present. The optoelectronic properties of a-C and
a-C:H films can be described in terms of transitions between the occupied π
states to the unoccupied π ∗ states. Previous studies have shown [3] that it is
possible to represent the occupied π and unoccupied π ∗ bands by Gaussian
functions centred at Eπ(π∗ ) . The σ and σ ∗ states, which are associated with
sp3 C bonding, lie further separated from each other. By deposition on suit-
able transparent substrates, such as Corning glass, it is possible to extract
the parabolic Tauc gap and the exponential Urbach energy.
• The Tauc band gap corresponds to optical transitions between extended
states, and for the a-C system in the absence of disorder, the Tauc gap
is given by the separation between the π–π ∗ bands with a magnitude of
2 Eπ . The larger the average size of the sp2 cluster, the smaller the Tauc
gap.
• An alternative band gap that is sometimes employed is to measure the

energy at which the absorption coefficient is 104 cm−1 ; this is known as
the E04 gap. Values associated with the E04 gap tend to be larger than
the Tauc gap, as the absorption at 104 cm−1 tends to occur in states
beyond the π and π ∗ band edges. E03 and E05 gaps are also sometimes
quoted in the literature.
• Transitions between extended-to-localised states in amorphous semicon-
ductors are usually characterised by the Urbach energy, E U . The value of
E U is found by fitting an exponential function to the slope of the absorp-
tion edge and has been used as a measure of disorder in other material
systems. For example, in the case of low defect density a-Si:H, the valence
band tail has an exponential slope of about 45 meV, whereas the slope
of the conduction band is 25 meV. In this manner the observed Urbach
energy of 55 meV for a-Si:H is mainly associated with the valence band
and with low-lying defect states.
Fanchini and colleagues have, however, suggested that while the experi-
mentally observed values can be extracted, the validity of this approach for
this dual mixed-phase system may be too simplistic. As a result, care must be
exercised in the interpretation of the Tauc band gap and the Urbach energy
that are obtained [4]. A second measure of disorder, based upon bond angle
distortion, is the full width half maximum of the Raman active G band, Γ G .
This Raman signal is found at around 1580 cm−1 and originates from the E 2g
vibration between sp2 C bonds [5]. Several complementary optical [6,7,8] and
Raman studies [7, 9] have revealed that as the Tauc gap increases the Urbach
energy increases monotonically, but that the width of the G band undergoes
a maximum of about 1.5 eV (Fig. 2).
In this respect, if the Urbach energy and the width of the G band are
valid measures of disorder in the a-C:H system, then they must measure
different aspects or types of disorder. Robertson previously proposed that the
Urbach energy is a measure of an inhomogeneous disorder associated with
different sp2 cluster sizes, whereas the G band width has been attributed to
a homogeneous disorder associated with bond angle disorder [10]. However,
such an interpretation is unable to explain:
(i) why the largest values of E U are observed at the highest Tauc gaps
(Fig. 2a), since the cluster sizes are smallest there is only a small limited
number of different cluster conformations possible;
(ii) the observation of a maximum in the G-band line width with Tauc gap
(Fig. 2b);
(iii) why the G-band position is also dependent on the excitation wave-
length, with the G-band position tending to saturate at ∼ 1600 cm−1
for excitation wavelengths below 350 nm and the peak shifting to lower
energies at longer wavelengths, as shown in Fig. 3a;
Fig. 2. Variation of (a) Urbach energy and (b) linewidth of Raman G band for
different a-C films as a function of optical Tauc gap. The different symbols refer to
different sets of films (see [2])
(iv) the observation that the G-band line width depends on the excitation
wavelength with the narrower values of Γ G observed at increasing ex-
citation energy as seen in Fig. 3b.
Since the Raman spectra are a result of resonant excitation of sp2 clusters,
dispersion of the G peak reflects excitation of clusters with different band
gaps. The saturation of the G-band position at ∼ 1600 cm−1 in Fig. 3a
reflects the maximum possible Raman shift for C atoms bonded in sp2 rings.
The observation of dispersion of the Raman line width implies that there is an
inhomogeneous distribution present in the bond angle disorder. Furthermore,
the reduction of the G width with increasing excitation energy implies that
there is a narrower distribution of higher-gap sp2 states, in contradiction with
the proposal that there must be a larger distribution of sp2 states required for
the large Urbach energy. The increase in the Urbach energy with increasing
π–π ∗ separation and small numbers of possible sp2 cluster configurations,
coupled with the dispersion in the G line width have lead Fanchini and
Tagliaferro to conclude that the Urbach energy in amorphous carbon is not
a good measure of disorder [11]. Furthermore, they went onto propose two
types of disorder:
(i) A structural disorder associated with clusters of the same size but with
different amounts of distortion. This type of inhomogeneous disorder
would increase the Urbach energy as well as broaden the G-band line
width.
(ii) A topological disorder arising from undistorted clusters but of different
sizes. This would also broaden the Urbach energy but have no effect on
the G line width.
Fig. 3. Variation of (a) G-band Raman shift and (b) G-band linewidth for a series
of polymer-like a-C:H films (), diamond-like a-C:H () and graphitic carbon (•)
thin films as a function of Raman excitation wavelength (see [5])
As a consequence, the differences between distorted and undistorted sp2

clusters needs to be explored. The location of the sp2 clusters in the energy
gap depends on two factors: whether they consist of even or odd numbers of
carbon atoms and whether they are distorted. Undistorted even-numbered
clusters will give rise to states near the Fermi level (E F ) only if they are
sufficiently large. This differs from the case of odd-numbered clusters, which
can give rise to gap states even if composed of a small number of atoms. In
general, distorted clusters give rise to states that are closer to E F than undis-
torted clusters. One method that is able to distinguish between the different
types of cluster is to measure the density of unpaired electron spins arising,
for example, from odd-numbered clusters. Such states can be measured by
electron paramagnetic resonance (EPR), also called electron spin resonance
(ESR). A high concentration of paramagnetic defects (1020 cm−3 ) is typically
found in films deposited under energetic conditions. Such energetic conditions
include samples deposited on the driven electrode of a PECVD system [12].
Low defect densities, typically ∼ 1017 cm−3 , tend to be found under less
energetic conditions [13]. The term “defect” is a misnomer since the ESR
active species are generally agreed to result from different configurations of
sp2 clusters. In this manner, ESR can give a measure of the density of sp2
states at the Fermi level, provided that a majority of these states have a net
unpaired electron spin present. It should be noted that ESR spin densities
do not themselves measure the sp2 content.
3 Intracluster Effects in Amorphous Carbon

Several studies have examined the variation of spin density with deposition
conditions, though rarely in terms of relating spin density to localisation of
Fig. 4. Variation of (a) spin density and (b) peak-to-peak linewidth with negative
self-bias for three series of a-C:H films (see [2])
the wavefunction [2]. The variation of the spin density, N s , and peak-to-peak
line width, ∆B pp , with negative self-bias for three different series of films are
shown Fig. 4.
It is evident that the spin density rises from a low value of ∼ 1017 cm−3
and tends to saturate at 1020 cm−3 . The ESR line width initially increases, of-
ten as the spin density increases, but at higher biases the line width decreases.
In this regime the line shape of the ESR signal is Lorentzian. Furthermore,
in these high-bias conditions, the spin densities are typically ∼ 1020 cm−3 ,
corresponding to ∼ 1 at. % of the film. As a result, the ESR signal can be
considered as originating from a dilute paramagnetic material. Abragam’s
formulism may be used to calculate the contribution to the peak-to-peak line
width, ∆B pp , from the dipole–dipole interaction between like spins [14]. Such
a broadening mechanism will produce a Lorentzian ESR line shape, which
is consistent with the ESR line shapes observed. Assuming a spin concen-
tration N s , measured in cm−3 , the contribution to the dipolar interaction
∆B pp , measured in mT, can be given as [12]
4π2
∆B pp = g µB N s . (1)
9
Barklie et al. applied (1) for a typical C-related g value of 2.0025 to ob-
tain the expression ∆B pp = 8.12 × 10−21 N s . The variation of ∆B pp against
N s directly for the same three data sets reported in Fig. 4, as well as the
predicted line width (dashed line) based solely on dipolar broadening us-
ing (1) is shown in Fig. 5. It is apparent that significant reductions in ∆B pp
are observed at high spin densities, indicating an additional mechanism is
responsible for reductions in the line width. This additional interaction is
due to the motional averaging of the electron wavefunction within the sp2
Fig. 5. Variation of peak-to-peak linewidth with spin density for a data in Fig. 4.
The dashed line represents the predicted dipolar contribution to the linewidth based
on (1)
cluster. As the sp2 cluster increases in size, the probability of the electron
being associated with a particular atom decreases. This spreading or delo-
calization of the wavefunction as the cluster size increases is accompanied by
a general reduction in the Tauc gap. Further evidence of delocalization can
also be found in an examination of the spin resonance relaxation times. The
spin–lattice relaxation time, T1 , decreases from 3 × 10−5 s for polymer-like
films down to 2 × 10−7 s for DLC films, which implies a greater interaction
between the spin system and lattice [12]. At the same time, the spin–spin
relaxation times increase, reflecting a change in the exchange frequency. It
has been reported that in films with a high sp2 content that the exchange
frequency is approximately ∼ 1010 rad s−1 .
In addition to the concentration of the sp2 clusters, the size of the sp2
clusters is therefore an important feature of this description. Indeed, there
have been attempts to infer the cluster size and shape using visible Raman
spectroscopy [15]. As a result, considerable interest was generated by the
report [16] of an apparent anisotropy in the ESR signal of defects in DLC
films measured at high frequency and low temperature. A fully amorphous
material with a single spin 12 paramagnetic centre should exhibit a single
isotropic line. In [16] it was proposed that the ESR signal could be assigned
to the sum of an anisotropic powder spectrum with axial g values of g ∼
2.005 and g⊥ ∼ 2.0025, and one symmetric unresolved line with g = 2.0025.
From previous ESR studies of nanocrystalline graphite, it is known that the
g tensor is anisotropic and the g values depend on the average crystallite size.
A value of g of 2.005 was associated [17] with a crystallite of a size less than
7.5 nm. In this way, it was believed that ESR was providing a way to quantify
the size of the nanoclusters. This is an extremely attractive feature since it
has been proposed that the ratio of the intensities of the G and D Raman
bands (the D band being located at 1350 cm−1 and associated with the A1g
breathing mode of 6-fold sp2 rings) can be related to the in-plane correlation
length of disordered graphite. The reason that no anisotropy was observed
at the more conventional 9 GHz room-temperature measurement was due
to the strong exchange narrowing. Unfortunately, in this analysis the role of
demagnetizing fields, which at room temperature and “low” magnetic fields
are negligible, but must be included when operating at 4 K and especially at
higher resonance fields, required at the higher microwave frequencies [18, 19].
When demagnetizing fields are taken into account, an apparent anisotropy of
the resonance can occur since the demagnetizing field results in an effective
shift downwards of the effective g value when the applied Zeeman magnetic
field is perpendicular to the film and results in an up-shift when the applied
field is parallel to the plane.
Apart from changes in deposition conditions, such as self-bias or the ad-
dition of N, nanoclustering within a-C films can be adjusted via thermal an-
nealing. Siegal et al. annealed nonhydrogenated a-C films produced by pulsed
laser ablation and reported the presence of regions 3–5 nm in size which were
5–10% more dense that the surrounding regions [20]. The as-deposited films
has stresses of 5–7 GPa and a film density of 3.0 g cm−3 , but near-complete
stress relief had occurred and the density had dropped to 2.8 g cm−3 with
thermal annealing to 600◦ C. Over the same annealing temperature range
the optical transparency of the film decreased, demonstrating that cluster-
ing of the sp2 within the film has occurred. By considering different bonding
topologies on the basis of Raman spectroscopy, Siegal et al. concluded that
the as-grown films consist of small and isolated sixfold clusters, but that
above annealing temperatures of 300◦ C the fraction of five-membered rings
begins to increase. The presence of five-membered rings was attributed to
the newly formed π-bonded atoms, rather than a reduction in the popula-
tion of the six-membered rings themselves. Despite the reduction in the film’s
density, nanoindentation measurements showed that the nanocomposite films
were 15% harder than the as-deposited films.
Computational studies showed that within this nanocomposite film, the
residual stress consists of an inhomogeneous stress distribution where the
three-fold coordinated atoms are under tensile stress, but that the four-fold
atoms are under a smaller compressive stress [21]. During annealing small re-
gions of high stress relax, leading to the preferential generation of three-fold
coordinated atoms. The stress relief is as a result of fourfold sp3 C bonds being
replaced by threefold sp2 C bonds. This change to the shorter sp2 C bond is
irreversible under normal conditions. If the near-zero stress is achieved due to
high temperature annealing, care must the exercised to prevent an excessive

thermal stress being introduced into the film on cooling back to room tem-
perature. The thermal stress can be eliminated by growing the film at above
room temperature, which introduces some tensile stress. Further evidence of
the formation of clustering in ta-C films was reported in the fluctuation mi-
croscopy studies recently undertaken by Chen et al. [22]. Annealing up to
600◦ C resulted in the increase in locally ordered regions, but the formation
of graphitic ordering did not occur until annealing to 1000◦ C. Much of the
research for stress relief in hard a-C films, especially ta-C films, is driven by
the development of carbon-based microelectromechanical systems.
Evidence for mixed-phase components in a-C has also been obtained by
examining the proton and 13 C relaxation times using nuclear magnetic res-
onance (NMR). Jäger et al. used cross-polarization 13 C NMR to measure
the effects on the proton relaxation times in different environments of films
deposited from acetylene and benzene [23]. A two-component 1 H spin–lattice
relaxation with time constants of 14 and 120 ms was found and interpreted
as being due to the presence of two differently relaxing proton systems. The
shorter relaxation time originates from CH groups in the sp2 and sp3 matrix
and the longer T1 time originating from short CH2 polymer units with an sp3
configuration. In this description the two environments are separated by re-
gions of nonhydrogenated sp2 C. It is was further found that for films grown
by rf glow discharge from benzene at a self-bias voltage of −200 V, the proton
relaxation time of ∼ 10 ms was virtually independent of temperature. This
temperature independence of the shorter T1 component was interpreted in
terms of nuclei being relaxed via spin diffusion to paramagnetic centers. It is
therefore the degree of electron aggregation or clustering around these para-
magnetic centers that determines the proton relaxation. The influence of spin
diffusion to paramagnetic centers as an explanation of the temperature inde-
pendence of the spin–lattice relaxation times implies that an understanding
of the factors that affect the spin density and line width are important.
Tamor et al. estimated from NMR measurements the protonated and
nonprotonated C concentration with either sp2 or sp3 hybridisation [24]. At
a self-bias of −500 V, the H content was estimated to be 36 at. % and the
fraction of protonated sp3 C atoms to be 0.16, nonprotonated sp3 C atoms to
be 0.20, protonated sp2 C atoms to be 0.25, and nonprotonated sp2 C atoms
to be 0.37. For a film density of 1.7 g cm−3 , this result suggests there are
5.1 × 1022 sp2 C atoms cm−3 in the film. If the signal observed from ESR
is associated with sp2 C centers, then for a value of N s of 1020 cm−3 , this
corresponds to 1 spin per 510 sp2 C atoms, or equivalently to 1 spin per 300
atoms of nonprotonated sp2 C atoms. If these spins were randomly distributed
throughout the film, cooperative effects, such as exchange, should not occur.
Since exchange effects have been attributed to films grown at high bias, the
combined use of NMR and ESR confirms and quantifies the clustering of the
sp2 phase has taken place.
4 Intercluster Interactions in Amorphous Carbon

The increase in intracluster delocalization also manifests itself in changes that
can be associated with intercluster interactions. Various transport studies of
a-C films reveal a commonality in behaviour with high resistivity associated
with samples with large Tauc gap. The resistivity decreases as the Tauc gap
decreases and the conductivity associated with various mechanisms such as
variable range hopping in the more conductive DLC films [1]. In the case of re-
sistive polymer-like amorphous carbon, the conductivity has been attributed
to a space charge-induced current mechanism [1]. Improvements in the con-
ductivity have also been reported with nitrogen incorporation and attributed
to a weak “doping” effect [25] or to a reduction of the band gap through
graphitization [26]. In terms of post-deposition processing, transport studies
of annealed films also reveal an improvement in the conductivity. However,
this is almost always accompanied by significant modification to the structure
of the film at a microscopic level. In this respect furnace annealing results in
an improvement in electrical properties but is accompanied by global changes
to the film. Ion implantation as a method of injection of thermal energy is
therefore an alternative process to modify the film properties in a highly
controlled manner with precise depth control by the correct choice of ion
species, ion energy and dose. Using high-dose B+ implantation into polymer-
like films, it is possible to reduce the low-field resistivity from 2 × 1014 Ω cm
(unimplanted) to 6 × 1012 Ω cm (2 × 1014 B+ cm−2 ). Over the same range the
Tauc gap remained constant at about 2.6 eV, indicating that no significant
increase in the mean sp2 cluster size has occurred. For doses greater than
2 × 1014 cm−2 , a fall in the optical gap finally ends with a total collapse to
0.2 eV at a high dose (2 × 1016 B+ cm−2 ), and the resistivity concomitantly
decreases to 5 × 106 Ω cm. The improvement in the electrical characterization
has been attributed to improvements in the cluster–cluster interaction via a
hopping related mechanism (Fig. 6).
During low-dose implantation, localized heating results in localized sp2
clusters, which remain small (as evidenced by the absence of a reduction in
the Tauc gap). Before the formation of damage cascades, the concentration
of the induced sites will be proportional to the ion dose. This nanoclustering
of the sp2 sites, which maintains the wide gap, increases the conductivity by
decreasing the hopping distance. At higher doses, nanostructuring of the bulk
of the film occurs and the band gap decreases. As a result, ion implantation
demonstrates that it is possible to improve the conductivity without inducing
extensive graphitizing of the film.
Further evidence for the idea of dielectric inhomogeneity of conductive
sp2 clusters embedded in a more insulating matrix can be found in the high-
resolution scanning tunneling microscope (STM) for a DLC film, where the
STM results show that there are nanometer-sized regions of higher conduc-
tivity surrounded by other regions of lower conductivity [2]. Similar localized
regions of conductivity have also been recently reported in an STM study of
Fig. 6. Schematic of how the different bias growth conditions can give rise to a
change in the size and concentration of sp2 clusters
sulphur-containing ultrananocrytalline diamond (UNCD) films [27]. UNCD

films consist of nanometer-sized grains surrounded by graphitic grain bound-
aries. The surface density of states was determined from the normalized dif-
ferential conductivity, which for sulphur-rich films showed an oscillatory be-
haviour. In UNCD films, it was reported that the sulphur acts both to dope
the films but also to introduce localized defects. In both cases increasing the
concentration and size of the clusters results in improved localized electrical
conductivity. This is an important result since conventional “large-area” con-
tacts used in current–voltage characteristics would mask the localized nature
of the conductivity.
5 Field Emission from Amorphous Carbon

The implication of this “dielectric inhomogeneity” can also be extended to
examine the field-induced electron emission from a-C:H films. Application
of an applied electric field can result in electron emission from the surface.
In the case of PAC, films it has been reported [28] that the threshold elec-
tric field for emission exhibited a dependence with the film thickness with
a minimum threshold observed for a thickness of 65 nm. This behavior was
attributed to the effects of internal high electric field effects within the film
and the absence of any significant screening of the applied electric field. Such
an explanation is possible when one considers that the electric field from
the anode will terminate on the “more conductive” sp2 clusters. At a low
density (∼ 1017 cm−3 ) screening of the bulk of the film will not occur. In
such a situation the controlling step in the electron emission process is at

the film/substrate. A different situation will result for films grown at higher
biases when the defect density increases. For two films deposited at −90 V
and −265 V self-bias, it was observed [29] that there was little dependence
on the film thickness and the lowest threshold field was reported for the film
grown at the highest bias, consistent with the model presented in Fig. 6.
Here, the film with the largest bias has the greatest intercluster interactions
since the clusters are larger and the cluster–cluster separation will be the
lowest. This gives rise to an emission mechanism that is dominated by the
front surface properties of the film. Furthermore, the local electric field in the
neighbourhood of the clusters is enhanced, resulting in lower than predicted
threshold fields for emission.
Finally, it has been reported in several studies that the onset of stable
field emission, often in more resistive samples, occurs only after several volt-
age cycles [30,31]. This conditioning behaviour often results in lower threshold
fields. It has been shown that it is possible to condition a PAC film by current
stressing, by intentionally passing a current of suitable magnitude through
the film [31]. Mercer et al. performed a similar experiment using a scanning
tunneling microscope tip to generate a highly spatially localized electric field
and effectively current stress ta-C films [32]. They observed that after ramp-
ing the tip–sample bias and current, nanostructures of about 100 nm in extent
form. By using high-resolution spatially-resolved electron energy loss spec-
troscopy (EELS), they showed that the predominant bonding configuration
changes from predominately fourfold coordinated C to threefold coordination.
Changes in the sp2 bonding have also been reported for heavy ion irradiation
of S-doped UNCD films [33]. High-energy (GeV/amu) Fe and Si ions were
used, and it was reported that there was a reduction in the threshold field af-
ter implantation for irradiated films and this improvement in threshold field
was accompanied by changes in the sp2 phase as inferred from by Raman
spectroscopy. The net effect of the conditioning, current stressing or heavy
ion treatment is to generate conductive sp2 -rich areas or filaments through
localized Joule heating. These conductive regions can then be viewed as ei-
ther isolated sp2 clusters or a continuous sp2 network that can lead to high
field enhancement factors, which result in reduced threshold fields.
Further evidence of the importance of the sp2 phase in field emission
comes from a comparative study of the threshold field with the ratio of the
I D /IG bands found in Raman spectroscopy. For a range of different types of
a-C films (including ta-C, ta-C:N, ta-C:H and DLC films [16] and S-doped
UNCD films [34]), it was observed that the threshold field is correlated with
the in-plane correlation length, suggesting sp2 cluster sizes of around 1–2 nm.
The in-plane correlation length is inferred from the ratio of the intensities of
D and G Raman bands.
6 Conclusions
In conclusion, the disorder inherent in amorphous carbon films plays a major
role in determining the electronic properties. An understanding of the role
of disorder is important. PAC films have small clusters with a small distri-
bution of cluster sizes and are dominated by topological disorder in which
the Urbach energy is not a good measure of disorder and topological disorder
dominates. In DLC films a reduction in the Tauc gap and blue shift of the
G-band implies a higher concentration of larger clusters is present. Further-
more, the increase in spin density (1020 cm−3 ) coupled with the increase in
the G band width implies more distorted clusters are present. At the higher
spin density the ESR line width is no longer determined by the dipole inter-
action, with intracluster exchange interactions determining the line width.
In this regime, both topological and structural disorder are present and can
effect the stress in the film. The disorder is dominated by structural disorder
with little topological disorder, thereby reducing the line width of G band.
The clustering can be adjusted by thermal annealing or ion implantation. In
addition, the sp2 phase plays an important role in the field emission from
these films.
Acknowledgements
The authors acknowledge funding from the EPSRC, and JDC acknowledges
the funding of an Advanced Research Fellowship from the EPSRC.
References
[1] S. R. P. Silva, J. D. Carey, R. U. A. Khan, E. G. Gernster, J. V. Anguita: Hand-
book of Thin Film Materials (Academic Press, New York 2002) and reference
therein 138, 147
[2] J. D. Carey, S. R. P. Silva: Phys. Rev. B. 70, 235417 (2004) 139, 141, 143,
147
[3] D. Dasgupta, F. Demichelis, C. F. Pirri, A. Tagliaferro: Phys. Rev. B. 43, 2131
(1991) 139
[4] G. Fanchini, S. C. Ray, A. Tagliaferro: Diam. Relat. Mater. 12, 891 (2003)
140
[5] A. C. Ferrari, J. Robertson: Phys. Rev. B. 61, 14095 (2000) 140, 142
[6] J. Ristein, J. Schafer, L. Ley: Diam. Relat. Mater. 4, 509 (1995) 140
[7] A. Zeinert, H.-J. von Bardeleben, R. Bouzerar: Diam. Relat. Mater. 9, 728
(2002) 140
[8] F. Giorgis, F. Giuliani, C. F. Pirri, A. Tagliaferro: Appl. Phys. Lett. 72, 2520
(1998) 140
[9] M. A. Tamor, W. C. Vassell: J. Appl. Phys. 76, 3823 (1994) 140
[10] J. Robertson: Mater. Sci. Eng. R 37, 129 (2002) 140
[11] G. Fanchini, A. Taliaferro: Appl. Phys. Lett. 85, 730 (2004) 141
[12] R. C. Barklie, M. Collins, S. R. P. Silva: Phys. Rev. B. 61, 3546 (2000) 142,
143, 144
[13] M. Collins, R. C. Barklie, J. V. Anguita, J. D. Carey, S. R. P. Silva: Diam.
Relat. Mater. 9, 781 (2002) 142
[14] A. Abragam: Principles of Nuclear Magnetism (Clarendon, Oxford 1996) 143
[15] A. Ilie, A. C. Ferrari, T. Yagi, J. Robertson: Appl. Phys. Lett. 76, 2627 (2000)
144
[16] H. J. von Bardeleben, J. L. Cantin, A. Zeinert, B. Racine, K. Zellama, P. Hai:
Appl. Phys. Lett. 78, 2843 (2001) 144, 149
[17] G. Wagoner: Phys. Rev. 118, 647 (1960) 145
[18] B. Druz, I. Zaritskiy, Y. Evtuikhov, A. Konchits, M. Y. Valakh, S. P. Kolesnik,
B. D. Shanina, V. Visotskij: Mat. Res. Soc. Symp. Proc. 593, 249 (2000) 145
[19] B. J. Jones, R. C. Barklie, G. Smith, H. E. Mkami, J. D. Carey, S. R. P. Silva:
Diam. Relat. Mater. 12, 116 (2003) 145
[20] M. P. Siegal, D. R. Tallant, P. N. Provencio, D. L. Overmyer, R. L. Simpson,
L. J. Martinez-Miranda: Appl. Phys. Lett. 76, 3052 (2000) 145
[21] P. C. Kelires: Phys. Rev. B. 62, 15686 (2000) 145
[22] X. Chen, J. P. Sullivan, T. A. Friedmann, J. M. Gibson: Appl. Phys. Lett. 84,
2823 (2004) 146
[23] C. Jäger, J. Gottwald, H. W. Spiess, R. J. Newport: Phys. Rev. B. 50, 849
(1994) 146
[24] M. A. Tamor, W. C. Vassell, K. Carduner: Appl. Phys. Lett. 58, 592 (1991)
146
[25] G. A. J. A. S. R. P. Silva: Appl. Phys. Lett. 68, 2529 (1996) 147
[26] R. U. A. Khan, J. D. Carey, S. R. P. Silva, B. J. Jones, R. C. Barklie: Phys.
Rev. B. 63, 121201 (2001) 147
[27] S. Gupta, B. R. Weiner, G. Morell: J. Appl. Phys. 97, 094307 (2005) 148
[28] R. D. Forrest, A. P. Burden, S. R. P. Silva, L. K. Cheah, X. Shi: Appl. Phys.
Lett. 73, 3784 (1998) 148
[29] J. D. Carey, R. D. Forrest, S. R. P. Silva: Appl. Phys. Lett. 78, 2339 (2001)
149
[30] A. A. Talin, T. E. Felter, T. A. Friedmann, J. P. Sullivan, M. P. Siegal: J. Vac.
Sci. Technol. A. 14, 1719 (1996) 149
[31] J. D. Carey, S. R. P. Silva: Appl. Phys. Lett. 78, 347 (2001) 149
[32] T. W. Mercer, N. J. DiNardo, J. B. Rothman, M. P. Siegal, T. A. Friedmann,
L. J. M. Miranda: Appl. Phys. Lett. 72, 2244 (1998) 149
[33] A. Gonzalez-Berrios, D. Huang, N. M. Medina-Emmanuelli, K. E. Kristian,
O. O. Ortiz, J. A. Gonzalez, J. D. Jesus, I. M. Vargas, B. R. Weiner, G. Morell:
Diam. Relat. Mater. 13, 221 (2004) 149
[34] S. Gupta, B. R. Weiner, G. Morell: Appl. Phys. Lett. 80, 1471 (2002) 149
Index
a-C, 137, 149 cluster size, 139–141, 144, 147
a-C:H, 138 distorted clusters, 142
2
sp -bonded clusters, 137, 139, 142 intercluster interaction, 137, 147
intracluster interaction, 137, 142, 150 G band, 140, 141

undistorted clusters, 142 G-band, 150
sp2 -chains, 139 G-band, 145, 149
sp2 -rings, 139, 141, 145 graphite-like carbon (GLC), 139, 142
five-fold rings, 145 graphitization, 147
six-fold rings, 145
ta-C, 138, 149 hydrogenated diamond-like carbon
ta-C:H, 149 (DLC:H), 138, 142, 149
ta-C:N, 149
nitrogen incorporation, 147
amorphous carbon thin films, 137
optical properties, 137
cluster-cluster interaction, 147
plasma-enhanced chemical vapour
cluster-cluster separation, 149
deposition (PECVD), 138, 142
clustering, 137, 145, 146
polymer-like carbon (PLC), 138, 139,
142, 144, 147
D band, 145, 149
D/G-band intensity ratio, 145, 149 Raman spectroscopy, 137, 140–142,
defects, 142, 144, 148 144, 145, 149
diamond-like carbon (DLC), 138, 139,
144, 147, 149, 150 scanning tunneling microscopy (STM),
disorder, 137, 139–141, 150 147
homogeneous disorder, 140
inhomogeneous disorder, 140, 141 Tauc gap, 137, 139–141, 144, 147, 150
tetrahedral amorphous carbon, 139
electron paramagnetic resonance
(EPR), 142 ultrananocrytalline diamond (UNCD),
electron spin resonance (ESR), 142–144, 148, 149
146, 150 Urbach energy, 137, 139–141, 150
Elastic and Structural Properties of Carbon
Materials Investigated by Brillouin Light
Scattering
Marco G. Beghi, Carlo S. Casari, Andrea Li Bassi, and Carlo E. Bottani
NEMAS-Center for “Nano-Engineered MAterials and Surfaces”, and

Dipartimento di Ingegneria Nucleare, Politecnico di Milano, Via Ponzio 34/3,
I-20133 Milano, Italy
marco.beghi@polimi.it
Abstract. Brillouin light scattering (BS) is a nondestructive technique exploited

to investigate collective excitations in solids: acoustic phonons, spin waves, magnons
and density fluctuations. In particular, it permits users to characterize dynamical
structural and nanomechanical properties of materials through the detection of ther-
mally excited surface and bulk acoustic phonons. This gives access to information
about both surface and bulk structural properties in the spatial range from a few
nanometers to several hundred nanometers. Here we show how BS has been widely
and successfully used to study various carbon-based materials ranging from ultra-
thin tetrahedral amorphous carbon films to carbon nanotubes. These materials are
interesting for both technological applications and fundamental research.
1 Introduction
Carbon-based materials are currently exploited, or being considered, for a
variety of applications. Suitable characterization techniques are therefore
needed in order to investigate a wide spectrum of properties, in particular
mechanical and structural properties are of paramount interest to study struc-
tural materials and films for coating applications. Brillouin scattering (BS)
permits us to characterize dynamical structural and mechanical properties of
materials through the detection of thermally excited surface and bulk acoustic
waves. This gives access to information about both surface and bulk struc-
tural properties at the mesoscale (from a few nanometers to several hundred
nanometers). Here we show how BS has been widely and successfully used
to study a large variety of carbon-based novel materials, of interest for both
technological applications and fundamental research aspects.
Inelastic scattering of visible light is a contactless, nondestructive mea-
surement technique that can be employed to investigate various types of
collective excitations in solids. Measurements are performed by illuminat-
ing a sample with a focused laser beam and analyzing the spectrum of the
scattered light. Most of the light is elastically scattered and remains at the
laser frequency, while the light that interacts with some kind of excitation
undergoes a frequency shift.
154 Marco G. Beghi et al.
In Raman scattering a photon interacts with a single optical phonon, i.e.,

an oscillation of local, atomic degrees of freedom, and thus provides infor-
mation mainly on local order at an atomic scale. In carbonaceous materials
Raman scattering involves frequency shifts in the 102 –103 cm−1 range, i.e.,
phonons of frequencies ranging from a few THz to one hundred THz. In
BS a photon interacts with an acoustic phonon having wavelength of the
same order of the optical wavelength, i.e., a bulk or surface acoustic wave of
submicron wavelength [1, 2, 3, 4, 5]. The involved frequency shifts are the fre-
quencies of the acoustic waves of such wavelengths; they depend on material
properties and on the type of acoustic mode, and typically lie in the GHz
to one-hundred GHz range, i.e., a fraction to a few cm−1 . These frequencies
are determined by the elastic properties at the wavelength scale, and thus
provide information on material properties at such a scale.
Interactions with single phonons, i.e., traveling excitations, give Stokes–
anti-Stokes doublets: couples of down-shifted and up-shifted peaks, at well-
defined frequencies. Interactions can also occur with nontraveling excitations
(quasi-elastic scattering), or involving more phonons (multiphonon Raman
scattering). Both these mechanisms give a single broad peak centered at null
frequency shift, having width typically comparable to the Brillouin shifts [6].
Other techniques to measure the characters of acoustic waves, like quan-
titative acoustic microscopy [7] and the so-called laser acoustic technique [8]
are based on the excitation of acoustic waves, respectively by a piezoelectric
transducer and by a laser pulse, and on the observation of their propagation
over millimeter distances. Such techniques typically operate in the tens to
hundreds of MHz frequency range. BS measurements do not involve the ex-
citation of acoustic waves, but exploit the spontaneous thermal fluctuations
of the acoustic modes. The amplitude of these fluctuations is small and re-
quires longer measurement times (tens of minutes up to some hours), but
the acoustic modes are observed at higher frequencies (tens of GHz) and lo-
cally. Light scattering occurs in the focusing spot of the laser and does not
imply acoustic wave propagation over macroscopic distances. These charac-
teristics qualify BS as a nondestructive, contactless testing technique that is
also applicable to small samples.
BS can occur by interaction with either bulk acoustic waves, in transpar-
ent or semitransparent media, or surface acoustic waves (SAWs), at the sur-
face of any kind of material. Bulk waves are characterized by their wavevec-
tor k, while SAWs have a wavevector k parallel to the surface. When acoustic
waves are detected, their velocity is measured: the wavevector is determined
by the scattering geometry (see Sect. 2), the circular frequency ω is measured
from the spectrum, and the acoustic velocity v = ω/k or v = ω/k is immedi-
ately obtained. In homogeneous solids the acoustic velocities are independent
of wavelength, while they depend on the mode polarization and, in anisotropic
media, on the propagation direction. At the surface of a homogeneous solid
only one type of SAW exists: the Rayleigh wave, of velocity v R . The dis-
Brillouin Light Scattering 155
placement field of this type of wave decays exponentially with depth, the
decay length being of the same order of the wavelength 2π/k . The Rayleigh
wave can be considered the prototype of the various SAW types. In layered
solids, beside a Rayleigh wave modified by the layered structure, other types
of SAWs can be present, depending on layer thickness and properties: Sezawa
waves, Love waves, pseudo surface waves [9]. Differently from homogeneous
solids, layered media have an intrinsic length scale; consequently, the veloc-
ity of all the modes supported by layered structures generally depends on
wavelength, and the dispersion relation v(k ) can be measured. In particular,
for a single supported film of thickness h on a homogeneous substrate, the
velocity of SAWs is a function of the nondimensional product k h [2, 3].
The acoustic properties of materials or of layered structures are the direct
outcome of BS measurements. The acoustic properties are of direct interest
only for specific applications, but they provide a useful tool to probe other
properties, namely the elastic properties, which are of interest in various
cases, like, e.g., for coatings. Indentation techniques directly give the hard-
ness, and instrumented indentation supplies one combination of the elastic
moduli [10, 11]. However, also in the simplest case, the isotropic medium, the
full characterization of the elastic properties requires two independent para-
meters, to be chosen among Young’s modulus E, shear modulus G, bulk mod-
ulus B, Poisson’s ratio ν, or among the elements Cij of the elastic constants
tensor. Any pair of these parameters uniquely determines all the others.
BS can be exploited to measure these parameters. This Chapter is mainly,
although not exclusively, devoted to a discussion of the measurement and data
analysis techniques. The velocity of bulk acoustic waves is easily computed as
a function of the mass density ρ and the elastic constants Cij of the material
supporting them. The velocity of SAWs in layered media can be computed
as a function of the material properties and the thickness of all the layers.
The properties of the layer(s) can be derived by fitting the computed dis-
persion relations to the measured ones. In reality the available information
is generally not sufficient to derive all the mentioned properties; the most
frequent applications concern either bulk nonmetallic samples or single sup-
ported films. In the first case the mass density is independently measured and
the elastic constants are obtained [4, 12, 13, 14]. In the second case one ex-
ploits the substrate properties and values of film thickness and mass density
independently obtained, e.g., by X-ray reflectivity measurements, to derive
the film elastic constants [5, 15, 16, 17, 18]. It can be noted that, since the
probed acoustic wavelengths are submicrometric, in films having thickness of
a few micrometers the bulk waves are fully present and measurable (if the
material is sufficiently transparent), while SAWs are entirely confined within
the film, and are unaffected by its finite thickness. In other words, from the
point of view of BS, films having thickness above a couple of microns are
indistinguishable from a semiinfinite medium [19, 20], except for the acoustic
modes having a wavevector normal to the surface [21].
Fig. 1. Scheme of the experimental set-up for Brillouin scattering experiment in

the backscattering configuration. Sp: specimen; θi : incidence angle; M: mirror; L1–
L4: lenses; P: pinhole for spatial filtering; FP: Fabry–Perot interferometer (a single
interferometer is shown, for clarity, instead of the tandem multipass interferometer
actually exploited); D: light detector
In the following sections the experimental and data analysis techniques

are first discussed. A general overview of the applications of BS to various
types of carbonaceous materials is then presented. Two specific cases are
finally reviewed in more detail to show the potential of the technique: the
measurement of the elastic constants of ultra-thin (down to 2 nm) tetrahedral
amorphous carbon films of high density and stiffness, and the characterization
of single-walled nanotubes (SWNT).
2 Experimental Technique
As mentioned above, BS measurements are performed illuminating a sam-

ple with a focused laser beam, collecting the scattered light and analyzing its
spectrum. The incidence and collection directions define the scattering geom-
etry. In the backscattering configuration the two directions coincide and are
at an angle θi from the normal to the specimen surface (Fig. 1). This config-
uration is the most common one, since it has some advantages: it maximizes
the acoustic wavevector k, and it exploits the same lens to focus the incident
laser beam and to collect the scattered light, instead of needing two separate
optics.
In the scattering event, energy conservation has the form
Ω s = Ω i ± ω, (1)
where Ω i and Ω s are the circular frequencies of the incident and scattered
light, respectively. The + and − signs refer to anti-Stokes (phonon annihi-
lation) and Stokes (phonon creation) events, respectively. The spectrum of
scattered light is dominated by the light at the same incident frequency Ω i ,
elastically scattered by static inhomogeneities. As indicated by (1) it also

contains one or more doublets, symmetric with respect to Ω i . In scattering
by bulk waves, wavevector conservation has the form
q s = q i ± k, (2)
where q i and q s are the wavevectors (in the material) of the incident and
scattered light; the sign refers again to anti-Stokes and Stokes events. In
scattering by SAWs the absence of translational invariance in the depth di-
rection leads to the nonconservation of the wavevector component in that
direction: (2) holds only for the two components parallel to the surface, in
the form
q s = q i ± k . (3)
Obviously, Ω i = (c/n) |q i | and ω = v |k|, where v is the velocity of the acoustic

wave and (c/n) is the velocity of light in the material, c being its velocity
in vacuum and n being the refractive index. Since (2) and (3) imply that
|k| and k are at most of the same order of |q i | and q i , it follows that
ω/Ω i is at most of the same order of the ratio v/(c/n) 1. This means that
Ω s Ω i and, accordingly, |q s | |q i |: the modulus |q i | = n (2π/λ0 ) being
determined by the laser wavelength λ0 , in a given experiment the probed
acoustic wavevector is fully determined by incidence and scattering direc-
tions. For backscattering by bulk waves, |q i | = 2 (2π/λ0 ) n. This marks a
difference with the other techniques to measure the acoustic properties [7, 8]:
in these techniques the experiment selects the frequency, and the wavelength
is determined by the material properties. Consequently, hard materials, which
are acoustically fast, are probed at longer wavelength. In BS experiments the
acoustic wavelength is instead selected, irrespective of the material proper-
ties, and the latter determine the acoustic frequency.
It must be remembered that Snell’s law means that upon refraction the
parallel component of the optical wavevector remains unchanged. Conse-
quently, q i and q s are completely determined by the incidence and scatter-
ing angles θi and θs , and are independent of the refractive index n. Therefore
the analysis of surface Brillouin scattering (SBS) results does not require the
value of n. In backscattering

k = 2 |q i | sin θi = 2 2π sin θi , (4)
λ0
and the relative difference |Ω s − Ω i | /Ω i = |q s − q i | /q i reaches its maximum
value (for scattering by SAWs) 2 (v SAW /c) sin θi , which is, for any material,
well below 10−4 , down to below 10−6 . A typical experimental set-up for this
scattering configuration is presented in Fig. 1. The laser beam is directed
onto the specimen by a small mirror and focused onto the specimen surface
by a lens, which also collects the scattered light. Spatial filtering by a pinhole
is typically needed, and the filtered scattered beam goes to the spectrometer.
A more complete analysis would consider the finite collection solid angle:
the lens collects all the scattered wavevectors belonging to a cone around
the nominal backscattering direction q s = −q i . Computation of the average
exchanged wavevector leads to a correction to (4) [3, 22].
The analysis of Brillouin spectra is difficult because the doublets to be
measured (1) have, as noted above, a very small relative frequency shift ω/Ω i
from the elastic peak at Ω i , whose intensity is larger by several orders of mag-
nitude. Diffraction gratings are not sufficient, and surface Brillouin spectra
became observable with the introduction of the tandem multipass Fabry–
Perot interferometer [1, 23, 24]. The interferometer is operated as a narrow-
bandwidth, high-contrast tunable band-pass filter [2, 25], and light is finally
detected by a photomultiplier, operated in the single-photon counting mode.
3 Derivation of the Elastic Constants

A Brillouin spectrum gives the velocities v = ω/k and/or v = ω/k of the
bulk and surface acoustic waves which give spectral doublets. SAWs are of
particular interest for supported films. Since the value of k is determined by
the incidence angle (4), a set of measurements at different incidence angles
allows us to measure several points of the dispersion relation v SAW (k ). The
bulk wavevector |k| is instead independent of the incidence angle (except at
most for anisotropy effects), but this is not a limitation because, since bulk
wave velocities do not depend on |k|, the measurement of a dispersion relation
is not needed.
In the simplest case, the isotropic medium, longitudinal
and transverse

bulk waves exist. Their velocities are simply given by C11 /ρ and C44 /ρ,
respectively, and the two moduli are immediately obtained if both types of
waves give measurable spectral peaks and if the mass density is known. In
an isotropic medium the whole tensor of the elastic constants is completely
determined by only two independent quantities, and the elastic behavior of
the material is thus fully characterized. In anisotropic media (single crys-
tals) the number of independent elastic constants is higher, and there are
generally three different bulk waves, because the two different transverse po-
larizations are no longer equivalent. The velocities of these waves have more
complex expressions, functions of the propagation direction. In the case of
bulk crystals the complete elastic characterization depends on the specific
crystal symmetry and typically needs measurements with different directions
of k, possibly requiring crystals with different cuts. The behavior of bulk
waves has to be kept in mind also for the analysis of films, because, as noted
in Sect. 1, for film thicknesses above about one micrometer, data analysis can
be performed as for a semi-infinite medium; this also means that in this case
a precise knowledge of film thickness is not required.
In the case of a semi-infinite medium, either isotropic or anisotropic, only

one type of SAW generally exists: the already-mentioned Rayleigh wave. It is
a true surface wave, meaning by this that its velocity is lower than that of any
bulk wave, that its displacement field vanishes at increasing depth (for the
Rayleigh wave it declines exponentially) and that its energy flow (the acoustic
Poynting vector) is strictly parallel to the surface. In the case of layered
media other types of SAWs can exist [9, 26, 27, 28]. Considering in particular
a single supported film, the Rayleigh wave generally exists, modified by the
presence of the film (modified Rayleigh wave, MRW) but preserving its main
character. If the film is acoustically slower than the substrate it can act as a
waveguide, supporting guided modes (Sezawa waves), which are reminiscent
of the acoustic modes in free-standing plates (Lamb waves) and have the
character of true surface waves. Both the Rayleigh and the Sezawa waves are
polarized in the sagittal plane, defined by the propagation direction and the
normal to the surface. Transverse waves, polarized in the surface plane (Love
waves) also exist.
Along a stress-free boundary pseudo surface waves (PSAWs) can also
exist. They do not have the character of true surface waves: their velocity is
not lower than that of any bulk wave, and they radiate part of their energy
towards the interior of the medium; accordingly, their displacement field does
not vanish at infinite depth. However, when their displacement field is still
mainly confined in the vicinity of the surface and the rate at which they
lose energy towards the bulk is low, they resemble to true SAWs, and in
BS measurement they give doublets exactly as true SAWs. Of particular
importance in the case of films acoustically faster than the substrate is the
wave called “leaky longitudinal surface wave” or longitudinal resonance: it is
a PSAW that, at the surface, is longitudinally polarized.
The velocities of all these waves can be computed, as functions of the
mass density and the elastic moduli of all the layers, and of the thickness of
the layers (of the product k h for a single supported film). If the dispersion
relation v SAW (k ) is measured for a set of values of k , obtaining a set of
values {vi meas. }, the same set of velocities can be computed as function of
some free parameter, namely the elastic moduli Cj , obtaining the values
{vi comp.(Cj )}. The free parameters Cj are then identified by a standard least
squares minimization procedure. The sum of squares is computed:
vi comp. (Cj ) − vi meas. 2
χ2 (Cj ) = , (5)
i σi
where the general case has been considered, in which the variances σi of each
vi meas. can be individually estimated. Following standard estimation theory,
the minimum of χ2 (Cj ) identifies the most probable values (C j ) of the para-
meters in the (Cj ) space, and the isolevel curves of the ratio χ2 (Cj )/χ2 (C j )
identify the confidence region at any predetermined confidence level [17, 29].
If the minimum of χ2 (Cj ) is well defined, the isolevel curves identify a lim-
ited region in the (Cj ) space, resulting in a rather precise measurement of
the Cj parameters [16, 17, 18]. In other cases the function χ2 (Cj ) has rela-
tively low values in a broad region [30]; for instance, in a two-dimensional
(Cj ) space, it can have the shape of an elongated valley [31]. In such cases
precise measurements of the Cj parameters are not achieved, although in
some cases a combination of the parameters exists, which is identified with a
better precision than the individual parameters [18].
The information that can be obtained on the Cj moduli depends obviously
on the amount of information available from the dispersion relations. If only
the modified Rayleigh wave is observed, the Cj parameters are generally not
identified with good precision, but if further acoustic modes are observed, ei-
ther Sezawa waves [32] or longitudinal guided modes [33, 34], the precision of
the measurement of the Cj moduli increases significantly [15,16,18,29,31,35].
In particular, in isotropic materials the longitudinal guided mode has a
wavevector parallel to the surface, and the same velocity of the longitudi-
nal bulk wave: without needing the value of the refractive index, it directly
supplies the value of the C11 elastic constant [33].
The information that can be derived when only the modified Rayleigh
wave is detected was assessed by a sensitivity analysis [36]. For the repre-
sentative model case of an isotropic film on a silicon substrate, it was found
that the modified Rayleigh wave velocity is indeed sensitive to the values of
the Young modulus E and the shear modulus G, but very slightly sensitive
to the Poisson’s ratio ν, the bulk modulus B and the C11 elastic constants.
This means that both E and G can be determined with good precision, and
that a similar precision cannot be achieved for ν, B and C11 . This is not con-
tradictory with the fact, remembered above, that in an isotropic medium the
values of two elastic constants are sufficient to determine all the others. The
moduli ν, B and C11 , viewed as functions of E and G, are functions charac-
terized by steep slopes. Consequently, even a small region in the (E, G) plane,
corresponding to a good identification of these two parameters, corresponds
to wide intervals of ν, B and C11 , corresponding to a poor identification of
their values. For Poisson’s ratio ν the interval can be numerically not very
wide, but easily covers a substantial fraction of the range (0, 0.5) which is
the meaningful range for this parameter. The same analysis showed that the
sensitivity to the film thickness and mass density is not high. This is an
interesting result, as it means that, for the precise measurement of the elas-
tic moduli, extremely accurate values of thickness and mass density are not
crucial.
It must be remembered that not all the (E, G) plane is physically mean-
ingful. Since ν = E/2G − 1, in the (E, G) plane the ν = const lines are
straight lines through the origin, and the ν = 0.5 and ν = 0 lines are the
G = E/3 and G = E/2 lines. These lines delimit the triangle that is fully
meaningful. The condition ν ≤ 0.5 is required for thermodynamic stability;
the condition ν ≥ 0 is not strictly required (thermodynamic stability only
requires ν ≥ −1), but is violated only in extremely rare and peculiar cases.
4 Brillouin Scattering from Carbonaceous Materials

BS has been extensively used to characterize carbonaceous materials, al-
though in several cases their scattering cross section seems to be relatively
low. In the case of diamond this possible limitation is overcome by the high
thermal conductivity, by which the samples can withstand a high incident
laser power without being damaged. In other cases long measurement times
allow researchers to reach a sufficient counting statistics also with weak sig-
nals, but for some types of “diamond-like carbon” the scarcity or absence of
published BS measurements is probably to be ascribed to the low scattering
cross section. A synthetic overview of published BS results is given here.
In crystalline graphite the elastic constants had already been measured
by ultrasonic techniques, but these techniques had proven to be unable to
reliably measure C44 , due to dislocation motion at the achievable ultrasonic
frequencies. Attempts to avoid this problem had been made by performing
measurements on graphite irradiated to pin the dislocations. BS avoids this
problem by operating in a much higher frequency range. The values of C44
obtained by BS by different authors [37, 38] are in reasonable agreement
among them, and similar to the values obtained from neutron scattering, but
significantly higher than those found by ultrasonic techniques. The remaining
difference among the BS results can be ascribed to crystallite size effects [39].
If the crystallites are not much larger than the acoustic wavelength, and
the measurement spot encompasses several, not perfectly aligned crystallites,
their boundaries have a significant influence on propagation properties [38].
If instead the crystallites are much larger than the the acoustic wavelength
and the measurement spot, the boundaries have no effect [37, 39].
In crystalline diamond the most accurate measurements were performed
by Grimsditch’s group, who also reviewed previous works [4, 13]. They mea-
sured with great precision the elastic constants of natural composition dia-
mond and isotopically enriched diamond [13], finding, for natural composition
diamond, C11 = 1080.4 GPa, C12 = 127 GPa and C44 = 576.6 GPa. These
values correspond to a bulk modulus B = 444.8 GPa. The temperature de-
pendence of these moduli was also measured [14], finding a decrease of only
8% for temperatures up to 1600 K.
Various measurements have been performed on CVD polycrystalline di-
amond films. Measurements on 400 µm thick films, in which the crystallite
size is much larger than wavelength, give values of the elastic moduli in
agreement with those found for crystalline diamond, with a Rayleigh velocity
v R = 10326 ± 470 m/s, consistent with the (110) texture of the film [40].
Measurements on different films give results that depend significantly on
film quality [20]. Exploiting a peculiar Brillouin scattering configuration and
known values of the stress coefficients for diamond single crystals, Krüger
et al. were able to evaluate the internal stresses in CVD diamond [41]. They
found stresses up to several GPa, without a clear difference between the nu-
cleation side (nanocrystalline) and the growth side (microcrystalline).
The ability of BS to perform measurements on submillimeter-sized sam-

ples was exploited to measure the elastic constants for two types of materials
that can be obtained only in small samples: fullerite fcc crystals [19] and the
hard amorphous carbon phases which can be synthetized, at high tempera-
tures and pressures, from fullerite [42]. In the latter case BS found, for both
bulk acoustic waves and SAWs, velocities that approach those of diamond,
up to around 90% of diamond values [42].
Hydrogenated amorphous carbon films were characterized by BS [43].
Measurements performed in conjunction with nanoindentation [44] were an-
alyzed with an a priori assumption about the value of Poisson’s ratio, while
in other cases [34] such an assumption was not needed. Films of non-
hydrogenated amorphous carbon, with different fractions of sp3 bonding,
were characterized by various authors [33, 34, 45]. In these works, besides
the Rayleigh wave, other surface or pseudo surface acoustic modes could be
measured. For films acoustically slower than the substrate it was a Sezawa
wave [45], while for films acoustically faster than the substrate it was a pseudo
surface mode, the so-called longitudinal guided mode [33, 34]. In both cases
the additional information coming from the second mode allowed a better
determination of the elastic constants. Ultra-thin films (down to 2 nm) of
tetrahedral amorphous carbon films of high density and stiffness only perturb
the Rayleigh wave of the substrate, without adding other modes; nevertheless
this perturbation could be measured, allowing researchers to characterize the
ultra-thin films, as reviewed in the next section.
The characterization by BS of low-density nanostructured carbon films
deposited by low-energy cluster beam deposition was also possible [46], de-
spite the relatively high porosity and roughness of such films. At least for
some types of such films acoustic phonon propagation is supported, and their
detection is possible for films deposited on highly reflective aluminum sub-
strates.
Single-walled nanotubes (SWNT) are a peculiar type of carbon material.
BS from aligned nanotubes could be detected, allowing a characterization of
the graphene sheet forming the tube wall, as reviewed below.
5 Ultra-Thin Carbon Films

Ultra-thin diamond-like carbon films, of thickness down to 2–5 nanometers,
have been developed as protective coatings of magnetic hard disks and the
corresponding reading/writing head. The distance between the recording
head and the magnetic material has to be decreased, in order to further
increase the storage density [47], and a reduction of the thickness of the
protective coatings of the disk and the head is required. Carbon coatings
are already a current choice to provide protection against corrosion, graft-
ing sites for the lubricant molecules and mechanical protection [47]. In order
to achieve a storage density of 100 Gbit/in2 , coatings having thickness re-

duced to 2–5 nanometers, but maintaining the peculiar properties of thicker
diamond-like films, are needed; coatings of even lower thickness are currently
being investigated. The characterization of such thin films is a challenging
task. The measurements achieved by BS are discussed here. With thicker
films the influence of film properties on the modified Rayleigh wave velocity
is higher: several cases in which the film properties could be determined were
mentioned above. Ultra-thin films induce instead only a small perturbation of
the wave velocity, which remains mainly determined by the substrate prop-
erties. The ability to detect these perturbations and to interpret them is a
good proof of the potential of the technique.
Ultra-thin carbon films are amorphous, and therefore isotropic: as already
noted, only two independent parameters are sufficient to fully characterize
their stiffness. For ultra-thin films their measurement becomes difficult. Me-
chanical characterization is often performed by instrumented indentation,
which directly gives the hardness, and one combination of the elastic mod-
uli [10,11]. However, when the film thickness is reduced to tens of nanometers
or even less, the requisite that indentation depth be significantly smaller than
film thickness cannot be met, and the results are determined by the substrate
rather than by the film. Other means have then to be exploited to measure
the elastic moduli. As discussed above, SBS, allowing the measurement of
acoustic waves of wavelength down to a quarter of a micron, is a precious
tool.
We review here BS measurements performed on tetrahedral amorphous
carbon (ta-C) films deposited using an S-Bend Filtered Cathodic Vacuum
Arc (FCVA) with an integrated off-plane double bend (S-Bend) magnetic
filter; the deposition chamber was evacuated to 10−8 Torr. This deposition
method proved able to deposit ta-C films with particle area coverage of less
than 0.01% and uniform thickness [48]. Samples of thickness ranging from
2 to 8 nm were considered, together with a bare silicon substrate, which is in
reality covered by a thin layer of native oxide due to exposure to air.
The thickness and mass density of the films were obtained by X-ray reflec-
tivity measurements, performed on a Bede GXR1 reflectometer with CuKβ
radiation. Reflectivity curves for supported films typically consist of a critical
angle peak, followed by a long decay. Superposed on this decay are oscilla-
tions (interference fringes) due to the interference of waves reflected from the
surface and the substrate interface. Usually, the density is directly derived
from the critical angle [48]. However, for thickness below about 20 nm the
X-ray evanescent wave reaches the substrate; consequently, the critical angle
turns out to be determined by the substrate rather than by the film. The film
density must then be derived by a fit of the whole experimental reflectivity
curve to a simulated scattering curve computed from model structures [49];
this was performed by the Bede REFS-MERCURY code [50]. For low thick-
ness the interference fringes on the decay become wider, and their number is
reduced. However, comparison of the reflectivity profiles from the bare sili-
con and from the coated samples shows evidence of at least one broad fringe,
from which the thickness can be measured.
It is known from measurements on thicker ta-C films [51] that in this kind
of films the properties can have a gradient in proximity of the outer surface
and the inner interface, the mass density declining when approaching the
two surfaces. Also in these nanometric films the X-ray measurements gave
indications of such a gradient: the fit to the experimental reflectivity curve
improved if a three-layer model was considered, with thin layers of lower
density on the two sides of the main carbon film. However, adopting such a
three-layer model the individual layer properties could not be identified with
good precision. Furthermore, characterization of the elasticity of a three-layer
model requires six independent parameters, and their identification would
remain an undetermined problem. In order to avoid these complexities, both
the X-ray and SBS results were interpreted in a consistent way, modeling the
films as single homogeneous equivalent films, with a sharp interface with the
substrate.
SBS measurements were performed at room temperature exploiting the
514.5 nm line of an Argon ion laser and a 3 + 3 passes tandem interferome-
ter [2], the type of spectrometer most commonly adopted in Brillouin spec-
trometry. Measurements were performed in backscattering, with incidence
angle ranging from 30◦ to 70◦ . SAW propagation was along the [100] direc-
tion on the (001) face of the Si substrate. The incident light was p-polarised,
the scattered light being collected without polarisation analysis. The incident
power on the specimen was around 100 mW, focused into a spot of the order
of 103 µm2 ; irradiation did not induce film modifications.
Films of such low thickness do not alter the spectrum of SAWs of the
substrate: the Rayleigh wave undergoes a slight modification of its velocity,
but remains the only type of SAW supported by this structure. As indi-
cated in Sect. 3, its velocity was computed [9] as a function of the elas-
tic constants and mass density of both the film and the substrate, of film
thickness and of the wavevector k . The anisotropy of crystalline silicon was
fully taken into account; accepted values of its properties (C11 = 166 GPa,
C12 = 63.9 GPa, C44 = 79.6 GPa, ρ = 2.33 g/cm3 [52]) were adopted. The
film thickness and mass density were obtained by X-rays, as indicated above,
and the SAW wavevector k was determined by the laser wavelength and
the scattering geometry. The Rayleigh wave velocity remained thus a func-
tion of the unknown elastic constants of the film. The film being isotropic,
the elastic properties were represented, as indicated in Sect. 3, by the (E, G)
couple of moduli. The velocity for each value of k is computed (see (5)) as
vi comp. (E , G), obtaining a sum of squares χ2 (E, G) whose minimum identifies
the most probable values of E and G, and whose isolevel curves identify the
confidence regions.
Fig. 2. 90%, 95% and 99% confidence regions, having a valley shape, obtained
fitting DLC films dispersion relation in the E, G plane. Continuous lines: ν = 0
and ν = 0.5 physical limits and ν = 0.1, 0.2, 0.3, 0.4 isolevel lines; dashed lines:
B = 100, 200, 300, 445 GPa isolevel lines. The line B = 445 GPa = B diam is taken
as the physical upper bound delimiting the acceptable part of the confidence regions
The χ2 (E, G) function turns out, in this case, to have the shape of a
valley that extends from the neighborhood of the ν = 0.5 line into the region
in which ν < 0. The bottom of this “valley” is relatively flat and does not
reliably identify a small region in the (E, G) plane; instead, it identifies a
“strip” running through wide ranges of E and G, and also of B and ν. In order
to delimit the range of acceptable results, only the part of any confidence
region is considered that corresponds to positive values of Poisson’s ratio
(see Sect. 3) and to values of bulk modulus B below a predetermined upper
limit. The bulk modulus of diamond, B diam = 445 GPa, was adopted as
physically plausible upper limit for B.
The confidence regions run across the region delimited by these accept-
ability limits, meaning that ν and B remain essentially undetermined, while
the part of the confidence region compatible with these limits corresponds
to reasonably well defined intervals of E and G. The acceptable part of the
confidence regions must be seen as the meaningful information that can be ob-
tained from the SBS measurements. Further resolution within the confidence
regions does not have real significance. The values of the elastic constants
corresponding to the center of the 90% confidence region allow us to fit the
measured dispersion relations, as shown in Fig. 2. The dispersion relations
computed by the most probable parameter values are compared to the mea-
sured ones in Fig. 3.
Fig. 3. Measured and computed dispersion relations for six thin films on silicon
substrate. Dispersion relations presented as function of the nondimensional product
k h are independent from film thickness and depend only on film properties. From
bottom to top: ∗ and ×, two films of native silicon oxide of different thickness and
indistinguishable properties (they fall on the same line); and , two DLC films of
3.5 nm and 4.5 nm thickness, of very similar properties (they fall on nearby lines);
◦, DLC film of 15 nm thickness. , DLC film of 8 nm thickness, stiffer than all the
other films
6 Nanotubes
Single-wall carbon nanotubes (SWNT) can be thought as formed by warping
a single graphitic layer into a cylindrical object [53]. Beyond possessing unique
electronic properties, depending on the symmetry, or chirality, of the carbon
lattice, nanotubes also attract much attention because they are expected to
be very strong and to have high elastic moduli.
The Young’s modulus of a carbon nanotube was estimated observing by
transmission electron microscopy (TEM) the vibrations of the free end of
a nanotube clamped at the other end [54], and by observing the nanotube
bending by an atomic force microscope (AFM) [55]. These estimates (val-
ues up to 1.25 TPa) are consistent with the exceptionally high values of the
Young’s modulus measured for a graphene sheet, i.e., about 1 TPa; such val-
ues imply very high stiffness-to-weight ratios. Acoustic vibrational properties
are strictly related to elastic properties of SWNT, and acoustic vibrational
modes in carbon nanotubes have been calculated both by ab initio calculation
(lattice dynamics, e.g., [56]) and by continuum elastic models [57]. AFM and
TEM measurements, beside being destructive, suffer from lack of precision in
determining the SWNT diameter, length, temperature and vibrational fre-
Fig. 4. Left: AFM image of the (top) oriented SWNT sample surface, and of the
(bottom) nonoriented SWNT sample surface. Square area is 1 µm × 1 µm, z range
is 1 nm. Right: scattering geometry with respect to the nanotube direction
quency; furthermore, the tendency of SWNT to form bundles complicates the

analysis.
Dynamic light scattering from longitudinal acoustic modes in aligned
SWNT, with a wavelength of the order of hundreds of nm, has been measured
by means of Brillouin spectroscopy; a model for the description of the phe-
nomenon and for the calculation of the scattered intensity was also proposed.
The model is based on a shell mechanical model and a surface elasto-optic
effect coupling the incident light to the tube thermal strains [58,59]. The sam-
ple was a free-standing film of aligned SWNT, produced by the laser-oven
method, purified and aligned with a strong magnetic field [60]. SWNT are
present in bundles or ropes, with a preferential orientation, as shown in Fig. 4.
SWNT have instead a complete random orientation in nonaligned samples.
The typical tube free length is on the order of several hundreds nanometer.
BS measurements were performed exploiting the λ0 = 514.5 nm line of an
Ar+ ion laser. The incident light was p-polarized, with a power of 40 mW;
the scattered light, collected without polarization analysis, was analyzed by
a Sandercock interferometer with a free spectral range of 100 GHz. Mea-
surements were performed in backscattering, with the scattering geometry
shown in Fig. 4: the scattering plane (determined by the scattering wavevec-
tor k = q s − q i and the normal z to the sample surface), the sample surface
plane and the nanotube direction x ; α is the angle between the SWNT direc-
tion and the scattering plane xz; θi is the incidence angle, between incident
light q i and z; β is the angle between x and q i (cos β = sin θi cos α). The
incidence angle θi was varied in the range 30◦ –70◦ in order to vary the scat-
tering wavevector k; measurements were performed with α = 0◦ and α = 90◦
(scattering plane parallel and perpendicular to the nanotube direction).
Fig. 5. (a) Brillouin spectra for the oriented SWNT sample along the tube direction
(α = 0◦ ; θi = 30◦ , 70◦ ), and perpendicularly to the tube direction (α = 90◦ ;
θi = 30◦ ). (b) Brillouin spectrum of a nonaligned SWNT (θi = 50◦ )
The experimental results show that scattering does not occur from SAWs
as in thin films, but from longitudinal acoustic modes of the single tube, hav-
ing a wavevector k ph parallel to the tube axis. In spectra from the oriented
SWNT sample (θi = 30◦ , 70◦ ), with α = 0◦ , a broad peak is detected at
about 45 GHz (Fig. 5), and its position is independent of the incidence angle.
When the scattering plane is instead perpendicular to the SWNT direction
(α = 90◦ ), the feature disappears (Fig. 5). A spectrum from a randomly
oriented SWNT sample is also shown (θi = 50◦ ), with a broad feature at
about 40 GHz. This feature does not change as the sample is rotated by 90◦
(α = α + 90◦ ), showing the in-plane isotropy of the film. The absence of
scattering from aligned tubes for α = 90◦ and the peak shift and width for
α = 0◦ confirm the relative independence between the tube bundles, and that
the material is not an effective medium where acoustic waves can propagate
in all directions, at least at the scale of the laser wavelength λ0 .
A cylindrical shell model provides an adequate description for this scatter-
ing phenomenon. For a long cylindrical thin shell the elastodynamic equations
can be written in cylindrical coordinates (x , r, φ) and a membrane approx-
imation can be used [61]. This model, described by 2D stress conditions and
membrane elastic constants having the dimension of a force per unit length,
is precise enough to describe thermally excited modes along the tube axis,
with frequency in the range 1–100 GHz, responsible for the observed BS. The
elastodynamic equations admit plane wave solutions traveling along the axial
direction with wavevector k ph , i.e., ux = A u0 exp[i(k ph x − ωt + mφ)], with
m an integer. The mode frequency ω depends on the 2D Young’s modulus
E 2D , the surface density ρ2D , the Poisson’s ratio ν, and the SWNT radius R.
For an infinitely long tube these are the appropriate solutions; for tubes of
finite length L the proper boundary conditions affect the form of the solution
and the corresponding frequencies. A surface elasto-optic effect is assumed,
coupling the light to the tube thermal phononic strains and producing a
fluctuating surface polarization density that radiates the scattered photons.
The intensity of the light scattered by a single tube is related to the Fourier
transform of the dielectric susceptibility tensor and to the polarization of the
incident light.
It results from the computed cross section that, in backscattering, only
longitudinal modes with m = 0 contribute to the scattered intensity, provided
that the component k = (4π/λ0 ) cos β = (4π/λ0 ) sin θi cos α of the scattered
wavevector k parallel to the tube is not zero (there must be a nonvanish-
ing component of transferred wavevector, i.e., the scattering plane must not
be perpendicular to the SWNT orientation). In the case of an infinite tube
(L 2π/k ph ) the cross section is nonzero only if the parallel wavevector is
conserved (k = k ph ). This would be a coherent “line Brillouin scattering”,
with only one longitudinal wave (determined by the selected value of k ph )
contributing to the scattered intensity, and the corresponding peak position
depending on the incidence angle (dispersion of the modes). Instead, in the
measurements published in [58], the θi dependence of the peak position is
negligible, because the free SWNT length is of the order of 1 µm or less:
the tubes cannot be considered as infinite with respect to the wavelength
(λ = λ0 /2 cos β L) of the modes selected by BS. Thus, all longitudinal
modes (fixed end boundary conditions) with m = 0 contribute to scattering,
due to confinement effects, with relative weights given by the computation.
This kind of scattering is often called low-frequency Raman scattering. Con-
tributions from tubes of different lengths (0.5–1.5 µm) were considered, and
assuming ρ2D = 7.6 × 10−7 kg/m2 , R = 6 × 10−10 (by Raman spectroscopy
measurements) and ν = 0.2, a feature with a shape similar to that observed
was obtained, independent of the scattering angle. The broadening of the peak
can be related to length and orientation distributions, tube interactions, and
real boundary conditions. The 2D Young’s modulus can be estimated by fit-
ting the computed peak position to the measured one and a value E 2D = 100
N/m is obtained, on the order of magnitude of the C–C atomic force constant
in a graphite plane [62]. Taking into account the thickness of a graphite layer,
the estimated E 2D value corresponds to a Young’s modulus for the curved
graphitic sheet of the order of 1 TPa, compatible with known values of C11
for graphite [37].
7 Conclusions
We have presented Brillouin scattering (BS) as a nondestructive, contact-
less, powerful characterization technique capable of studying structural and
mechanical properties of bulk materials, thin films and nanostructures. In
particular elastic properties are accessible through the observation of ther-
mally excited acoustic modes propagating in the medium. The experimental
technique and the derivation of the elastic constants have been described to-
gether with some results on different carbon materials in order to show the
capabilities of this technique. In conclusion, BS turns out to be an interesting
tool which proved to be able to characterize a wide variety of materials, and
namely carbon-based materials.
References
[1] J. R. Sandercock: Trends in Brillouin scattering, in M. Cardona (Ed.): Light
Scattering in Solids III, vol. 31, Top. Appl. Phys. (Springer, Berlin, Heidelberg
1982) p. 173 154, 158
[2] F. Nizzoli, J. R. Sandercock: Surface Brillouin scattering from phonons, in
G. K. Horton, A. A. Maradudin (Eds.): Dynamical Properties of Solids, vol. 6
(Elsevier 1990) pp. 281–335 154, 155, 158, 164
[3] P. Mutti, C. E. Bottani, G. Ghislotti, M. Beghi, G. A. D. Briggs, J. R. Sander-
cock: Surface Brillouin scattering - Extending surface wave measurements to
20 ghz, in Advances in Acoustic Microscopy (Plenum, New York 1995) p. 249
154, 155, 158
[4] M. Grimsditch: Brillouin scattering, in M. Levy, H. E. Bass, R. R. Stern (Eds.):
Handbook of Elastic Properties of Solids, Liquids and Gases, vol. 1 (Academic,
New York 2001) p. 331 154, 155, 161
[5] J. Comins: Surface Brillouin scattering, in M. Levy, H. E. Bass, R. R. Stern
(Eds.): Handbook of Elastic Properties of Solids, Liquids and Gases, vol. 1
(Academic, New York 2001) p. 349 154, 155
[6] H. E. Jackson, R. T. Harley: Phys. Rev. Lett. 54, 459 (1985) 154
[7] P. V. Zinin: Quantitative acoustic microscopy of solids, in M. Levy, H. E. Bass,
R. R. Stern (Eds.): Handbook of Elastic Properties of Solids, Liquids and Gases,
vol. 1 (Academic, New York 2001) pp. 187–226 154, 157
[8] A. Lomonosov, A. P. Mayer, P. Hess: Laser controlled surface acoustic waves,
in M. Levy, H. E. Bass, R. R. Stern (Eds.): Handbook of Elastic Properties
of Solids, Liquids and Gases, vol. 1 (Academic, New York 2001) pp. 137–186
154, 157
[9] G. W. Farnell, E. L. Adler: Elastic wave propagation in thin layers, in
W. P. Mason, R. N. Thurston (Eds.): Physical Acoustics, vol. 9 (Academic,
New York 1972) pp. 35–127 155, 159, 164
[10] G. M. Pharr, D. L. Callahan, S. D. McAdams, T. Y. Tsui, S. Anders,
J. W. Ager, I. G. Brown, C. S. Bhatia, S. R. P. Silva, J. Robertson: Appl.
Phys. Lett. 68, 779 (1996) 155, 163
[11] T. A. Friedmann, J. P. Sullivan, J. A. Knapp, D. R. Tallant, D. M. Follstaedt,
D. L. Medlin, P. B. Mirkarimi: Appl. Phys. Lett. 71, 3820 (1997) 155, 163
[12] R. Vacher, L. Boyer: Phys. Rev. B 6, 639 (1972) 155

[13] R. Vogelgesang, A. K. Ramdas, S. Rodriguez, M. Grimsditch, T. R. Anthony:
Phys. Rev. B 54, 3989 (1996) 155, 161
[14] E. Zouboulis, M. Grimsditch, A. K. Ramdas, S. Rodriguez: Phys. Rev. B 57,
2889 (1998) 155, 161
[15] P. Zinin, M. H. Manghnani, S. Tlechev, V. Askarpour, O. Lefeuvre,
A. G. Every: Phys. Rev. B 60, 2844 (1999) 155, 160
[16] J. D. Comins, A. G. Every, P. R. Stoddart, X. Zhang, J. C. Crowhurst,
G. R. Hearne: Ultrasonics 38, 450 (2000) 155, 160
[17] M. G. Beghi, C. E. Bottani, R. Pastorelli: High accuracy measurement of
elastic constants of thin films by surface Brillouin scattering, in C. Muhlstein,
S. B. Brown (Eds.): Mechanical Properties of Structural Films, ASTM STP
1413 (ASTM, Conshohoken, PA 2001) pp. 109–126 155, 159, 160
[18] X. Zhang, P. R. Stoddart, J. D. Comins, A. G. Every: J. Phys.: Condens.
Matter 13, 2281 (2001) 155, 160
[19] D. Fioretto, G. Carlotti, G. Socino, S. Modesti, C. Cepek, L. Giovannini,
O. Donzelli, F. Nizzoli: Phys. Rev. B 52, R8707 (1995) 155, 162
[20] P. Djemia, C. Dugautier, T. Chauveau, E. Dogheche, M. I. de Barros, L. Van-
denbulcke: J. Appl. Phys. 90, 3771 (2001) 155, 161
[21] X. Zhang, R. Sooryakumar, A. G. Every, M. H. Manghnani: Phys. Rev B 64,
81402 (2001) 155
[22] P. R. Stoddart, J. C. Crowhurst, A. G. Every, J. D. Comins: J. Opt. Soc. Am.
B 15, 2481 (1998) 158
[23] J. R. Sandercock: Opt. Commun. 2, 73 (1970) 158
[24] S. M. Lindsay, M. W. Anderson, J. R. Sandercock: Rev. Sci. Instrum. 52, 1478
(1981) 158
[25] J. M. Vaughan: The Fabry–Perot Interferometer (Adam Hilger, Bristol, UK
1989) 158
[26] G. W. Farnell: Types and properties of surface waves, in A. A. Oliner (Ed.):
Acoustic Surface Waves (Springer, Berlin, Heidelberg 1978) pp. 15–60 159
[27] A. H. Nayfeh: Wave propagation in layered anisotropic media, in , North-
Holland Series in Applied Mathematics and Mechanics 39 (Elsevier, Amster-
dam, New York 1995) 159
[28] A. G. Every: The elastic properties of solids: static and dynamic principles,
in M. Levy, H. E. Bass, R. R. Stern (Eds.): Handbook of Elastic Properties of
Solids, Liquids and Gases, vol. 1 (Academic, New York 2001) p. 3 159
[29] O. Lefeuvre, W. Pang, P. Zinin, J. D. Comins, A. G. Every, G. A. D. Briggs,
B. D. Zeller, G. E. Thompson: Thin Solid Films 350, 53 (1999) 159, 160
[30] X. Zhang, J. D. Comins, A. G. Every, P. R. Stoddart: Int. J. Refr. Metals and
Hard Mater. 16, 303 (1998) 160
[31] S. Makarov, E. Chilla, H. J. Frölich: J. Appl. Phys. 78, 5028 (1995) 160
[32] M. G. Beghi, C. E. Bottani, P. M. Ossi, T. A. Lafford, B. K. Tanner: J. Appl.
Phys. 81, 672 (1997) 160
[33] M. Chirita, R. Sooryakumar, H. Xia, O. R. Monteiro, I. G. Brown: Phys. Rev.
B 60, 5153 (1999) 160, 162
[34] A. C. Ferrari, J. Robertson, M. G. Beghi, C. E. Bottani, R. Ferulano, R. Pas-
torelli: Appl. Phys. Lett. 75, 1893 (1999) 160, 162
[35] T. Wittkowski, J. Jorzick, K. Jung, B. Hillebrands: Thin Solid Films 353, 137
(1999) 160
[36] R. Pastorelli, S. Tarantola, M. G. Beghi, C. E. Bottani, A. Saltelli: Surf. Science
468, 37 (2000) 160
[37] M. Grimsditch: J. Phys. C: Solid State Phys. 16, L143 (1983) 161, 169
[38] S. A. Lee, S. M. Lindsay: Phys. Stat. Sol. B 157, K83 (1990) 161
[39] M. Grimsditch: Phys. Stat. Sol. B 193, K9 (1996) 161
[40] X. Jiang, J. V. Harzer, B. Hillebrands, C. Wild, P. Koidl: Appl. Phys. Lett.
59, 1055 (1991) 161
[41] J. Krüger, J. Embs, S. Lukas, U. Hartmann, J. Brierley, C. Beck, R. Jiménez,
P. Alnot, O. Durand: J. Appl. Phys. 87, 74 (2000) 161
[42] M. H. Manghnani, S. Tkachev, P. V. Zinin, X. Zhang, V. V. Brazhkin,
A. G. Lyapin, I. A. Trojan: Phys. Rev. B 64, 121403 (2001) 162
[43] X. Jiang: Phys. Rev. B 43, 2372 (1991) 162
[44] X. Jiang, J. W. Zou, K. Reichelt, P. Grünberg: J. Appl. Phys. 66, 4729 (1989)
162
[45] T. Wittkowski: Thin Solid Films 368, 216 (2000) 162
[46] C. Casari, A. L. Bassi, C. E. Bottani, E. Barborini, P. Piseri, A. Podestà,
P. Milani: Phys. Rev. B 64, 85417 (2001) 162
[47] J. Robertson: Thin Solid Films 383, 81 (2001) 162
[48] A. C. Ferrari, A. LiBassi, B. K. Tanner, V. Stolojan, J. Yuan, L. M. Brown,
S. E. Rodil, B. Kleinsorge, J. Robertson: Phys. Rev. B 62, 11089 (2000) 163
[49] B. K. Tanner, A. Libassi, A. C. Ferrari, J. Robertson: Mat. Res. Soc. Symp.
Proc. 675, W11.7 (2001) 163
[50] M. Wormington, I. Pape, T. P. A. Hase, B. K. Tanner, D. K. Bowen: Phil.
Mag. Letts. 74, 211 (1996) 163
[51] A. C. Ferrari, J. Robertson, R. Pastorelli, M. G. Beghi, C. E. Bottani: Mat.
Res. Soc. Symp. Proc. 593, 311 (2000) 164
[52] H. J. McSkimin, P. Andreatch: J. Appl. Phys. 35, 3312 (1964) 164
[53] M. S. Dresselhaus, G. Dresselhaus, P. C. Eklund: Science of Fullerenes and
Carbon Nanotubes (Academic, New York 1996) 166
[54] M. M. J. Treacy, T. W. Ebbesen, J. M. Gibson: Nature 381, 678 (1996) 166
[55] J.-P. Salvetat, G. A. D. Briggs, J.-M. Bonard, R. R. Bacsa, A. J. Kulik,
T. Stockli, N. A. Burnham, L. Forro: Phys. Rev. Lett. 82, 944 (1999) 166
[56] V. N. Popov, V. E. van Doren, M. Balkanski: Phys. Rev. B 59, 8355 (1999)
166
[57] D. Kahn, K. W. Kim, M. A. Stroscio: J. Appl. Phys. 89, 5107 (2001) 166
[58] C. E. Bottani, A. L. Bassi, M. G. Beghi, A. Podestà, P. Milani, A. Zakhidov,
R. Baughman, D. A. Walters, R. E. Smalley: Phys. Rev. B 67, 155407 (2003)
167, 169
[59] A. L. Bassi, M. G. Beghi, C. S. Casari, C. E. Bottani, A. Podestà, P. Milani,
A. Zakhidov, R. Baughman, D. A. Walters, R. E. Smalley: Diam. Relat. Mater.
12, 806 (2003) 167
[60] D. A. Walters, M. J. Casavant, X. C. Qin, C. B. Huffman, P. J. Boul, L. M. Er-
icson, E. H. Haroz: Chem. Phys. Lett. 338, 14 (2001) 167
[61] S. Markus: The Mechanics of Vibrations of Cylindrical Shells, Studies in Ap-
plied Mechanics (Elsevier, New York 1998) 168
[62] C. Mapelli, C. Castiglioni, G. Zerbi, K. Müllen: Phys. Rev. B 60, 12710 (1999)
169
Index
ta-C, 162, 163 graphite, 161
acoustic phonon, 154 hydrogenated amorphous carbon, 162

acoustic waves, 153
acoustic wave velocity, 154, 155, 157, inelastic scattering, 153
159, 160, 163
surface acoustic waves (SAWs), 154, light scattering, 154
155, 157–159, 162, 164, 168 low-frequency Raman, 169
pseudo surface waves (PSAWs), 159
SAW velocity, 154, 155, 159 mechanical properties, 153
nanostructured carbon, 162
Brillouin scattering, 153–159, 161–164, nanotubes, 166
167–170 single-walled nanotubes (SWNT),
surface Brillouin scattering, 157, 166, 167
163–165 single-walled nanotubes (SWNTs),
156, 162, 167–169
diamond, 161
diamond-like carbon (DLC), 162, quasi- elastic scattering, 154
164–166
dispersion, 166 Raman scattering, 154
Rayleigh wave, 154, 155, 159, 162, 164
elastic constants, 155, 156, 158, modified Rayleigh wave, 159, 160
160–162, 164, 165, 168 modified Rayleigh wave velocity,
elastic moduli, 155, 159–161, 163, 166 160, 163
elastic properties, 154, 155, 170 Rayleigh wave velocity, 154, 161, 162,
elastic scattering, 153 164
fullerite, 162 Sezawa wave, 155, 159, 160, 162
graphene, 162 ultra-thin carbon films, 156, 162–166

Electrical Resistivity and Real Structure
of Magnetron-Sputtered Carbon Films
Alexei A. Onoprienko
Institute for Problems of Materials Science, National Academy of Sciences of

Ukraine, 3 Krzhyzhanovsky St., UA-03142 Kiev, Ukraine
onopr@materials.kiev.ua
Abstract. The resistivity and structure of DC magnetron-sputtered carbon films

is studied. The film structure is examined by Raman spectroscopy, transmission
electron microscopy, and electron diffraction. The films reveal prominent resistivity
anisotropy relating to film structure. In the range of deposition temperatures 20–
450◦ C, ordered regular aromatic rings and graphite-like clusters form. In the range
450–650◦ C, the growth mechanism changes and graphite phase forms directly on
the substrate. Annealing of films in the range 50–300◦ C results in transformation of
distorted aromatic rings into regular ones, and in ordering them without formation
of graphite fragments. Further increase in the annealing temperature causes little
changes the film structure. Ion bombardment in the range of substrate bias voltage
−20 ≤ V bias ≤ −150 V decreases cluster size and results in disordering in their
internal structure. At V bias > −150 V the cluster structure tends to be more
ordered towards graphitization. The results obtained open the way for deposition
of carbon films with controllable electrical properties.
1 Introduction
The great interest in diamond and diamond-like carbon (DLC) films as
promising materials for microelectronics stems from their specific electrical
properties. However, when using diamond films, problems arise connected
with their polycrystalline structure. Diamond films have a well-developed
net of grain boundaries that affect the electrical properties of films. In this
respect DLC or amorphous carbon (a-C) films are preferable since they are
smooth and do not contain well-developed grain boundaries.
In many studies [1, 2, 3, 4, 5, 6, 7, 8, 9], it has been shown that the substrate
temperature and energy of carbon species constituting the growing film are
the parameters that govern the structure and properties of carbon films the
most. Depending on the deposition method and conditions, the carbon films
can be deposited with a variety of properties ranging from diamond-like to
graphite-like. The film resistivity, being a structure-sensitive property, is also
dependent on deposition conditions. However, in spite of the fact that a num-
ber of works deal with the effects of deposition [1,2,3,4,5] and heat treatment
temperature [10, 11, 12, 13] and ion energy [5, 6, 7, 8, 9] on the microstructure
and properties of a-C films prepared by various techniques, there do not
seem to be any systematic studies on the electrical resistance of a-C films.
176 Alexei A. Onoprienko
Fig. 1. Schematic draw of specimen with Ni contacts (1–1 and 2–2 ) and carbon
film for resistance measurements
In the aforementioned studies, the electrical resistance was measured only as

a subsidiary parameter along with other properties of a-C films. Moreover,
in all of them the resistance (or conductivity) was measured only in one di-
rection: either parallel or perpendicular to the substrate surface. Conversely,
in our previous study [14] we reported, for the first time, the phenomenon
of prominent resistivity anisotropy in a-C films deposited by DC magnetron
sputtering, the extent of which depends on deposition temperature.
The present study focuses on the relationship between structure and
electrical resistivity evolution with substrate temperature and ion bombard-
ment (IB) intensity of carbon films, which may favour understanding the
mechanism of film microstructure formation. Electron diffraction (ED) and
Raman scattering were chosen as the methods sensitive to the structure.
These studies were supplemented by measurements of film resistivity and by
transmission electron microscopy (TEM).
2 Experimental
The a-C films were sputter-deposited using a planar DC magnetron system
of balanced type with graphite target, and argon at pressure of 1 Pa as a
sputtering gas. As substrates, cleavages of NaCl single crystals (for electron
diffraction and electron microscopic studies) and polished platelets of Si–Ti–
Al–O ceramics (for Raman spectroscopy and resistance measurements) were
used. For resistance measurements, specially configured 150 nm thick nickel
film contacts (1–1 and 2–2 , Fig. 1) were magnetron sputter-deposited onto
substrates before (1–1 ) and after (2–2 ) C-film deposition.
Three series of specimens were prepared for study. In one series, films
were deposited onto substrates preheated to a given temperature in the range
20 − 650◦C. In the second series, the carbon films were deposited onto sub-
strates approximately at room temperature followed by annealing for 1 h in
a vacuum of 1.3 × 10−3 Pa at different temperatures in the range 50–650◦C.
During deposition in the above two series, the substrates were at a floating
potential of approximately V bias −20 V. In the third series, carbon films
Resistivity and Structure of Magnetron-Sputtered Carbon Films 177
were deposited onto substrates at room temperature, but DC negative bias

voltage in the range V bias = −25 V to −175 V was applied to lower metal
contacts (1–1 in Fig. 1) during film deposition. It has to be noted that carbon
films in all series were deposited at low magnetron power in order to prevent
uncontrollable heating of the substrates by plasma radiation [15].
The resistance of C-films was measured in two directions: parallel (R )
and perpendicular (R⊥ ) to substrate surface (see Fig. 1). Film resistivities
ρ and ρ⊥ were calculated from the measurements of R and R⊥ , the surface
area and thickness of carbon film between respective metal contacts. The
resistivity anisotropy coefficient was defined as κ = ρ⊥ /ρ .
The film thickness was measured with an optical interferometer, 400–
600 nm, on ceramic substrates. Thickness of the films deposited onto NaCl
for TEM and ED was 60–80 nm. Raman scattering was induced by the λ =
514.5 nm line of argon laser. The recorded spectra were fitted to the sum of
two Gaussian-shape lines (G band centered at approx. 1580 cm−1 and the
D band centered at approx. 1360 cm−1 ) and a linear background term.
3 Results and Discussion

3.1 Substrate Temperature Effects
Figure 2a shows the dependence of logarithm ρ⊥ on the substrate temperature

during deposition (T dep) and annealing (T ann ) temperature (T ann ). As seen,
ρ⊥ decreases with increase in both T dep and T ann , but this change is not
large.
At the same time, those dependencies for ρ exhibit an essentially different
character (Fig. 2b). An increase in T dep results in immediate decrease in ρ at
T dep ≥ 50◦ C. It decreases with temperature change up to T dep ≈ 400◦C, but
with further increase in T dep up to 650 ◦C it remains unchanged. Contrary to
this, a-C films deposited at room temperature and further annealed exhibit
constancy of ρ in the range T ann = 50 to 300◦ C, and only at T ann > 300◦ C
does ρ decrease with the increase in annealing temperature.
In Fig. 2c the dependence of the logarithm of resistivity anisotropy coef-
ficient on T dep and T ann is shown. For films deposited at different values of
T dep, κ increases in the range of deposition temperatures 20–400◦C and then
slightly decreases at higher deposition temperatures. For films deposited at
room temperature and annealed at different temperatures, κ is nearly con-
stant in the range 50–300◦C and then it increases with further increase in
T ann .
In Fig. 3a the dependence of integral intensity ratio I D /I G of D and
G peaks on the T dep and T ann is shown. For films deposited at different
substrate temperatures, I D /I G increases in the range 20–400◦C and then de-
creases with further increase in T dep. For films deposited at room temperature
and then annealed, the intensity ratio I D /I G increases rapidly with increase
Fig. 2. Carbon film resistivity ρ⊥

(a), ρ (b), and resistivity aniso-
tropy coefficient κ = ρ⊥ /ρ (c) as
a function of (1) deposition and
(2) annealing temperatures
in annealing temperature in the range 50–300◦C, but then it increases at a

lower rate.
Figure 3b shows the dependence of the G-peak position on temperatures
T dep and T ann. With increase in temperature of deposition or annealing, the
G-peak position shifts towards higher frequencies in both cases.
Electron diffraction revealed that carbon films deposited at room tem-
perature and then annealed in the range 50–650◦C were amorphous, and
ED patterns consisted of three–four diffused halos corresponding to d ∼ =
0.30, 0.20, 0.16, 0.11 nm that are close to the interplanar distances of a
graphite (0.338, 0.212, 0.169, 0.115 nm, respectively). The carbon films de-
posited at substrate temperatures in the range 20–400◦C were also amor-
phous. Only at T dep > 500◦ C were halos transformed into broadened diffrac-
tion lines peculiar to the structure of polycrystalline graphite.
The observed anisotropy and the character of changing electrical resis-
tance in carbon films studied can be explained in terms of C-film microstruc-
ture. In accordance with the model proposed earlier [14], upon magnetron
sputtering the C-films form with predominantly sp2 -bonds that are distrib-
Fig. 3. Integral intensity ratio

I D /I G (a) and G-peak position
(b) as a function of (1) deposition
and (2) annealing temperatures
uted chaotically within the film volume. ED shows that films deposited at
low (room) temperature are amorphous. However, the presence of a D peak
in the Raman spectra indicates that a-C films contain aromatic rings [16].
At the same time, the coefficient of resistivity anisotropy κ ∼ = 1 for such
films (Fig. 2c) is peculiar to isotropic structure. On the basis of this, it can
be assumed that aromatic rings that form within the structure of C-films
at low temperature are strongly distorted and mutually disordered. Such a
model of a-C film microstructure is justifiable since there is information that
C-films deposited by DC magnetron sputtering and laser ablation of graphite
at low temperature also contain distorted fivefold and sixfold rings [17, 18].
As deposition temperature increases in the range 20–400◦C, the formation of
regular (nondistorted) aromatic rings becomes more probable and ordering
of them into sp2 -bonded clusters takes place. This is supported by growth of
I D /I G (Fig. 3a) in this temperature range because, as pointed out in [16],
the intensity of the D peak increases with the number of aromatic rings with
sixfold symmetry and degree of ordering them. The number of sp2 -bonded
clusters increases with an increase in T dep, and the film forms with a more
graphite-like structure, which is indicated by increase in κ (Fig. 2c). Shift-
ing of the G-peak position towards higher frequencies with increase in T dep
(Fig. 3b) also indicates graphitization of the a-C film. The graphitization of
C-films deposited by various techniques at elevated substrate temperatures

was also observed in [2, 4, 18, 19].
Since electron diffraction patterns from C-films deposited at different sub-
strate temperatures in the range 20–400◦C are similar (halos), it is clear that:
(i) the short-range atomic order in a-C films with a structure based on sp2
bonding remains unchanged, and (ii) the sizes of graphite-like clusters are
so small that structural changes within the clusters are below the sensitiv-
ity limit of the method used (ED). The graphite crystallites of very small
size (nanocrystallites) have been also detected in electron microscopic and
scanning tunnelling microscopic examination of a-C films deposited by DC
magnetron sputtering and laser-arc evaporation at elevated substrate tem-
peratures [17, 20].
Further increases in T dep over 400◦C results in direct nucleation and
growth of a fine-grained graphite phase onto the substrate. This is indicated
by ED patterns, which contain broadened lines peculiar to a polycrystalline
graphite phase, and also the temperature dependence of ratio I D /I G and G-
peak position at elevated temperatures (Figs. 3a, 3b). Nucleation and growth
of graphitic films at high temperatures is supported also by near constancy
of the resistivity anisotropy coefficient at T dep ≥ 450◦ C (Fig. 2c).
The nature of changes in the microstructure of a-C films annealed at var-
ious temperatures differs from that of films deposited at different substrate
temperatures. As seen from Fig. 2a, the dependence of ρ⊥ on temperature
and the respective values of resistivity are virtually the same for both de-
position and annealing. This means that structural changes are similar in
films of both types in the direction perpendicular to substrate surface. At
the same time, annealing in the range 50–300◦C does not appreciably change
ρ , but its absolute value remains higher than for films deposited at the same
temperatures. Only at T ann > 300◦ C does ρ decrease (Fig. 2b). Accord-
ingly, the dependence of κ on T ann is different: up to 300◦ C coefficient κ is
nearly constant and then changes similarly to the case of film deposition at
various temperatures. The ratio I D /I G increases with increase in T ann up to
∼ 300◦ C and then changes little (Fig. 3a). This indicates that postdeposition
annealing also results in changing a-C film microstructure towards graphiti-
zation. However, in contrast to the films deposited at various temperatures,
in this case only transformation of distorted aromatic rings into regular ones
occurs. This process is the most intense in the annealing temperature range
50–300◦C. An increase in the number of regular aromatic rings in the film
structure, in accordance with ideas stated in [16], should give an increase in
D-peak intensity and a shift of G-peak position towards higher frequencies,
as actually observed in Figs. 3a, 3b. However, lengthy graphite fragments
do not form in this case, and it is only at T ann > 300◦ C that the forma-
tion of graphite-like fragments seems to occur, reflected in the corresponding
variation of κ with temperature (cf. Fig. 2c). Since the energy required for
structure rebuilding with cluster formation is higher than for their formation
during deposition, then the microstructure of a-C films changes little with
further increase in T ann, corresponding to the dependence for κ and I D /I G
in that temperature range. This conclusion is also supported by the results
of the electron diffraction study: even after annealing at 650◦ C the respective
ED pattern contain very diffused halos.
3.2 Substrate Bias Effects
In this set of experiments NaCl single-crystal cleavages with Ni contacts

(1–1 in Fig. 1) were used as substrates, and DC bias voltage in the range
V bias = −25 V to −175 V was applied to contacts during film deposition.
Because NaCl is essentially a low-conductivity material, such a configuration
resulted in bombardment with ions of growing film on contacts and in a close
vicinity of contacts, while areas far from contacts were not subject to ion
bombardment (IB) [21].
In Figs. 4a, 4b the TEM micrographs of carbon film deposited at V bias =
−175 V are shown for the areas in which the film was subject and was not
subject to IB, respectively. As can be seen, the microstructures of these parts
of the C-film are quite different. The C-film area grown without the IB con-
sists of round-shaped partially faceted “grains” (Fig. 4b), whereas the film
area grown under intense IB consists of elongated “grains” with clearly seen
internal structure (Fig. 4a). ED patterns from both areas are similar and
coincide with those for amorphous carbon films. However, in some cases ED
patterns taken from film areas affected by intense IB contained an additional
halo with d ∼= 0.30 nm that is close to the d-spacing of graphite (002) basal
plane. Note that such a prominent difference in film structure was observed
only at V bias > −100 V.
Raman spectra were taken from the area in the middle between the con-
tacts for the films grown at different values of substrate bias [21]. Analysis
of Raman spectra recorded for the films deposited at different V bias revealed
a nonmonotonous dependence of I D /I G on bias voltage (Fig. 5). The I D /I G
ratio increases with increased bias voltage and reaches a maximum value at
V bias ≈ −150 V, and then drops to the initial level at −175 V. It was not
possible to investigate the dependence of resistivity anisotropy of C-films on
V bias due to particularities of the film microstructure and variation of the
thickness of such films caused by the IB [21].
Only the ρ⊥ resistivity of these films could be measured. The ρ⊥ values
of C-films deposited at different V bias are listed in Table 1. From these data
one can see that the IB results in an increase of ρ⊥ by more than an order
of magnitude.
Based on the interpretation of Raman spectra developed in [16, 22, 23], it
is possible to assert that the increase in I D /I G with increasing bias voltage
in the range of −25 < V bias < −150 V indicates the bond-angle disorder
of internal structure of graphite-like clusters along with a decrease in their
size. This is indicated also by an increase in film resistivity ρ⊥ in this V bias
Fig. 4. TEM micrographs of carbon films subject to (a) and not subject to (b) in-
tense IB during deposition
Fig. 5. Integral intensity ra-

tio I D /I G as a function of
substrate bias voltage V bias
Table 1. Resistivity values for C-films deposited in different bias conditions
Substrate bias V bias Resistivity ρ⊥

V Ω · cm
Floating 1.5 × 104
−100 1.2 × 105
−150 2.9 × 105
−175 2.2 × 105
range. An increase of V bias over −150 V results in a decrease in I D /I G and

simultaneously in a decrease in film resistivity ρ⊥ (drop-down branch on the
plot in Fig. 5, and data in Table 1). The critical value of V bias ∼
= −150 V for
maximum resistivity in our work is close to that found in [6, 24] for maximum
values of various film properties (optical gap, atomic density, etc.).
The effect of a decrease in both I D /I G and ρ⊥ with increasing IB inten-
sity (above V bias = −150 V) relates to a certain bond-angle ordering within
the graphite-like clusters (i.e., their graphitization) and to some increase of
their size. A similar dependence of I D /I G on V bias was observed during a-C
film deposition by ion-beam sputtering of a graphite target with simultaneous
bombardment by argon ions of the growth surface [25]. It was revealed that
at low ion energy both film density and sp3 concentration are low. The max-
imal value of the film density and the minimal size of film-forming clusters
were achieved at about 100 eV. A further increase in the ion energy promoted
the decrease of film density, and the appearance and increase in the size of
graphite regions together with local microstructural ordering within them.
The most probable reason for these effects could be the development of lo-
cal high-temperature spikes caused by the collision cascades arising during
enhanced IB [26].
TEM investigations of our films allow us to assume that the dense, round-
shaped or partially faceted “grains,” observable in Fig. 4b, are the cross-
sections of the columns disposed normally to the substrate, which form the
film in the area where the IB is almost absent. The “grains” observable in
micrographs taken from a-C film area subject to intense IB also have colum-
nar structure (Fig. 4a), but in contrast to Fig. 4b, these columns are strongly
tilted to the substrate. As it follows from the micrographs, the column width
in Fig. 4a is approximately equal to the diameter of the columns in Fig. 4b.
This means that the columns from both areas are similar. At the same time,
on some of ED patterns from an area of film subject to intense IB the halo
appeared with d -spacing of about 0.30 nm, which is close to (002) d -spacing
of graphite. This halo was never observed in ED patterns from film that was
not subject to IB. The (002) reflection in the transmission ED pattern indi-
cates that the graphite-like planes are arranged nearly parallel to the electron
beam, i.e., these planes are perpendicular to the substrate. In other words,
the “c-axis” of graphite-like clusters forming the columns that are arranged
along the substrate (columns in Fig. 4a) is parallel to the substrate surface,
whereas the graphite-like clusters forming the columns arranged perpendic-
ular to the substrate (Fig. 4b) have “c-axis” directed perpendicular to the
substrate surface.
4 Conclusion
On the basis of the results obtained, it is possible to deduce what follows.
Magnetron-sputter deposition at low temperature results in the formation of
amorphous carbon films with predominantly sp2 bonds between atoms. The
microstructure of such films consists of chaotically displaced and distorted
aromatic rings and graphite fragments. This structure is close to an isotropic
one, so there is virtually no resistivity anisotropy. As the deposition temper-
ature is raised in the range 20–450◦C, more regular sixfold aromatic rings
and graphite-like clusters form on the substrate. The extent of aromatic ring
ordering and cluster growth is greatest at T dep ∼ 450◦C. Graphitization of
the film occurs mainly in the direction parallel to the substrate surface that
is indicated by electron diffraction patterns and the dependence of resistivity
on the deposition temperature. An increase in deposition temperature above
450◦ C results in changing the formation mechanism for carbon films, and the
nuclei of the graphite phase form directly on the substrate. Increasing the
temperature in the range 450–650◦C also results in growth of clusters with
an ordered inner structure. This is evidenced by the drop in the dependence
of I D /I G on deposition temperature, the appearance of ED patterns peculiar
to fine-grained polycrystalline graphite, and the stabilization of the resistivity
anisotropy coefficient at maximum level.
Annealing in the range 50–300◦C of a-C films deposited at room temper-
ature results in the intense transformation of distorted aromatic rings into
regular ones, and in ordering them in the substrate plane without formation
of graphite fragments. Further increase in the annealing temperature up to
650◦ C has little influence on the structural changes in film, and graphite-like
clusters do not form in this case. This inference is supported by the depen-
dencies for ρ and κ, and the absence of “graphite” lines in the ED patterns
over the entire annealing temperature range studied.
Ion bombardment of growing C-film affects its structure and therefore
resistivity. In the range of substrate bias voltage −20 < V bias < −150 V
the size of graphite-like clusters within C-films decreases and simultaneously
increases their internal structural disorder. At V bias > −150 V the clus-
ter structure tends to be more ordered towards its graphitization. The mi-
crostructure of C-films is of columnar type, and columns are built of packs of
graphite-like clusters that are ordered somehow within the columns. When
the intensity of IB is low (or absent) the c-axis of these clusters is oriented
predominantly perpendicular to the growth surface, while the enhancement
of the IB promotes the reorientation of the c-axis predominantly along the
growth surface. The observed variations in C-film resistivity occur according

to film microstructure changes.
References
[1] N. H. Cho, D. K. Veirs, J. W. Ager III, et al.: J. Appl. Phys. 71, 2243 (1992)
175
[2] E. Mounier, F. Bertin, M. Adamik, et al.: Diamond Relat. Mater. 5, 1509
(1996) 175, 180
[3] J. Koskinen, J. P. Hirvonen, J. Keranen: J. Appl. Phys. 84, 648 (1998) 175
[4] B. K. Tay, X. Shi, E. J. Liu, et al.: Thin Solid Films 346, 155 (1999) 175, 180
[5] M. Chhowalla, J. Robertson, C. W. Chen, et al.: J. Appl. Phys. 81, 139 (1997)
175
[6] P. J. Fallon, V. S. Veerasamy, C. A. Davis, et al.: Phys. Rev. B 48, 4777 (1993)
175, 183
[7] Y. Lifshitz, G. D. Lempert, S. Rotter, et al.: Diamond Relat. Mater. 3, 542
(1994) 175
[8] M. Gioti, S. Logothetidis: Diamond Relat. Mater. 7, 444 (1998) 175
[9] F. Rossi: Chaos, Solitons & Fractals 10, 2019 (1999) 175
[10] L. G. Jacobson, R. Priorli, F. L. Freire Jr., et al.: Diamond Relat. Mater. 9,
680 (2000) 175
[11] T. A. Friedman, K. F. McCarty, J. C. Barbour, et al.: Appl. Phys. Lett. 68,
1643 (1996) 175
[12] S. Anders, J. W. Ager III, G. M. Pharr, et al.: Thin Solid Films 308-309, 186
(1997) 175
[13] A. C. Ferrari, B. Kleinsorge, N. A. Morrison, et al.: J. Appl. Phys. 85, 7191
(1999) 175
[14] A. A. Onoprienko, L. R. Shaginyan: Diamond Relat. Mater. 3, 1132 (1994)
176, 178
[15] A. A. Onoprienko, V. V. Artamonov, I. B. Yanchuk: Surf. Coat. Tech. 200,
4174 (2006) 177
[16] A. C. Ferrari, J. Robertson: Phys. Rev. B 61, 14095 (2000) 179, 180, 181
[17] B. Marchon, M. Salmeron, W. Siekhaus: Phys. Rev. B 39, 12907 (1989) 179,
180
[18] S. Bhargava, H. D. Bist, A. V. Narlikar, et al.: J. Appl. Phys. 79, 1917 (1996)
179, 180
[19] M. A. Capano, N. T. McDevitt, R. K. Singh, F. Qian: J. Vac. Sci. Technol. A
14, 431 (1996) 180
[20] I. Alexandrou, H. J. Scheibe, C. J. Kiely, et al.: Phys. Rev. B 60, 10903 (1999)
180
[21] L. R. Shaginyan, A. A. Onoprienko, V. F. Britun, V. P. Smirnov: Thin Solid
Films 397, 288 (2001) 181
[22] S. Fujimaki, M. Kitoh, K. Furusawa, et al.: Diamond Films and Technol. 2, 17
(1992) 181
[23] D. Beeman, J. Silverman, R. Lynds, M. R. Anderson: Phys. Rev. B 30, 870
(1984) 181
[24] J. Ishikawa, Y. Takeiri, K. Ogawa, T. Takagi: J. Appl. Phys. 61, 2509 (1987)
183
[25] F. Rossi, B. Andre, A. van Veen, et al.: J. Appl. Phys. 75, 3121 (1994) 183
[26] L. K. Cheah, X. Shi, B. K. Tay, E. Liu: Surf. Coat. Tech. 105, 91 (1998) 183
Index
a-C, 175, 176, 181 graphitization, 175, 179, 180, 184
sp2 -bonded clusters, 179–181, 183
ion bombardment (IB), 176, 181–184
annealing, 175–180, 184
magnetron sputtering, 176
clustering microstructure, 175, 176, 178–181, 183,
graphite-like clusters, 175, 180, 181, 184
183 graphite-like structure, 179, 183
D band, 177, 179, 180
D/G band intensity ratio, 177, 179–184 Raman spectroscopy, 176, 179–181
diamond-like carbon (DLC), 175 resistivity, 175–178, 180, 181, 183
resistivity anisotropy, 175–181, 184
electron diffraction (ED), 176–181, 183,
184 substrate bias, 181
G band, 177–180 transmission electron microscopy

graphite-like carbon (GLC), 175 (TEM), 176, 177, 181–183
Formation, Atomic Structures and Properties
of Carbon Nanocage Materials
Takeo Oku1 , Ichihito Narita1 , Atsushi Nishiwaki1 , Naruhiro Koi1 ,

Katsuaki Suganuma1 , Rikizo Hatakeyama2,
Takamichi Hirata2 , Hisato Tokoro3, and Shigeo Fujii3
1
Institute of Scientific and Industrial Research, Osaka University, Mihogaoka 8-1,
Ibaraki, Osaka 567-0047, Japan
2
Graduate School of Engineering, Tohoku University, Sendai 980-8579, Japan
3
Hitachi Metals Co. Ltd., Advanced Electronics Research Laboratory, Mikajiri
5200, Kumagaya, Saitama 360-0843, Japan
oku@sanken.osaka-u.ac.jp
Abstract. Various carbon nanocage fullerene materials (clusters, metallofullerenes,

onions, nanotubes, nanohorns and nanocapsules) were synthesized by electron-beam
irradiation, chemical reaction, hybrid arc discharge, and self-organization. Atomic
structures and structural stability of these materials were investigated by high-
resolution electron microscopy, molecular dynamics, and molecular orbital calcu-
lations. Photoluminescence as well as magnetic and electronic properties of these
fullerene materials were also investigated. The present work indicates that the new
carbon nanocage fullerene materials with various atomic structures and properties
can be produced by various synthesis methods, and a guideline for designing the
carbon fullerene materials is summarized.
1 Introduction
Carbon has various hollow-cage nanostructures such as C60 , giant fullerenes,
nanocapsules, onions, nanopolyhedra, cones, cubes, and nanotubes [1, 2, 3,
4, 5, 6, 7]. These C structures show different physical properties, and have
the potential of studying materials of low dimensionality within an isolated
environment. In particular, cluster-included fullerene materials are intrigu-
ing for both scientific research and device applications such as cluster pro-
tection, cluster separation, nano-ball bearings, nano-electronic–magnetic de-
vices, gas storage, and biotechnology [8, 9, 10, 11, 12, 13, 14]. By controlling
of the size, layer numbers, helicity, compositions, and included clusters, the
cluster-included C nanocage structures with band-gap energy of 0–1.7 eV and
nonmagnetism are expected to show various electronic, optical, and magnetic
properties such as Coulomb blockade, photoluminescence, and superparamag-
netism, as shown in Fig. 1a. Possible applications of the C fullerene materials
are also shown in Fig. 1b.
The direction of the research on carbon nanocage materials is threefold.
The first is to prepare the new C nanocage fullerene materials. In the present
188 Takeo Oku et al.
Fig. 1. (a) Structures and properties of C fullerene materials. (b) Applications and
future of C fullerene materials
Carbon Nanocage Materials 189
work, various methods such as electron-beam irradiation, chemical reaction,

hybrid arc discharge, and self-organization techniques were selected for the
formation of the fullerene materials. Nanocapsules and onions are expected to
form at low temperatures compared to the ordinary arc-discharge method.
Au clusters were selected for the C nanocapsule formation. Arc-discharge,
electron-beam irradiation and chemical synthesis techniques were applied for
the fullerene formation. One-dimensional carbon nanocage arrangement was
also tried by using self-organization of metal nanoparticles with a size below
10 nm. Gold colloids have been used for the formation of single-electron tran-
sistors [15, 16]. In the present work, Au nanoparticles were selected because
of easy control of cluster size [17]. Fabrication of one-dimensional Au nan-
odots and nanowires encapsulated in carbon nanocage structures was tried
by annealing the one-dimensional self-organized nanoparticles on carbon thin
films.
The second direction is to understand the formation mechanism and mi-
crostructure of these C nanocage fullerene materials from high-resolution elec-
tron microscopy (HREM), which is a powerful method for structure analysis
at the atomic scale [18, 19, 20]. Energy dispersive spectroscopy (EDS) was
carried out for composition analysis the fullerene materials. In order to con-
firm the atomic structure proposed by HREM, theoretical calculations by
molecular mechanics, molecular dynamics, and molecular orbital calculations
were carried out. The third direction is to understand the various properties
of the C nanocage materials. Photoluminescence (PL), electronic properties
and magnetic properties of these fullerene materials were investigated. Down-
sizing of these materials is expected to show the various properties [21]. These
studies will give us a guideline for design and synthesis of the fullerene ma-
terials, which are expected for the future nanoscale devices.
2 Synthesis Methods
Carbon nanocapsules and clusters were produced by the following method [22,
23]. β-SiC nanoparticles (Sumitomo Cement), were dispersed in deionized
water with polyvinyl alcohol (PVA-706, Kuraray) at 60◦ C. This polymer is
a copolymer of polyvinyl alcohol and polyvinyl acetate. The solution with
clusters and polyvinyl alcohol was dried in a drying oven prior to loading
into the vacuum chamber. The annealing chamber was first evacuated to
1 × 10−6 Pa, and the samples were annealed at 400–800◦C for 30 min in an
Ar atmosphere of 0.12 MPa.
Commercially available a-Fe2 O3 (hematite, 99.7%) and graphite carbon
(99.5%) powders were also used as starting materials [24, 25]. Their particle
sizes were 30 nm and 5 µm, respectively, which were measured by using a
particle counter (HORIBA LA-920). Samples for Fe nanoparticles coated with
C nanolayers were prepared by annealing mixtures of a-Fe2 O3 and C powders
at 1000◦C for 2 h in the same manner. Ratios in weight of a-Fe2 O3 to B
were varied from 7 : 3 to 3 : 7. Those of a-Fe2 O3 to C were from 8 : 2

to 5 : 5. Each mixture was mixed in a V-shape mixer. The temperature
elevation rate was 3◦ C min−1 at annealing. For the formation of carbon
onions, amorphous carbon produced from with polyvinyl alcohol (PVA-706,
Kuraray) was used [26, 27].
Carbon nanocapsules were also prepared by a direct current and radio-
frequency (DC-RF) hybrid arc discharge using pair anodes in order to en-
hance the Ge and Si evaporation [28, 29]. A graphite hollow cylinder (0.6 cm
in diameter, 30 cm in length) is used for the main DC arc discharge. The
gap distance between the main discharge electrodes is approximately 0.4 cm.
A subanode of a carbon cylinder (1.5 cm in diameter, 10 cm in length) with
the semiconductor (Si or Ge) powder is used, and a main anode of a carbon
rod is partially mixed with the semiconductor powder. The DC arc discharge
is initiated between a cathode electrode and the pair anodes. An RF antenna
of water-cooled spiral copper pipe is installed 3 cm above the arc point to
generate auxiliary plasma produced by an RF discharge around the DC arc
plasma. When RF power is applied to an antenna through a matching circuit,
the antenna DC current is also generated. The experimental parameters are
as follows: DC arc current, 100 A; subanode current, 80 A; RF power, 600 W
(13.56 MHz); helium gas pressure, 100 Torr. A detailed illustration of the
apparatus was provided in the previous work. The plasma produced by this
DC-RF hybrid discharge is much more voluminous than that by the ordinary
DC arc discharge.
Au nanoparticles (ULVAC) with a size of ca. 5 nm are selected in the pres-
ent work for the one-dimensional positioning of carbon nanocapsules [30, 31].
The surface of these nanoparticles was stabilized by α-terpineol (C10 H18 O)
in toluene solution. The solution with Au nanoparticles was dispersed on
holey carbon grids with thickness ca. 15 nm (Oken Syoji.). After drying
the specimens, they were loaded into the vacuum chamber, and annealed at
∼ 200–400◦C for 30 min in vacuum of ca. 7 × 10−4 Pa.
3 Fullerene Clusters and Metallofullerenes

Hollow carbon clusters with diameters in the range of 0.7–1.0 nm are often
on the carbon nanocapsules. Two of the carbon clusters are shown in Figs. 2a
and 2b. The size of the carbon cluster in Fig. 2a is in the range of 0.5–0.7 nm,
which corresponds to that of C36 [32,33]. The high-resolution image of Fig. 2b
also shows a carbon cluster with the size of 0.7–0.8 nm, and the number of
carbon atoms would be ∼ 70 (C70 ) [1], which could be estimated from the size
of the cluster. In addition, dark contrast is observed inside the cluster, which
would indicate the existence of several atoms inside the carbon cluster. The
carbon cluster is directly connected at the step edge of the graphite sheets,
which results in the stabilization of the carbon cluster. These C36 and C70
carbon clusters have elliptical structures [1,32,33], which agrees well with the
Fig. 2. HREM images of (a) C36 and (b) C70 clusters. Atomic structural models
of (c) C36 and (d) C70 clusters
present high-resolution observation. Structure was optimized by molecular

orbital method using MOPAC software (Fujitsu).
The possibility of direct detection of doping atoms in C60 solid clusters
by HREM was investigated using calculated residual indices (RHREM ) and
difference images [1, 34, 35]. The light element nitrogen (Z = 7) and heavy el-
ement rubidium (Z = 37) were selected as doping atoms in the present work.
It was reported that nitrogen atoms were included at the center of the C60
clusters [36], which was detected by electron paramagnetic resonance [37].
It was also reported that Rb atoms were doped at the octahedral sites of
C60 solid clusters, which were determined by X-ray diffraction [38]. Based on
the crystal structure of C60 solid clusters, difference images between C60 and
doped C60 (C60 /doped C60 ) were calculated for image analysis using residual
indices. HREM images of C60 doped with nitrogen (N) and rubidium (Rb)
atoms were calculated as functions of accelerating voltage of microscope,
crystal thickness, and defocus values. The structure models and calculated
images are shown in Fig. 3. Difference images of N@C60 /C60 and RbC60 /C60
showed N and Rb atomic positions clearly at high accelerating voltage and
large crystal thickness. RHREM values of RbC60 are strongly dependent on
accelerating voltage and crystal thickness, which would be due to the dynam-
ical diffraction effect of Rb atoms. Defocus values also influence the RHREM
Fig. 3. Structural models of C60 , N@C60 , and Rb@C60 , and HREM simulated
images
values, and the Scherzer defocus would be the best for atomic detection. The
present work indicates that the doping atoms at the inside and interstitial
site of C60 solid clusters could be detected by HREM with difference images
and RHREM values.
Various types of endohedral C clusters have been discovered and investi-
gated. Metallofullerenes, which have atoms (such as La, Y, Sc, Li, Na, N, and
other lanthanoids) inside the C60 , C82 , C84 , and other giant fullerenes, are ex-
pected as superatoms with various structures and physical properties [39, 40].
In addition, encapuslation of Sc [41], La [42], Ba [43], Ca [40, 44], Eu [45],
Gd [46], Y [47], Sm [47] and Tb [47] atoms inside the C74 cluster have been
reported.
Silicon (Si) atoms have also been successfully introduced in carbon clus-
ters by using the DC-RF hybrid arc discharge. Mass analysis of the soot
indicated high-yield production of C74 [48] and existence of Si@C74 [28], as
shown in Fig. 3a. The C74 fullerene has D3h symmetry satisfying the isolated
pentagon rule, and the possibility of inserting a Si atom inside the C74 cage
was investigated by using ab initio molecular dynamics simulation [49].
Hollow carbon clusters with diameters of ∼ 0.8 nm are often observed on
carbon nanotubes and nanocapsules. A HREM image of a carbon cluster on
a carbon nanotube is shown in Figs. 3b and 3c is an enlarged image of the
carbon cluster. The carbon cluster has an elliptical structure, and the size is
∼ 0.8 nm, which agrees well with the size of Si@C74 [50]. Dark contrast is
often observed in the carbon clusters. This indicates the existence of Si atoms
inside the carbon clusters, which was also supported by the mass spectra
analysis.
Fig. 4. (a) Mass spectra of the soot indicated high-yield production of C74 .
(b) HREM image of Si@C74 on a carbon nanotube. (c) Enlarged HREM image
of the Si@C74 cluster. Calculated HREM images of (d) C74 and (e) Si@C74 clus-
ters, as a function of defocus values, respectively
Table 1. Calculated values for carbon clusters

Total steric energy Heat of formation εHOMO εLUMO Energy gap
kcal kcal kcal kcal
mol mol · atom mol mol · atom
eV eV eV
C60 179 2.98 811 13.5 −1.527 1 0.606 4 2.133 5
C70 177 2.53 886 12.7 −1.503 9 0.454 1 1.958 0
C72 197 2.74 928 12.9 −1.399 6 0.572 9 1.972 5
C74 170 2.30 950 12.8 −0.468 3 −0.369 8 0.098 5
Si@C74 168 2.24 994 13.3 −0.564 9 −0.466 6 0.098 3
C76 182 2.39 937 12.3 −1.106 3 0.272 8 1.379 1
The cluster structures are calculated theoretically, and the structure mod-
els of the C74 and Si@C74 are shown in Figs. 4d and 4e, respectively. The
structures are optimized by molecular orbital calculations using PM3. Based
on the projected structure models, image calculations on the C74 and Si@C74
clusters were carried out for various defocus values to investigate the imaging
condition of the HREM images, as shown in Figs. 4d and 4e, respectively. The
Scherzer defocus value of the present 300 kV electron microscope is ∼ 41 nm.
The experimental data agree well with HREM images calculated at the de-
focus values in the range of −40 to −50 nm, which indicates the observed
HREM images were taken close to the Scherzer defocus.
Electronic structures of C74 , Si@C74 , and a single Si atom were investi-
gated by DV-Xa, as shown in Figs. 5a–c, respectively. Calculated results for
the structural stability and electronic structure are listed in Table 1. The
C60 shows a large energy gap between the highest occupied molecular orbital
(HOMO) and the lowest unoccupied molecular orbital (LUMO). Although
the HOMO–LUMO gap of C60 is calculated to be 2.134 eV, the energy gaps
of C74 and Si@C74 are 0.099 and 0.098, respectively. The energy gap of a
single Si atom was also calculated to be 0.452 eV. The C2p orbital is strong
around the energy gap for C60 and C74 , and the Si3p orbital shows the density
of states (DOS) around the energy gap for Si@C74 . The Si@C74 cluster in
Fig. 4b is directly connected on the carbon {002} sheets, which would result
in the stabilization of the carbon cluster. Total energies of the C clusters were
calculated as listed in Table 1, and 60–76 total steric energy of the Si@C74
was the lowest in the present calculation, which indicates that the insertion
of the Si atom into the C74 cluster stabilizes the structure. Heat of formation
per carbon atom for the Si@C74 is almost the same as that of C60 cluster,
which indicates the structure is fairly stable.
It has been reported that the energy gap of C74 clusters is small (∼ 0.2 eV)
compared with the ordinary C60−80 fullerenes [19]. In the present work,
the energy gaps of the C74 and Si@C74 were calculated to be 0.0985 and
0.0983 eV, respectively. Although the Si3d orbital affects the energy levels
of the C74 , the Si@C74 shows almost the same HOMO–LUMO gap. For the
metallofullerenes such as Sc@C74 , La@C74 , Ca@C74 , Eu@C74 , and Gd@C74 ,
Fig. 5. Energy level diagram and density of states of (a) C74 , (b) Si@C74 , and
(c) a single Si atom
encapsulated atoms are ionized and change the electronic states. In the pres-
ent work, the energy gap of a single Si atom is calculated to be 0.452 eV,
as shown in Fig. 5c, and electronic state of the Si atom is isolated by the
C74 cluster. The Si atom would keep the original electronic state inside the
C74 cluster, which would be interesting for investigation of atomic properties
of Si.
4 Onions and Nanotubes

Carbon was formed from polyvinyl alcohol after annealing at 400◦ C in Ar,
and the carbon matrix was found to have an amorphous structure from the
HREM image, as shown in Fig. 6a. The amorphous carbon was irradiated by
an electron beam for 20 min under a beam current of 100 A/cm2 at 1250 kV.
This beam current is ∼ 20 times higher compared to that of the ordinary
HREM observation. Graphitization of amorphous carbon is observed, which is
confirmed by the lattice fringes of carbon layers with a spacing of ∼ 0.34 nm,
as shown in Fig. 6b. Two carbon onions were produced from amorphous
carbon as shown in Fig. 6b. Lattice fringes of the carbon {002} planes are
observed.
Figure 6c is an HREM image of tetrahedral carbon onion produced from
that of Fig. 6b [27]. By electron-beam irradiation for 10 min on the onions
of Fig. 6b, a carbon onion with tetrahedral structure was formed as shown
in Fig. 6c. The edge lengths of the first and second internal shell of the
tetrahedral carbon onion are approximately 0.8 and 1.4 nm, respectively.
The distance between onion layers is 0.35 nm. Although a similar structure
consisting of C264 @C660 @C1248 was calculated [15], the size of the present
tetrahedral carbon onion is smaller compared to the C264 .
Figure 6d is a six-layered onion structure, and the outside is calculated to
be C2160 from the equation 60n2 (n is the number of the shell). The lattice
fringes of the carbon {002} planes are smeared at the edges of the onions with
pentagonal carbon bonding, as indicated by arrows. “Atom clouds” [51, 52],
which have vague contrast, are also observed at the vertices of onion edges as
indicated by arrows. The carbon “vibration” was also observed at the surface
of the onions. The shell structure is not perfect, but has some defects. The
same onion after 60 s is shown in Fig. 6e. Change of the vertices into spherical
structure is observed on the onion surface. At the center of the onion, there
would be a C28 cluster from the size of ∼ 0.4 nm. Although this structure
would be unstable, the C28 cluster would be stabilized by compressive tension
of outer onion. The structure model is proposed as shown in Fig. 6f.
At the center of the onion in Fig. 6c, there would be a tetrahedral carbon
cluster and the basic arrangements of carbon atoms are proposed, as shown
in Fig. 7a. Each vertex consists of a hexagonal ring (as indicated by a star
mark), and three pentagonal rings exist around the vertex along the edge.
Other parts consist of only hexagonal rings. Edge lengths of the C84 and
Fig. 6. (a) HREM image of as-prepared amorphous carbon. (b) Two carbon onions
produced from amorphous carbon (a) by electron-beam irradiation for 20 min.
(c) Tetrahedral carbon onion produced from (b) by electron-beam irradiation for
30 min. (d) HREM images of onion structures. (e) Six-layered carbon onion; 60 s
after (d). (f ) Structure model of C24 @C84
Table 2. Calculated values for tetrahedral carbon onions

C84 Td C168 C276 C408 C84 @C276 C84 D2
(No. 20) (No. 22)
Calculation method PM3 PM3 MM3 MM3 MM3 PM3
Layer number 1 2 3 4 1, 3 –
Edge length (nm) 0.8 1.0 1.4 2.0 1.4 –
Interatomic distance of
pentagonal rings (nm) 0.1445 0.1444 0.1399 0.1399 – 0.1438
Interatomic distance of
hexagonal rings (nm) 0.1452 0.1429 0.1407 0.1406 – 0.1464
Total steric energy
(kcal/mol) 362 395 389 345 823 335
Total steric energy
(kcal/mol · atom) 4.31 2.35 1.41 0.85 2.29 3.99
Heat of formation
(kcal/mol) 991 1413 – – – 967
Heat of formation
(kcal/mol · atom) 11.8 8.41 – – – 11.5
C276 in optimized structures are 0.8 and 1.4 nm, respectively. The total num-
bers (N ) of carbon atoms of the present tetrahedral onions are represented
by the equation N = 12(n + 2)2 − 24 (n = 1, 2, ...). Edge lengths of the C84
and C276 agree with the first and second internal shells of tetrahedral carbon
onion in the HREM image, respectively. Steric energy by molecular mechan-
ics calculations and heats of formation by semiempirical molecular orbital
calculations on tetrahedral carbon clusters are summarized in Table 2.
On the basis of the projected structure model of Fig. 7c, image calcu-
lations on the C84 @C276 cluster were carried out for various defocus values
to investigate the cluster structures. The smallest tetrahedral onion in the
HREM image in Fig. 6c agrees well with the calculated images of C84 @C276
in Fig. 7c. In this optimized model, the distance between layers is close to
the observed value of 0.35 nm. Further refinement has to be carried out in
which the van der Waals interaction is considered. Energy levels and density
of states of C84 were calculated, as shown in Fig. 7b. Energy gaps of C60 ,
C84 , and C168 clusters were calculated to be 2.13, 2.06, and 1.44 eV, respec-
tively. The experimental energy gap of C60 is 1.7 eV [53]. Since the difference
between the present calculation and the experimental data is approximately
0.4 eV, the energy gap of a C84 cluster would be approximately 1.6 eV, which
is a little smaller compared to that of the C cluster. The increase of numbers
of the 60 carbon atoms decreased energy gaps of the clusters, as calculated
for C186 (1.44 eV). In addition, DOS for C186 is broader than that of C84 .
The formation mechanism can be described with three effects: cutting of car-
Fig. 7. (a) Proposed structure models of tetrahedral carbon onions: C84 , C168 , C276
and C408 (◦, pentagonal ring; , vertex (hexagonal ring)). (b) Energy level diagram
and density of states for C84 . (c) Structure model and (d) calculated HREM images
of C84 @C276 onion as a function of defocus values
bon bonds by electron irradiation, knock-on of carbon atoms, and increase of

sample temperature. In the present work, bonding cuts and knock-on effects
would be important. These effects cause fluidization of the structure, which
provides a spherical structure by the surface tension. In Fig. 6b, two onions
were observed and their collision would occur during onion formation. The
onion would be formed from the outside because C84 is too unstable to be
fabricated, compared to C60 and related giant fullerenes. The outsidemost
layer would be formed first, and inside layers would be formed gradually. In
the end, the number of remaining carbon atoms is important to construct
the structure at the center of the onion, so that C84 would be formed in the
present work. Concerning the shape and the size of the onion, the center of
the onion would be C84 , and the onion would have any effect from the carbon
layer located outside. There are two important theorems of the Euler rule [54]
and the isolated pentagon rule (IPR) [55] in the formation of the fullerene.
Even if the closed fullerene size becomes larger, the number of pentagonal
rings is always 12 by the Euler rule. Pentagonal rings do not adjoin to others
based on the IPR rule when the size becomes bigger than C60 . There are
24 structural isomers in C84 [56], and C84 D2 and C84 D2d with the ratio of
2 : 1 are the main structural isomers. The first inside shell of the tetrahedral
carbon onion is equivalent to the C84 Td structural isomer. Heats of forma-
tion of C84 D2 and C84 D2d by the PM3 calculation are 967 and 968 kcal/mol,
respectively. This means that a spherical structure is more stable than the
tetrahedral structure (heat of formation 994 kcal/mol), and the Stone–Wales
transformation [57] would be introduced during the growth of C84 Td to sat-
isfy the Euler rule and the IPR rule. The position of the pentagonal ring
can be moved by this transformation keeping the fullerene size. Although
this tetrahedral structure (C84 Td ) might transform into spherical structure
(C84 D2 and C84 D2d ), the energy barrier between them would prevent the
transformation. The barrier would be the van der Waals force from outside
shell and distortion due to the collision of two onions.
The smallest carbon nanotube with a diameter of 0.4 nm was reported
in [58]. They showed a HREM image of the multiwalled carbon nanotube,
and they proposed an atomic structure model of C20 for the cap structure.
However, the nanotube was not an isolated single-walled nanotube.
Figure 8a is an HREM image of isolated carbon mininanotube on the
carbon layers. The diameter and length of the nanotube are 0.5 and 3.1 nm,
respectively. Figure 8b is a proposed atomic structure model of carbon mini-
nanotube C180 . The size of the C180 agrees well with the observed image of
Fig. 8a. The cap of the nanotube consists of a part of the structure of a C36
cluster, which are indicated by dark gray. Figure 8c is a 30◦ -rotated model
of Fig. 8b along the nanotube axis. The atomic structure model of the C180
carbon mininanotube along the nanotube axis is shown in Fig. 8d. Carbon
atoms corresponding to C36 clusters are indicated by dark gray. A structure
model of C36 is also shown in Fig. 8e. The distortion of the mininanotube
would be mitigated by the surface of the carbon layers, which may stabilize
the cap structure.
5 Carbon Nanocapsules
Although many carbon nanocapsules with various elements and compounds
have been prepared by an ordinary arc-discharge method, a few nanocapsules
filled with Ge and SiC have been reported.
A HREM image of Ge nanoparticles encapsulated in graphite sheets is
shown in Fig. 9a. In Fig. 9b, the Ge nanoparticle is surrounded by one to three
graphite sheets. Figure 9c is a HREM image of a Ge nanowire encapsulated
in a carbon nanotube with one to two graphite sheets. The diameter is 10 nm,
Fig. 8. (a) HREM image of carbon mininanotube on the carbon layers. (b) Pro-
posed atomic structure model of carbon mininanotube C180 . (c) 30◦ -rotated model
of (b) along nanotube axis. (d) Structure model of C180 carbon mininanotube along
nanotube axis. (e) Structure model of C36
Fig. 9. (a) HREM image of Ge nanoparticles encapsulated in carbon sheets. (b) En-
larged image of the nanocapsule. (c) Ge nanowire encapsulated in carbon nano-
tube. HREM images of (d) carbon nanocapsules with SiC nanoparticles, (e) car-
bon nanocapsule and nanohorn. (f ) Enlarged image of SiC cluster encapsulated in
carbon cage
and the length is 70 nm. The Ge nanowire has microtwin structures, and the
growth direction of the nanowire is < 111 > of Ge [29].
A HREM image of carbon nanocapsules with SiC nanoparticles is shown
in Fig. 9d. SiC nanoparticles with sizes of 6–10 nm are observed. All SiC
nanoparticles are encapsulated in graphite sheets, and the number of graphite
sheets was in the range of three to ten layers. A HREM image of the carbon
nanocapsule and nanohorn is shown in Fig. 9d. Five to six graphite layers are
observed around the nanoparticle, and the size of the carbon nanocapsule is
Fig. 10. Structure models for carbon nanohorns. Three and five pentagonal C rings
are introduced for model (a) C67 and (b) C74 , respectively. (c) Density of states
for C74 nanohorn
4 nm. Lattice fringes with a distance of 0.25 nm, which corresponds to the
distance of the {111} planes of β-SiC, are observed in the cluster. A car-
bon nanohorn with three graphite sheets is observed. The tip of the cage is
smeared. Figure 9e is a HREM image of a carbon nanocapsule filled with a
SiC nanocluster of size 2 nm.
In this work, a special DC-RF arc-discharge method was used, which re-
sults in their formation. The SiC nanoparticles would be formed by reaction of
Si and C atoms in the arc plasma. The DC-RF arc-discharge plasma would
be effective for the formation of semiconductor nanoparticles encapsulated
in graphite sheets. In the present work, the sizes of the semiconductor Ge
and SiC nanoparticles were reduced down to 10 nm, which indicates that
the widening of the band-gap energy is expected by quantum size effects,
and peculiar optical properties will be expected. Ge nanowires encapsulated
in carbon nanotubes are also expected as one-dimensional devices. Si ox-
ide and Ge oxide layers are not observed at the surface of the SiC and Ge
nanoparticles, which indicates that the carbon nanocapsules are effective for
cluster protection against oxidation in air. Figure 10 show structure models
for carbon nanohorns observed in Fig. 9e. Three and five pentagonal C rings
are introduced for model Figs. 10a and 10b, respectively, and the model in
Fig. 10b is equivalent to the previously proposed model [59].
Heats of formation of the C67 and C74 clusters were calculated to be
22.18 and 20.53 kcal/mol · atom, and the stability of C74 is higher than that
of C67 . DOS of C74 nanohorns were calculated, as shown in Fig. 10c. Energy
level diagram of C74 nanohorn shows narrow energy gaps of 0.49 eV between
the HOMO–LUMO gap. Carbon nanohorns are expected as catalytic elec-
Fig. 11. (a) HREM image of carbon nanocapsules with SiC cluster. (b) Photolumi-
nescence spectrum of (a). (c) HREM image of surface of carbon nanocapsule with
SiC together with calculated image based on the proposed interfacial model (d)
trode materials for fuel cells in next generation, which separate hydrogen
and electrons from methanol [8].
6 Properties of Carbon Nanomaterials

6.1 Photoluminescence of Carbon Nanocapsules
SiC nanoparticles prepared with polyvinyl alcohol were annealed at 500◦ C in

an Ar atmosphere. The particle size of SiC is in the range of 5–50 nm. All the
SiC particles are surrounded by graphite sheets, and the number of graphite
sheets is in the range 1–6. An enlarged HREM image of a carbon nanocap-
sule is shown in Fig. 11a. Lattice fringes with a distance of 0.25 nm, which
corresponds to the distance of {111} planes of β-SiC, are observed in the
nanoparticle. Two {002} planes of graphite grow on the β-SiC nanoparticle.
The PL spectrum of the carbon nanocapsules encasing SiC clusters is

shown in Fig. 11b [12]. A luminescence peak is observed at 382 nm, which
corresponds to ∼ 3.2 eV. The bulk SiC has an energy gap of 3.00 eV. The
present carbon nanocapsules with SiC clusters have higher energy compared
to the bulk SiC, which would be due to the quantum size effect of SiC as
shown in Fig. 11a. A HREM image of the surface of a carbon nanocapsules
with SiC nanoparticle is shown in Fig. 11c. Carbon {002} planes grow epi-
taxially on the SiC {111}, but disordered carbon structure is observed on
the SiC {200}. Epitaxial growth of carbon {002} on the β-SiC {111} planes
was often observed at the nanocapsule/nanoparticle interfaces in the pres-
ent work. A similar epitaxial relationship was observed at the carbon/β-SiC
in previous study [60]. The information on atomic arrangement at the car-
bon/SiC interface was obtained from the HREM images in the present work,
and two structural models of the carbon/SiC interface were constructed, as
shown in Fig. 11d. In Fig. 11d, carbon atoms that belong to the SiC structure
connect with carbon atoms of the graphite. From the present high-resolution
observation, the carbon {002} is parallel to the SiC {111}, and carbon [1,−1,0]
is nearly parallel to SiC [0,1,−1]. (If the incidence is parallel to carbon [0,1,0],
two-dimensional lattice fringes should be observed.) The distance between
the SiC crystal and graphite layer was assumed to be 0.335 nm, which is the
same as that of graphite {002} planes. Based on these models, HREM images
are calculated as shown in Fig. 11c as an inset. In the observed images, the
distance between the dark contrast of the first carbon layer and the top of
the SiC crystal is almost the same as that of carbon {002}, which indicates
that carbon atoms of the graphite-type structure directly connect with car-
bon atoms that belong to the SiC structure. Graphitization of amorphous
carbon on SiC {111} planes would be easier compared to the SiC {200}. It is
believed that the carbon surface of SiC would have a low activation energy
for graphite growth, which results in the heterogeneous nucleation growth of
graphite.
6.2 Magnetic Properties of Carbon Nanocapsules
Magnetic nanoparticles are suitable for potential applications like high-den-

sity magnetic recording media [61], magnetic fluids [62], electromagnetic wave
absorbing materials [63], magnetic carriers in clinical cures, and other novel
magnetic devices. Iron (Fe) or Fe-based alloy nanoparticles have an advan-
tage of high saturation magnetization for these applications, whereas deteri-
oration by oxidation has kept them from practical usage. To overcome this
problem, nanocoating techniques for the metal nanoparticles have been re-
ported [24, 25].
Carbon nanocapsules encapsulated Fe are shown in Fig. 12a. Fe nanopar-
ticles with 100–200 nm diameters are observed as shown in Fig. 12a. An
EDX spectrum of the nanoparticles confirmed that they are Fe nanoparticles.
HREM makes it clear that the Fe nanoparticles are coated with well-oriented
Fig. 12. Morphology of C-coated Fe particles. (a) TEM image of C wire with Fe
particle. (b) Microstructure of coating layers encapsulating Fe particle. (c) Mag-
netic hysteresis loops of C-coated Fe particles with maximum DC applied field of
0.8 MAm−1 . (d) Dependence of magnetization on temperature for commercially
available carbonyl Fe, BN-coated Fe and C-coated Fe particles. Magnetization at
each temperature is normalized with reference to that at 25◦ C
layers of ∼ 5 nm in thickness (Fig. 12b). The spacings of the lattice fringes

in the layers revealed that the crystal structure is graphitic carbon with the
spacing of 0.34 nm of {002} plane. The C-coated Fe nanoparticles also grew
nanotubes of 50–100 nm in diameter, ∼ 1 µm in length, and ∼ 10 nm in wall
thickness (Fig. 12a). Detailed observation of these tubes reveals that they
have a multiwalled structure with something like joints inside. Multiwalled
carbon nanotubes have been formed with diameters in the range from 1.4 nm
to at least 100 nm [64,65]. So, the carbon nanotubes in this sample have large
diameters.
The magnetic properties of C-coated Fe nanoparticles in Fig. 12c were
measured at room temperature. A magnetic field higher than 0.6 MAm−1
must be applied in order to saturate magnetizations, because the demag-
netizing factor of Fe nanoparticles would be larger than that of bulk Fe.

Coercivities are very low, so the samples exhibit soft magnetic properties.
Magnetization (Ms ) and coercivity (Hc ) values of the sample were measured
to be 193 Am2 kg−1 and 1.1 kAm−1 under an applied field of 1.6 MAm−1 .
The Ms value is 87% of the value of bulk Fe [66]. Ms values of Fe or Co
nanoparticles with C nanocoatings have been reported to be 40–60% of those
of the bulk metals, and the Hc values are around several hundred oersteds
(∼ 20 kAm−1 ) [67, 68]. In particular, the C-coated Fe nanoparticles in this
work have much higher Ms values than the reported values. The Hc values of
the C-coated Fe nanoparticles are lower as well. Herzer showed that the Hc
of soft magnetic metal particles depends on the particle size [69]. Lower Hc
values would be due to large particle sizes. The particle sizes in this work are
100–500 nm, about ten times larger than reported nanoparticle sizes. Con-
sidering that coercivity disturbs soft magnetic properties, it is obvious that
this C-coated Fe nanoparticles is excellent as soft magnetic materials.
In order to investigate resistance to oxidation, Ms values were measured,
heating the C- and BN-coated Fe nanoparticles from 27◦ C to 800◦C in air
(Fig. 12d). Commercially available carbonyl Fe powders, which have an av-
erage diameter of 3 mm, were also measured as a reference. The Ms values of
all three samples decrease gradually as temperature increases, and fall to zero
abruptly at elevated temperatures. The Ms of the carbonyl Fe disappears at
∼ 600◦ C. The BN- and C-coated Fe samples still keep their Ms at more than
60% of the initial value even at 600◦ C, and decrease to zero at ∼ 800◦ C. As
the Curie temperature of magnetite (Fe3 O4 ) is 587◦ C (‘T1’ in Fig. 12d), the
degradation of Ms in carbonyl Fe could be explained as oxidation and trans-
formation of Fe into Fe3 O4 . The temperature of ∼ 800◦ C coincides with the
Curie temperature of iron (767◦ C, indexed as ‘T2’ in the figure). Accordingly,
the BN- and C-coated Fe nanoparticles exhibit resistance to oxidation.
6.3 Possibility of H2 Gas Storage in Carbon Nanocages
Clean hydrogen energy is expected as substitute for oil energy in twenty-first

century. LaNi5 H6 is already used as H2 gas storage material. However, the H2
gas storage ability is only 1 wt. % because of large atomic number of La and
Ni. On the other hand, fullerene materials, which consist of light elements
such as boron, carbon, and nitrogen, would store more H2 gas compared to
the metal hydrides. It was reported that H2 gas was stored in alkali-doped
carbon nanotubes under ambient pressure and moderate temperature [70].
Collision between C60 and alkali-metal ions was also calculated to investigate
a possibility of Li@C60 by molecular dynamics calculations [71]. Here, the
possibility of H2 gas storage in fullerene materials was investigated.
The condition of H2 gas storage into C60 cage and H2 gas discharge
from C60 were calculated by molecular dynamics software (MASPHYC). The
conditions were set as follows: pressure (p) and temperature (T ) are con-
stant, or volume (V ) and temperature are constant. H2 gas storage into C60
Fig. 13. (a) Molecular dynamics calculation of H2 gas storage in C60 cage.
(b) Atomic structure models of C186 nanotube. (c) Structure models wherein the
H2 molecule passes from pentagonal and hexagonal rings of C60 . (d) Structural
model of H atom chemisorbed on C60
cages is shown in Fig. 13a. This model is calculated under the condition of
V = 8.0 × 10−27 m3 and T = 150 K. The H2 molecules pass through the
hexagonal rings of C60 cage at 20 ps. After introducing H2 molecule into the
C60 cluster at 40 ps, H2 is stored and stable in C60 . The condition of H2 gas
discharge from C60 is similar to that of H2 gas storage, which is as follows:
T = 300 K and p = 0.15 GPa. Although electronic structure calculation of
hydrogen in C60 was reported [72], the present work showed the molecular
dynamics of H2 molecule storage in C60 cluster.
The H2 molecules are considered to enter from the cap of nanotubes,
where the curvature is finite, as shown in Fig. 13b. The possibility of H2 gas
storage in C nanotubes was investigated by using first-principle calculations.
In the present work, C60 was used to calculate the end-cap structure of the
nanotube. In this case, only cap structures should be in the calculations. In
the nanotubes, H2 molecule would invade from the cap of tube in the same
way because of coulomb barriers of the tube wall. C60 clusters would have
energy barriers for H2 molecules to pass through pentagonal and hexagonal
rings. Figure 13c shows structure models wherein the H2 molecule passes
from pentagonal and hexagonal rings of C60 . Single-point energies were cal-
culated with changing set points of H2 molecule from the center of the cage
at intervals of 0.1 nm. Energy barriers were calculated for H2 molecules to
pass through pentagonal and hexagonal rings, and they are 23 eV and 16 eV,
respectively [73].
Figure 13d is a structural model of hydrogen atoms chemisorbed on a
C60 cluster. Atoms bonded with hydrogen are moved outside from the clus-
ters. The hydrogenated energies for C60 H and C60 H2 were calculated to be
−0.18 and 0.6 eV, respectively [74]. Hydrogen bonding with a carbon atom
of C60 is more stable than C60 . When two hydrogen atoms were chemisorbed
on carbon clusters, energies of carbon clusters increased.
To investigate stability of H2 molecules in clusters, energies were calcu-
lated for H2 molecules placed in the center of a C60 cluster, as summarized
in Table 3. Structural models are shown in Fig. 13e. When more than 22 H2
molecules were introduced in C60 , hydrogen atoms chemisorbed inside the
cluster [75]. When more than 26 H2 molecules were introduced in the C60
clusters, 22 H2 molecules were in the cluster and 8 hydrogen atoms chemi-
sorbed inside the clusters and the C–C bond was broken.
6.4 One-Dimensional Self-Organization of Nanocapsules
The recent speed-up of ultra-large scale integrated circuits (ULSI) of semi-

conductor devices is based on the minimization of the design rule. New prop-
erties are also expected by the formation of low dimensional arrays of quan-
tum dots. The photolithography technique is used for the formation of these
nanostructures, and the ULSI with 0.07 µm rule will be produced in 2009 [76].
However, the limit of the design rule is reported to be 0.05 µm (= 50 nm) by
using extreme ultraviolet lithography [77]. It is believed that establishment
Table 3. Energy of C60 cluster with hydrogen. A C–C bond was broken in case of
8 chemisorbed hydrogen atoms (∗ )
Introduced Optimized Heat of formation Hydrogen atoms Hydrogen storage

H2 H2 (eV) chemisorbed (wt. %)
inside cluster
0 0 35.21 0 0
20 20 127.54 0 5.6
22 22 143.01 0 6.1
25 25 164.87 4 6.9
26 26 169.63 8 (∗ ) 7.2
30 24 165.59 4 6.7
of nanostructure-formation techniques by self-organization of nanoparticles

is worth pursuing in order to overcome the lithography limit. Spontaneous
formation of well-organized nanostructures is intriguing for both scientific
research and future electronic-device application. However, dimensionality of
this self-organization has been two and three [78, 79], and one-dimensional
self-organization has been needed for the formation of “nanowiring” for the
future ULSI devices.
Gold nanoparticles and nanowires encapsulated in carbon nanocap-
sules and nanotubes were spontaneously formed from one-dimensional self-
organized gold nanoparticles on carbon thin films by annealing at low temper-
atures of 200–400◦C, as shown in Figs. 14a–c [30, 31]. The gold crystals inside
the nanotubes are distorted by the crystal growth of the nanowires, as shown
in Figs. 14d and 14e. Lattice expansion and reduction are observed parallel
and perpendicular to the growth axis, respectively. It is believed that the dis-
tortion of the nanowire would be due to the coalescence of nanoparticles along
the growth direction at the low temperatures. A conductivity (dI/dV ) curve
of a single carbon nanocapsule with Au nanoparticles is shown in Fig. 14f,
which indicates the V-shape behavior with the gap of ∼ 0.5 V [80]. This would
be due to the disordered carbon layers at the carbon/Au interface as shown
in Fig. 14g.
A schematic illustration of self-organized Au nanoparticles, carbon nano-
capsules, and carbon nanotube is shown in Fig. 14h. For the one-dimensional
positioning, adhesive force due to the step edge and substrate are mainly
used. The ordinary ordering mechanism for nanoparticle self-organization is
due to the internanoparticle force by α-terpineol and adhesion by the sub-
strate. The width of the one-dimensional line could be controlled by the
interactive force of α-terpineol. The present result indicates that the one-
dimensional arrangement is strongly dependent on the atomic step edge of
the substrate. The difference in the formation of carbon nanocapsules and
nanotubes would be due to the distance between the gold nanoparticles. Al-
though the atomic step edge is formed by the common sample preparation
a b c
2nm
2nm
1-11 e
d 200
C
5nm
000 Au
11-1
㨪5%
g
dI/dV (arb. units)
(f) 0.5 V
Au
-0.5 -0.25 0 0.25 0.5 1nm
Sample bias (V)
(h) ǩ-terpineol
1D arrangement of CNC
Au
3nm
a-carbon 15nm
graphite layers
Au
ǩ-terpineol 5nm
Au Au
3nm
͠
a-carbon 15nm amorphous carbon
200-400
CNT with Au nanowire
1D model
graphite layers
C layer 10nm
Au
Au
3nm
amorphous carbon
a-carbon 15nm
Fig. 14. (a) Low-magnification image of one-dimensional self-organized carbon

nanocapsules with Au nanoparticles on amorphous carbon film. (b) Alternate
arrangement of different size of carbon nanocapsules with gold nanoparticles.
(c) HREM image of Au nanowire in carbon nanotube. (d) Fourier transform of
nanotube. (e) Structure model of Au{111}/C{002} interface. (f ) Conductivity
curve of a single carbon nanocapsule with Au nanoparticle. (g) Enlarged HREM
image of carbon/Au interface. (h) Schematic illustration of one-dimensional self-
organized Au nanoparticles, carbon nanocapsules and carbon nanotube
technique in the present work, the atomic step could be controlled by or-
dinary lithography techniques such as electron beam, and is expected as a
nanofabrication technique for future electronic devices. The present result is
expected as a future fabrication technique for self-assembled ULSI nanowires
and quantum dots protected by nanocapsules and nanotubes at scales beyond
the limits of current photolithography.
7 Conclusion
Various C fullerene materials were synthesized by polymer pyrolysis, chem-
ical reaction, DC-RF hybrid arc discharge, and electron-beam irradiation.
Atomic structure and formation mechanisms were investigated by HREM,
and structural models were proposed. Measurements of PL and magnetic
properties of C nanocapsules showed a higher energy shift of luminescence
and good soft magnetic properties. The present work indicates that the new
C fullerene materials with various atomic structures and properties can be
produced by various synthesis methods, and are expected to be the future
nanoscale devices.
Acknowledgements
The authors would like to thank T. Hirano, M. Kuno, H. Kitahara, K. Hiraga,

E. Aoyagi, H. Kubota for their useful discussion, experimental help, and
collaborative work.
References
[1] H. W. Kroto, J. R. Heath, S. C. O’Brien, R. F. Curl, R. E. Smalley: Nature
318, 162 (1985) 187, 190, 191
[2] S. Iijima: Nature 354, 56 (1991) 187
[3] D. Ugarte: Nature 359, 707 (1992) 187
[4] Y. Saito, T. Yoshikawa, M. Inagaki, M. Tomita, T. Hayashi: Chem. Phys. Lett.
204, 277 (1993) 187
[5] A. Krishnan, E. Dujardin, M. M. J. Treacy, J. Hugdahl, S. Lynum, T. W. Ebbe-
sen: Nature 388, 451 (1997) 187
[6] Y. Saito: Nature 392, 237 (1998) 187
[7] C. N. R. Rao, R. Seshadri, A. Govindaraj, R. Sen: Mater. Sci. Eng. R 15, 209
(1995) 187
[8] K. Suzuki, A. Kasuya, Y. Nisina: J. Phys. Chem. Solid 54, 1849 (1993) 187,
204
[9] J. Sloan, J. Cook, M. L. H. Green, J. L. Hutchinson, R. Tenne: J. Mater.
Chem. 7, 1089 (1997) 187
[10] L. Rapoport, Y. Bilik, Y. Feldman, M. Homyonfer, S. R. Cohen, R. Tenne:
Nature 387, 791 (1997) 187
[11] B. R. Elliott, J. J. Host, V. P. Dravid, M. H. Teng, J. H. Hwang: J. Mater.

Res. 12, 3328 (1997) 187
[12] T. Oku, T. Hirano, M. Kuno, T. Kusunose, K. Niihara, K. Suganuma: Mater.
Sci. Eng. B 74, 206 (2000) 187, 205
[13] T. Oku, M. Kuno, H. Kitahara, I. Narita: Int. J. Inorg. Mater. 3, 597 (2001)
187
[14] T. Oku, I. Narita, A. Nishiwaki, N. Koi: Defects and Diffusion Forum 226-228,
113 (2004) 187
[15] T. Sato, H. Ahmed, D. Brown, B. F. G. Johnson: J. Appl. Phys. 82, 696 (1997)
189, 196
[16] D. L. Klein, R. Roth, A. K. L. Lim, A. P. Alivisatos, P. L. McEuen: Nature
389, 699 (1997) 189
[17] G. Schmid: Chem. Rev. 92, 1709 (1992) 189
[18] T. Oku: J. Ceram. Soc. Jpn. 109, S17 (2001) 189
[19] T. Oku: Sci. Tech. Adv. Mater. 5, 657 (2004) 189
[20] T. Oku: J. Phys. Chem. Solids 65, 363 (2004) 189
[21] T. Oku, I. Narita, A. Nishiwaki: Mater. Manuf. Process. 19, 1215 (2004) 189
[22] T. Oku, K. Niihira, K. Suganuma: J. Mater. Chem. 8, 1323 (1998) 189
[23] T. Oku, T. Hirano, S. Nakajima, K. Suganuma: J. Mater. Res. 14, 4266 (1999)
189
[24] H. Tokoro, S. Fujii, T. Oku: J. Mater. Chem. 14, 253 (2004) 189, 205
[25] H. Tokoro, S. Fujii, T. Oku: J. Magn. Magn. Mater. 290, 141 (2005) 189, 205
[26] I. Narita, T. Oku, K. Suganuma, K. Hiraga, E. Aoyagi: J. Mater. Chem. 11,
1761 (2001) 190
[27] T. Oku, I. Narita, A. Nishiwaki: Diamond Relat. Mater. 13, 1337 (2004) 190,
196
[28] T. Hirata, N. Motegi, R. Hatakeyama, N. Sato, T. Oku, T. Mieno, N. Y. Sato,
H. Mase, M. Niwano, N. Miyamoto: Jpn. J. Appl. Phys. 39, L1130 (2000) 190,
192
[29] T. Oku, T. Hirata, N. Motegi, R. Hatakeyama, N. Sato, T. Mieno, N. Y. Sato,
H. Mase, M. Niwano, N. Miyamoto: J. Mater. Res. 15, 2182 (2000) 190, 202
[30] T. Oku, K. Suganuma: Chem. Comm. p. 2355 (1999) 190, 210
[31] T. Oku, K. Suganuma: Microelectron. Eng. 51–52, 51 (2000) 190, 210
[32] C. Piskoti, J. Yarger, A. Zettl: Nature 393, 771 (1998) 190
[33] J. C. Grossman, M. Côté, S. G. Louie, M. L. Cohen: Chem. Phys. Lett. 284,
344 (1998) 190
[34] T. Oku, H. Kubota, T. Ohgami, K. Suganuma: Carbon 37, 1299 (1999) 191
[35] T. Oku, J. O. Bovin, S. Nakajima, H. Kubota, T. Ohgami, K. Suganuma:
Nanostructured Materials 12, 563 (1999) 191
[36] W. I. F. David, R. M. Ibberson, T. Matsuo: Proc. R. Soc. Lond. A 442, 129
(1993) 191
[37] T. A. Murphy, T. Pawlik, A. Weidinger, M. Höhne, R. Alcala, J. M. Spaeth:
Phys. Rev. Lett. 77, 1075 (1996) 191
[38] P. W. Stephens, G. Bortel, G. Faigel, M. Tegze, A. Jánossy, S. Pekker, G. Os-
zlanyi, L. Forró: Nature 370, 636 (1994) 191
[39] H. Shinohara, H. Sato, M. Ohchochi, Y. Ando, T. Kodama, T. Shida, T. Kato,
Y. Saito: Nature 357, 52 (1992) 192
[40] S. Stevenson, G. Rice, T. Glass, K. Harich, F. Cromer, M. R. Jordan, J. Craft,

E. Hadju, R. Bible, M. M. Olmstead, K. Maitra, A. J. Fisher, A. L. Balch,
H. C. Dorn: Nature 401, 55 (1999) 192
[41] H. Shinohara, H. Yamaguchi, N. Hayashi, H. Sato, M. Ohko-hchi, Y. Ando,
Y. Saito: J. Phys. Chem. 97, 4259 (1993) 192
[42] K. Sueki, K. Akiyama, T. Yamauchi, W. Sato, K. Kikuchi, S. Suzuki,
M. Katada, Y. Achiba, H. Nakahara, T. Akasaka, K. Tomura: Fullerene Sci.
Tech. 5, 1435 (1997) 192
[43] C. Moeschel, M. Jansen: Chem. Ber. Recl. 130, 1761 (1997) 192
[44] T. S. M. Wan, H. W. Zhang, T. Nakane, Z. Xu, M. Inakuma, H. Shinohara,
K. Kobayashi, S. Nagase: J. Am. Chem. Soc. 120, 6806 (1998) 192
[45] P. Kuran, M. Krause, A. Bartl, L. Dunsch: Chem. Phys. Lett. 292, 580 (1998)
192
[46] J. Lu, X. Zhang, X. Zhao: Appl. Phys. A 70, 461 (2000) 192
[47] Y. Lian, Z. Shi, X. Zhou, X. He, Z. Gu: Carbon 38, 2117 (2000) 192
[48] R. Hatakeyama, T. Hirata, H. Ishida, N. Sato: Appl. Phys. Lett. 73, 888 (1998)
192
[49] K. Shiga, K. Ohno, Y. Kawazoe, Y. Maruyama, T. Hirata, R. Hatakeyama,
N. Sato: Mater. Sci. Eng. A 290, 6 (2000) 192
[50] T. Oku, I. Narita, R. Hatakeyama, T. Hirata, N. Y. Sato, T. Mieno, N. Sato:
Diamond Relat. Mater. 11, 935 (2002) 192
[51] J. O. Bovin, R. Wallenberg, D. J. Smith: Nature 317, 47 (1985) 196
[52] T. Oku, G. Schmid, K. Suganuma: J. Mater. Chem. 8, 2113 (1998) 196
[53] Y. Wang, G. T. Hager, P. C. Eklund: Phys. Rev. B 45, 14396 (1992) 198
[54] X. Liu, T. G. Schmalz, D. J. Klein: Chem. Phys. Lett. 188, 550 (1992) 200
[55] H. Kroto: Nature 329, 529 (1987) 200
[56] D. E. Manolopoulos, P. W. Fowler: J. Chem. Phys. 96, 7603 (1992) 200
[57] A. J. Stone, D. J. Wales: Chem. Phys. Lett. 128, 501 (1986) 200
[58] L. C. Qin, X. Zhao, K. Hirahara, Y. Miyamoto, Y. Ando, S. Iijima: Chem.
Phys. Lett. 408, 50 (2000) 200
[59] S. Iijima: Carbon nanotubes: past, present, and future, Physica B 323, 1 (2002)
203
[60] M. Kusunoki, M. Rokkaku, T. Suzuki: Appl. Phys. Lett. 71, 2620 (1997) 205
[61] F. E. Kruis, H. Fissan, A. Peled: J. Aerosol Sci. 29, 511 (1998) 205
[62] J. Ding, W. F. Miao, P. G. McCormick, R. Street: Appl. Phys. Lett. 67, 3804
(1995) 205
[63] Z. Turgut, N. T. Nuhfer, H. R. Piehler, M. E. McHenry: J. Appl. Phys. 85,
4406 (1999) 205
[64] T. Katayama, H. Araki, K. Yoshino: J. Appl. Phys. 91, 6675 (2002) 206
[65] R. Baughman, A. Zakhidov, W. Heer: Science 297, 787 (2002) 206
[66] Maruzen Co. Ltd., Tokyo: National Astronomical Observatory, Chronological
Scientific Tables 2000 (1999), p. 491 207
[67] T. Hayashi, S. Hirono, M. Tomita, S. Umemura: Nature 381, 772 (1996) 207
[68] X. X. Zhang, G. H. Wen, S. Huang, L. Dai, R. Gao, Z. L. Wang: J. Magn.
Magn. Mater. 231, L9 (2001) 207
[69] G. Herzer: IEEE Trans. Magn. 26, 1397 (1990) 207
[70] P. Chen, X. Wu, J. Lin, K. L. Tan: Science 285, 91 (1999) 207
[71] K. Ohno, Y. Maruyama, K. Esfarjani, Y. Kawazoe: Phys. Rev. Lett. 76, 3590
(1996) 207
[72] J. Cioslowski: J. Am. Chem. Soc. 113, 4139 (1991) 209
[73] T. Oku, I. Narita: Physica B 323, 216 (2002) 209
[74] N. Koi, T. Oku: Sci. Tech. Adv. Mater. 5, 625 (2004) 209
[75] N. Koi, T. Oku: Sol. State Comm. 131, 121 (2004) 209
[76] The Semiconductor Industry Association, California: National Technology
Roadmap for Semiconductor (1997) 209
[77] J. E. Bjorkholm, J. Bokor, L. Eichner, R. R. Freeman, J. Gregus, T. E. Jewell,
W. M. Mansfield, A. A. Mac Dowell, E. L. Raab, W. T. Silfvast, L. H. Szeto,
D. M. Tennant, W. K. Waskiewicz, D. L. White, D. L. Windt, O. R. Wood: J.
Vac. Sci. Technol. B 8, 1509 (1990) 209
[78] C. J. Kiely, J. Fink, M. Brust, D. Bethell, D. J. Schiffrin: Nature 396, 444
(1998) 210
[79] K. Pohl, M. C. Bartelt, J. de la Figuera, N. C. Barttelt, J. Hrbek, R. Q. Hwang:
Nature 397, 238 (1999) 210
[80] H. Kitahara, T. Oku: J. Ceram. Process. Res. 5, 89 (2004) 210
Index
atomic structures, 187 C74 :Si, 192–195
C76 , 194
C168 , 198, 199 C82 , 192
C180 , 200, 201 C84 , 192, 198, 199
C186 , 198 C84 :C24 , 197
C20 , 200 C84 D2 , 198, 200
C2610 , 196 C84 D2d , 200
C264 , 196 C84 Td , 198
C276 , 198, 199 C84 Td , 200
C276 :C84 , 198, 199 carbon clusters, 194
C28 , 196 endohedral clusters, 192
C36 , 190, 200
C408 , 198, 199
fullerene, 187–192
C60 , 187, 191–194, 198, 200
fullerene materials, 188, 189, 207
C60 :Li, 207
giant fullerene, 192, 199
C60 :N, 192
C60 :Rb, 192
C60 with hydrogen, 207–211 graphitization, 205
doped C60 , 191
C67 , 203 high-resolution electron microscopy
C70 , 190, 194 (HREM), 191, 192
C72 , 194 highest occupied molecular orbital
C74 , 192–195, 203 (HOMO), 194
C74 :Ca, 194 HOMO-LUMO gap, 194, 203
C74 :Eu, 194 hybrid arc-discharge, 187, 189, 190,
C74 :Gd, 194 192, 203
C74 :La, 194 hydrogen storage, 207–209, 211
lowest unoccupied molecular orbital mininanotube, 200

(LUMO), 194 multi-walled nanotubes (MWNTs),
200
metallofullerene, 187, 190, 192 multiwalled nanotubes (MWNTs),
molecular dynamics, 207 206
nanocages, 187, 189, 207 single-walled nanotubes (SWNTs),
nanocage materials, 189 200
nanocapsules, 187, 190, 192, 200, nanowires, 200, 203, 210, 211
202–205, 207–211
self-organization, 187, 209 onions, 187, 196–199
one-dimensional self-organization, tetrahedral onions, 196–198
209
nanohorns, 187, 202, 203 photoluminescence (PL), 187, 189, 204
nanoparticles, 210
gold nanoparticles, 210, 211 SiC, 200, 202–205
nanostructures
hollow-cage-nanostructures, 187 ultra-large scale integrated circuits
nanotubes, 187, 192, 196, 201–203, 208, (ULSI), 209
210, 211 self-assembled ULSI, 212
Hard Amorphous Hydrogenated Carbon Films
and Alloys
Fernando L. Freire Jr.
Pontifı́cia Universidade Católica do Rio de Janeiro, Departamento de Fı́sica

Rua Marquês de São Vicente 225, Rio de Janeiro, BR-22452-900, RJ, Brazil
lazaro@vdg.fis.puc-rio.br
Abstract. Hard hydrogenated amorphous carbon films are presently being used
in a wide variety of applications as protective coatings. The properties of the car-
bon films are essentially controlled by the ratio between the number of carbon
atoms in sp2 and sp3 hybridization states. The properties of the films, which are
closely related to the film microstructure, can be tuned by the deposition tech-
nique employed and by the growth conditions, with the energy of the impinging
species playing the main role in the control of the carbon bonding hybridization.
Another way to improve the properties of carbon films is through the incorpora-
tion of several elements. In this work, the attention is focused on films deposited
by plasma-enhanced chemical vapor deposition, discussing the effects of nitrogen
and fluorine incorporation on the film microstructure and on the mechanical and
tribological properties.
1 Introduction
Amorphous carbon films are nanostructured materials made of clusters of
sp2 -hybridized carbon atoms, with a typical size of a few nanometers, con-
nected to each other by sp3 -hybridized carbon atoms. The electronic and op-
tical properties of these materials are mainly controlled by the size of the sp2
clusters, while the mechanical properties are given by the degree of intercon-
nectivity of the amorphous skeleton, i.e., the fraction of sp3 -hybridized carbon
atoms present in the matrix [1]. Film properties can be tuned by choosing
the deposition technique with the appropriated deposition conditions. Among
the different deposition parameters, the energy of the impinging ions plays
the more important role and controls the sp2 /sp3 -hybridized carbon atoms
ratio. The subplantation model suggests that the deposition of hyperthermal
carbon species is a shallow implantation process, where the incorporation of
carbon species in subsurface layers followed by internal stresses is the dom-
inant mechanism, responsible for the formation of a sp3 -carbon-rich dense
phase, the diamondlike phase [2]. Recently, this model was improved with
the inclusion of the role of the interstitials, which are physically and ener-
getically different from the bulk sp2 and sp3 carbon atoms [3]. In this way,
films with properties typical of polymers, graphite or diamond can be syn-
thesized. Carbon films are used in a wide variety of applications as protective
218 Fernando L. Freire Jr.
coatings. In particular, films deposited by magnetron sputtering, plasma-en-

hanced chemical vapor deposition (PECVD) and filtered cathodic vacuum
arc (FCVA) are used as overcoats for biomedical implants [4], computer hard
disks [5] and as spacer tools for electron guns of cathode ray tubes [6].
The incorporation of hydrogen in the amorphous skeleton also plays an
important role. The amorphous hydrogenated carbon films (a-C:H) can be
deposited by both sputtering and PECVD techniques in a large range of
sp2 /sp3 -hybridized carbon atom ratios. The hydrogen content in these films
can vary from 0 to 50 at. %. Depending on the deposition conditions, sub-
strate bias voltage, substrate temperature, and pressure and gas precursor
atmosphere, they can be either diamond-like (sp3 -rich films) or graphitic-like
(sp2 -rich films). However, a-C:H films with high H content and high frac-
tions of carbon atoms with sp3 hybridization have a polymeric character. It
was shown that the subplantation model could also be applied to describe
the mechanisms of a-C:H growth by PECVD. In fact, the internal stress and
hardness of a-C:H films deposited by rf-PECVD in pure methane atmosphere
show a dependence with the self-bias voltage that is well described by the
subplantation model [7].
One of the ways of changing, in a controlled way, a-C:H film properties
is through the incorporation during the film growth of different elements,
such as nitrogen (N), fluorine (F), silicon (Si) and many metals, for example,
titanium (Ti). The incorporation of Si, for example, seems to stabilize sp3 -
carbon bonds and to improve the thermal stability of a-C:H films [8]. On the
other hand, the incorporation of one or more of these elements can induce
the formation of new nanostructured materials, as the fullerene-like carbon
films, films with strongly interacting curved graphene planes. The fullerene-
like structure results from the curving of the graphene planes induced by
nitrogen incorporation and is similar to that found in carbon nanotubes and
fullerenes [9]. In recent years, an important fraction of the research effort was
dedicated to the study of nitrogen incorporation on carbon films [10]. The
main reason of this effort was the intention to synthesize the C3 N4 , proposed
by Liu and Cohen [11], to have mechanical properties comparable to those
of crystalline diamond. In spite of this, no clear experimental evidence of
the formation of C3 N4 has been presented until now [12]. Nitrogen incor-
poration into a-C:H films was found to modify the structure [13] and the
mechanical properties [14] of these films, as well as their electrical and opti-
cal properties [15]. In fact, the fraction of carbon atoms in sp3 hybridization
state decreases upon nitrogen incorporation [16]. In terms of the mechanical
properties, a reduction on the internal compressive stress was observed, with
minor changes in the mechanical hardness [14]. Concerning the modification
of electrical and optical properties, it was found that nitrogen could electron-
ically dope a-C:H films, with the simultaneous reduction of the electronic
defect density [15]. This reduction makes possible the use of a-C:H:N films as
Hard Amorphous Hydrogenated Carbon Films and Alloys 219
a semiconductor material. In addition, a-C:H:N films deposited by PECVD

showed a promising performance as electron field-emitters [17].
The incorporation of fluorine in a-C:H films has also attracted inter-
est [18]. In spite of the lubricant properties of poly(tetrafluoroethylene)
(PTFE) the recent research on the fluorine incorporation into a-C:H films
was mainly motivated by its electrical properties. In fact, to improve the
switching performance of future ultra-large scale integrated circuits (ULSI),
insulator films with dielectric constants less than that of SiO2 are needed to
reduce the capacitance of interlayer insulators [19]. Fluorinated amorphous
carbon (a-C:H:F) films have been proposed as possible candidates due to
their dielectric constant [20]. These films can be plasma deposited in a wide
compositional range, the film fluorine content being determined primarily by
the atmosphere during deposition [21]. It has been found that the dielectric
constant decreases with the increase of the F/H ratio of the precursor at-
mosphere for films deposited by PECVD [22]. Notwithstanding the interest in
the dielectric properties of fluorinated amorphous carbon films, few attempts
to combine the mechanical properties of a-C:H films with the tribological
characteristics of PTFE have been made. The effect of fluorine incorpora-
tion on the mechanical [23] and tribological [24] properties of a-C:H films
has received more attention only in the last few years. As expected, fluorine
incorporation increases the surface hydrophobicity and reduces the friction
coefficient [25]. Besides, the effect of ionic bombardment during a-C:H:F film
growth has not been investigated in detail [26].
Up to now, only a few studies have concentrated on the simultaneous
incorporation of F and N into a-C:H films [27]. They were focused to the
structural characterization and the electrical properties of the films [28]. In
a recent investigation, it was shown that N atoms were incorporated into the
aromatic rings and as electronegative groups directly bonded to sp2 -carbon
in a-C:H:F films deposited by PECVD [29]. The incorporation of nitrogen,
which occurs at the expense of carbon atoms, induces modifications on the
surface hydrophobicity and friction behavior of the films, indicating a route
to control the surface wettability of these films [30].
In this review, a comparative study of the incorporation of N and F into
a-C:H films deposited by PECVD using CH4 –N2 and CH4 –CF4 mixtures
as precursor atmospheres is presented. The effects on the film microstructure
and mechanical properties will be discussed emphasizing the role of the energy
of the bombarding species. Tribological properties will also be discussed, as
well as the thermal stability of those films. Results of an investigation on the
incorporation of N into a-C:H:F films deposited by PECVD using CH4 –CF4 –
N2 mixtures as precursor atmospheres will also be presented, with special
attention to the tribological properties of a-C:H:F:N films.
Fig. 1. Nitrogen and fluorine content in a-C:H:N and a-C:H:F films as a function
of the methane partial pressure
2 Film Growth and Chemical Composition
a-C:H:N and a-C:H:F films were deposited by the plasma-assisted decompo-

sition of CH4 –N2 and CH4 –CF4 mixtures, onto Si substrates mounted on the
water-cooled copper cathode feed by an rf power supply (13.56 MHz). In this
method, the capacitive coupling allows the development of an average in time
DC negative potential at the powered electrode, the self-bias potential, V b ,
which extracts and accelerates ions from the plasma towards the powered
electrode. For deposition chamber pressure around 10 Pa, the mean energy
of those ions is around (1/3)eV b , where e is the electron charge [1]. Results
obtained from two series of films will be presented. In the first set of samples,
during the deposition V b was kept fixed at −350 V, while the CF4 and N2
pressures were changed from 0 to 80% and 0 to 50%, respectively. For the
second series, V b ranged from −100 V to −700 V with the N2 partial pres-
sure fixed at 40% of the total pressure for a-C:H:N films, while the partial
pressure of CF4 was 67% in the case of a-C:H:F film deposition. For the sake
of comparison, a-C:H films were deposited in pure methane atmosphere in
the same range of self-bias voltage. The desired V b value was adjusted by
changing the rf power in the range of 10–80 W. The total gas inlet flux was
fixed at 3 sccm. The pressure in the deposition chamber was 10 Pa. A 10 nm
thick a-C:H buffer layer was deposited without breaking the vacuum prior to
the deposition of the film in order to enhance the film adhesion. Without the
buffer layer, the a-C:F:H films peel off. Details of the experimental apparatus
can be found elsewhere [23]. The film composition was determined by ion
beam analysis (IBA): Rutherford backscattering spectrometry (RBS), elastic
recoil detection analysis (ERDA) and nuclear reaction analysis (NRA) [31].
Fig. 2. Deposition rate of a-C:H:N and a-C:H:F films as a function of the methane
partial pressure
The chemical compositions of the films deposited while changing the pre-
cursor atmosphere are shown in Fig. 1, where the nitrogen and fluorine con-
tent of a-C:H:N and a-C:H:F films, respectively, were plotted as functions
of the partial pressure of methane. An increase of the F and N content in
the films occurs upon the the increase of the partial pressure of CF4 and
N2 . While in the case of a-C:H:N films the nitrogen incorporation occurs at
the expenses of the carbon content of the films, the hydrogen concentration
remaining constant within experimental errors around 20 at. %, in the case
of a-C:H:F films the incorporation of fluorine occurs at the expenses of the
hydrogen content, which decreases from 19 at. % to 1 at. %, in the range of
partial pressures considered in Fig. 1. The observed upper limits for the incor-
poration of nitrogen and fluorine in the films have different reasons and can
be understood with the help of Fig. 2, where the deposition rates of a-C:H:N
and a-C:H:F films are plotted as functions of the methane partial pressure.
The a-C:H:N deposition rate presents a sharp decrease when nitrogen par-
tial pressure increases, while the deposition rate of a-C:H:F films increases
by a factor of 2 when the partial pressures of CF4 increases from 0 to 80% of
the total pressure. For higher CF4 partial pressures, no film deposition but
substrate erosion occurs. It is important to note that for both a-C:H:N and
a-C:H:F film deposition, in order to keep V b constant at −350 V, the power
applied to the plasma is almost constant in the entire range of partial pres-
sures (CF4 or N2 ) studied, around 50 W. The deposition rate reduction due
to nitrogen incorporation observed in a-C:H:N film deposition was attributed
to the onset of a kind of chemical sputtering process arising from energetic
N+2 ion bombardment of the film growing surface [32]. In fact, it was found
that low-energy N+ 2 ion bombardment of pure amorphous carbon results in
carbon atom removal, at a rate of about 0.5 C atom per N+ 2 ion [33]. However,
there is no justification for the absence of the symmetric situation in carbon–
nitrogen film deposition process, i.e., nitrogen atoms being sputtered out by
carbon-carrying ions. We must consider both processes in the simulation of
ion beam deposition of carbon–nitrogen films. Besides, the possibility of N2
evaporation during film growth must be included. This situation is likely to
occur since N-atom subsurface penetration must favour N–N bond formation
with N atoms already in the film. The formation of other volatile species, such
as CN can also occur. These mechanisms are increasingly more probable for
deposition conditions that result in increasing nitrogen contents, reducing
the deposition rates. Therefore, the growth kinetics of carbon–nitrogen films
may be pictured as a competition between aggregation and erosion resulting
from the impinging of different species on the film-growing surface [34].
The increase of the deposition rate of a-C:H:F films with higher CF4 par-
tial pressure can be explained as follows: the growth of a-C:H films is mainly
due to the physisorption of hydrocarbon radicals onto the surface followed
by chemisorption of such radicals induced by energetic processes such as ion
bombardment. The existence of dangling bonds at the surface is expected
to enhance the sticking probability of carbon-carrying radicals [35]. As the
partial pressure of CF4 increases, the hydrogen concentration in the plasma
decreases together with the increase of the fluorine concentration. Fluorine
atoms can react on the surface by chemically eroding H from hydrocarbon
radicals, forming volatile HF and leaving a dangling bond on the surface. In
fact, a similar mechanism for dangling bond formation during the deposition
of a-C:H films has been reported [36]: atomic hydrogen chemically erodes hy-
drogen bonded on the surface of the growing film, leaving a dangling bond and
enhancing the sticking probability of CH3 radicals by a factor of 100. The hy-
drogen erosion induced by fluorine atoms is expected to be more efficient than
that induced by hydrogen atoms due to their much higher electronegativity.
In this way, the higher the CF4 concentration, the higher the probability of
hydrocarbon radicals sticking onto the surface. Besides, the higher the CF4
partial pressure, the higher the probability for large molecule formation in
the gas phase due to the agglomeration of CFx radicals [37], providing a new
channel for the formation of a fluorinated film up to the limit where there
are not enough hydrocarbon radicals to prevent F-induced substrate etching,
which occurs for CF4 partial pressure higher than 80%.
The a-C:H:N film composition is almost constant, within experimental
errors, for samples deposited in CH4 –N2 mixtures with N2 partial pressure
of 40% and with V b in the range of −100 V to −700 V. The nitrogen con-
tent is around 12 at. %, while the hydrogen content is typically 20 at. %. On
the contrary, fluorine and hydrogen contents in a-C:H:F films show a strong
dependence with the self-bias voltage, as shown in Fig. 3. An increase of the
fluorine content for higher V b was observed. Since the sum of the fluorine and
hydrogen concentrations is roughly constant around 20 at. %, these results
Fig. 3. Fluorine and hydrogen content in a-C:H:F films as a function of the self-bias
voltage
Fig. 4. Deposition rate as a function of the self-bias voltage
suggest that fluorine is incorporated by substituting hydrogen. On the other

hand, the deposition rate increases with the increase of V b , as one can see
in Fig. 4. It can be understood as a result of the higher degree of plasma
dissociation due to the higher power needed to achieve higher V b . Also, the
contribution to a higher generation rate of nucleation sites on the film surface
due to the more energetic ion bombardment should be taken into account.
Fig. 5. Atomic density of a-C:H:N and a-C:H:F films as a function of the methane
partial pressure
3 Film Microstructure
The influence of the nitrogen and fluorine incorporation on the film micro-
structure was usually investigated following a multitechnique approach. Ra-
man and X-ray photoelectron (XPS) spectroscopies [38], as well as, infrared
absorption spectroscopy (IR), electron energy loss spectroscopy (EELS) [16]
and thermal-induced gas effusion experiments were used to probe film mi-
crostructure [39]. The atomic density can be evaluated by combining the
areal atomic density provided by IBA and the thickness of the samples, de-
termined by stylus profilometry [31].
The modifications in the film microstructure are illustrated by the results
of the atomic density presented in Fig. 5. In this range of nitrogen incorpora-
tion, the atomic density remains constant, but for fluorine content higher than
15 at. % (corresponding to a-C:H:F films deposited in a CH4 –CF4 atmosphere
with 33% of partial pressure of methane) the atomic density decreases from
1.3 to 1.1 × 1023 atoms/cm3 , indicating the occurrence of structural modifi-
cations in these films. As will be discussed in more detail below, this more
opened and relaxed structure has a polymeric character.
Three representative Raman spectra obtained from films deposited at the
same V b (−350 V) are shown in Fig. 6. The two spectra at the bottom
part of the figure obtained from a-C:H and a-C:H:N films are typical of a
diamond-like film, while the one at the top, obtained from a film with fluo-
rine content of 19 at. %, has a larger luminescence background, characteristic
of a polymer-like material [21]. The main features of the Raman spectra are
the two partially overlapping bands characteristic of all a-C:H films. They
are usually referred as the D (disorder) and the G (graphite) bands. The first
Fig. 6. Raman spectra obtained from a-C:H, a-C:H:N (11 at. % of N) and a-C:H:F
(19 at. % of F) films. The position of the D and G bands are indicated by arrows
band, centered at around 1370 cm−1 , corresponds to the so-called D band

and is associated with disorder-allowed zone edge modes of graphite that be-
come Raman active due to the lack of long-range order [40]. The second one,
which peaks around 1550 cm−1 , is known as the G band and is attributed to
E 2g -symmetry optical modes occurring at the Brillouin-zone center of crys-
talline graphite [40]. This band is usually observed in the Raman spectra
of the a-C:H films as well as in ta-C films, but it may peak at slightly dif-
ferent energies in the two forms of amorphous carbon, depending on their
sp3 fractional content [41]. No other band besides the second-order Raman
was observed in the spectra. We can interpret the Raman results according
to the approach adopted by Dillon et al. for the analysis of the spectra ob-
tained from a-C films [42]. Recently, Ferrari et al. discussed in details the
Raman results obtained from carbon nitride films. This work is also taken
into account in the present analysis [43]. It was proposed that both the in-
crease in the ratio of the intensities of the D and G bands (I D /I G ) together
with the shift to higher frequencies of the position of the maximum and the
decrease of the width of the G band are indications of an increase in the
size of the sp2 -hybridized carbon domains. To obtain the evolution of these
Fig. 7. (a) I D /I G intensity ratio and (b) sp2 fraction as functions of the incorpo-
rated nitrogen. The lines are eye-guides only
bands against the changes in the film composition, the Raman spectra were
decomposed using two Gaussian functions, after the removal of the lumines-
cence background by means of a polynomial best-fit procedure [44]. In the
case of a-C:H:F films, a continuous shift of the position of the maximum of
the G band from 1538 to 1556 cm−1 (typical fitting errors of 5 cm−1 ) and
an increase of the I D /I G ratio by a factor of 2 was observed, together with
a decrease of the G bandwidth.
The Raman results obtained from the a-C:H:N films show a similar
trend [13]. These results are presented in Fig. 7, together with the sp2 fraction
as a function of the nitrogen content in the films.
The G band position shifts from 1538 to 1550 cm−1 and the G band-
width reduces from 165 to 153 cm−1 , while I D /I G increases by 60% upon
nitrogen incorporation up to 11 at. %. These results are interpreted as being
due to an increase of the size of the sp2 clusters. The increase of the fraction
of sp2 -hybridized carbon atoms measured by EELS (Fig. 7b) supports this
interpretation.
The increase of CF4 and N2 partial pressures, and consequently the in-
corporation of F and N, induces an increase in the size of the sp2 -hybridized
carbon domains. However, it should be pointed out that visible Raman spec-
troscopy is 50–230 times more sensitive to sp2 sites than to sp3 sites [45]. In
this way, the above analysis cannot completely describe the structural mod-
ifications induced by changing the deposition conditions. To this aim, more
information can be obtained, perhaps, from the evolution of the luminescence
background, underlying the a-C:H:F Raman spectra. In fact, while the lumi-
nescence intensity is negligible in the spectra obtained from a-C:H:N films or
a-C:H:F films with fluorine contents up to 14 at. %, it increases in films with
greater fluorine contents. The increase in the luminescence background has
Fig. 8. XPS spectrum obtained

from an a-C:H:F film with 20
at. % of F. The main features of
the spectrum are indicated by ar-
rows
been discussed in terms of a structural rearrangement towards a polymer-like

structure [21], and this interpretation is reinforced by both the reduction of
the a-C:H:F film density upon fluorine incorporation and the XPS identifi-
cation of C–F2 bands, characteristic of fluorinated polymers, observed in the
films containing about 20 at. % of fluorine. The C1s XPS spectrum shown in
Fig. 8 can be fitted using four bands: C–C and C–H at 285.1 eV, C–CF at
287.1 eV, C–F at 289.3 eV and C–F2 at 291.5 eV [38]. The fluorine peak at
around 690 eV does not provide any information about the different chemical
bonds. XPS spectra taken from films deposited at lower CF4 partial pressure
are not shown in this figure. However, they show that higher CF4 partial
pressures lead to a relative increase in the intensity of fluorinated carbon
peaks. In fact, only for films with 20 at. % of fluorine, we observed the bands
corresponding to both C–F and C–F2 bonds. These trends are in good agree-
ment with the infrared results, which show an increase in the intensity of the
C–Fx bands when the fluorine content increases [23].
As discussed above, the ionic bombardment during a-C:H film growth
controls the final film properties. According to mass spectrometry measure-
ments in CH4 and CF4 low-power rf capacitively coupled plasmas, the main
ionic species in our plasmas are expected to be CF+ +
3 and CH3 ions [46]. In
+
the case of N2 plasma, N2 is the more important ionic species [47]. For V b on
the order of several hundred volts, these molecular ions promptly break up
upon colliding with the film surface. From simple mechanical arguments, it is
possible to show that the energy sharing between the atoms is proportional
to the mass ratio of the atom and the molecule. Clearly, for CH+ 3 ions this
energy is almost totally concentrated in the carbon atom, while for CF+ 3 ions
it is more equally divided between all atoms. In a few words, the CF+ 3 bom-
bardment is much less effective than that due to CH+ 3 or even N +
2 ions. In
the case of CF4 –CH4 atmospheres, when the methane partial pressure in the
plasma is reduced, the efficiency of the ionic bombardment is also reduced,
resulting in films with polymeric character.
Fig. 9. Atomic density as a func-

tion of V b
Fluorinated a-C:H films were also investigated by thermal-induced gas

effusion [39]. A structural transition from diamond-like to polymer-like films
as a result of the incorporation of fluorine could be inferred from the effusion
spectra. At low fluorine contents (up to 10 at. %) the material is relatively
compact and the effusion of hydrogen-related species (hydrogen molecules
and hydrocarbons) dominates. As the fluorine content is increased, hydro-
gen-related effusion is progressively substituted by the effusion of CF4 -related
species, confirming that fluorine atoms substitute hydrogen ones in the amor-
phous network. For high enough fluorine concentrations a strong change in
the effusion characteristics indicates that an interconnected network of voids
is present. Strong effusion of CF4 -related species is consistent with a surface
desorption process and can only be observed when CF2 and CF3 bonds are
present in the film microstructure and the void dimensions are large enough,
i.e., for films with the highest fluorine contents.
Thermal-induced gas effusion experiments are also used to study the ther-
mal stability of the films. Gas evolution experiments performed on a-C:H:N
films showed that any possible application for such films is limited to temper-
atures below 300◦ C. Above this temperature H and N losses provoke severe
structural changes [48]. These results confirm a previous study performed on
films annealed in vacuum, which reveals a decrease in thermal stability when
the amount of nitrogen incorporated in the film increases [49]. It was shown
that a-C:H films deposited by PECVD are thermally stable up to 450◦C, com-
pared with the 350◦C determined for a-C:H:N films with 8 at. % of N. The
thermal stability of a-C:H:F was also studied and the incorporation of fluo-
rine in the amorphous skeleton reduces the thermal stability of the films [50].
Concerning the application of a-C:H:F films as interconnect dielectric mate-
rials, the thermal stability of low-κ materials is of major importance since
they are submitted to thermal annealing during subsequent integrated circuit
manufacturing steps. Among other requirements, such as gap-fill capability,
adhesion and minimal moisture absorption, thermal stability up to 400◦ C is
necessary. However, despite some claims that plasma-deposited carbon flu-
orinated films are thermally stable at those temperatures, the usefulness of

a-C:H:F films have yet to be proved by integration in an ULSI structure.
The effects of the self-bias voltage are illustrated in Fig. 9, where the film
density of a-C:H, a-C:H:N and a-C:H:F films are presented as functions of
V b . Our results for low-V b depositions correspond to low-density films. Also,
these films present high luminescence Raman background, while IR results
show important contents of CHn groups. These features are characteristic of
a polymer-like structure, with high concentration of saturated sp3 carbons.
As the self-bias is increased, the film properties change. Within the self-bias
range from −100 V to −300 V, the most remarkable structural changes occur.
They can be identified by the densification of the structure as a result of the
higher generation rate of nucleation sites on the film surface due to the more
energetic ion bombardment. These films have a diamond-like character, with
high hardness and high internal stress. Films deposited at higher V b have a
slightly lower density.
4 Mechanical and Nanotribological Properties

The study on the nanometer scale of the physical phenomena related to the
interaction of surfaces in contact and in relative motion, nanotribology, was
made possible with the invention of the atomic force microscope in 1986 [51].
These phenomena are extremely important in situations like manipulation
of atoms and molecules in surfaces, as well as in the operation of electrome-
chanical devices. In the last decade, with the development of friction force
microscopy, the study of wear and friction properties of materials has received
increasing attention [52]. In spite of that, the study of the lubricant and tri-
bological properties of carbon-based films and other carbon-nanostructured
materials is still in the beginning [53]. The friction and wear measurements
at nanometer scale that are discussed in this section were carried out us-
ing an atomic force microscope (AFM) operated in the lateral force mode.
During the experiments the temperature and the relative humidity were kept
constant at 23◦ C and 40%, respectively. The cantilever normal force was ob-
tained by multiplying the measured cantilever bending by its normal bend-
ing constant. In order to obtain the absolute values of the friction coeffi-
cients, the cantilever bending constant was calibrated using the formula of
Neumeister et al. [54], while the AFM was calibrated with the method pro-
posed by Liu et al. [55].
In this section we present both the internal stress and film hardness re-
sults. The internal stress determination was made by measuring the curvature
of the substrate by means of stylus profilometry and by applying Stoney’s
equation, as described in detail elsewhere [56]. The hardness of the films was
measured employing a nanoindenter. The film hardness was obtained accord-
ing to the Oliver and Pharr method [57].
Fig. 10. (a) Compressive internal stress and (b) hardness as functions of the
incorporated nitrogen
The discussion in the previous section shows that a-C:H:N films with up
to 11 at. % of nitrogen retained their diamond-like properties, despite the pro-
gressive graphitization induced by nitrogen incorporation. The compressive
internal stress and hardness are presented in Fig. 10 for a-C:H:N films. It is
well known that the high internal compressive stress observed in a-C:H films
may be viewed as a result of the material overconstraining. This means that
the relatively high carbon sp3 fraction observed in hard a-C:H films causes
the mean atomic coordination number to be higher than the ideal value pre-
dicted for a fully constrained network [58]. In this scheme, any stress relief
process may be strongly coupled to a reduction of the coordination number.
In the case of a-C:H:N films, the important decrease in the internal compres-
sive stress upon nitrogen incorporation (Fig. 10a) is conceived as a combina-
tion of the chemical composition, since nitrogen atoms admit a coordination
number equal to 3 at most, and hybridization states. Besides the nitrogen
incorporation itself, which occurs at the expense of the carbon content, the
sharp increase in the fraction of carbon atoms in sp2 -hybridization state is
a source of decrease in the mean coordination number. Thus, for 9 at. % of
nitrogen content in a-C:H:N films, the hardness shows only a slight reduction
(Fig. 10b), while the internal stress is reduced by a factor of two [14].
The film hardness and the internal stress of a-C:H:F films were shown
in Fig. 11. Fluorine-poor films present high hardness and high compressive
internal stress, which is typical of diamond-like materials. As the amount of
incorporated fluorine increases, the magnitudes of these quantities decrease.
Films with fluorine content of 22 at. % are nearly stress-free and their hard-
ness was reduced by a factor of 3 when compared to a-C:H films. This hard-
ness is still more than a factor of 10 higher than the values obtained for
PTFE. Further incorporation of fluorine leads to soft films with slightly ten-
Fig. 11. (a) Compressive internal stress and (b) hardness as functions of the in-
corporated fluorine
Fig. 12. Contact angle as a function of the fluorine content of the films
sile stress. The reduction of the internal stress in a-C:H:F can be explained
by the more relaxed and less dense polymeric structure of these films [23].
The hydrophobicity of these films was investigated by contact angle mea-
surements. The results indicate an increase in the contact angle as the fluorine
content of the films increases (Fig. 12). The increase in the contact angle in
fluorinated amorphous carbon materials was attributed not only to a higher
incorporation of fluorine but mostly to the presence of CFn groups on the
surface [59]. In fact, the fluorine richest films have a F/C ratio of only 0.5
Fig. 13. Friction coefficient as a function of the contact angle
but present contact angles of 95◦ . On the other hand, PTFE has a F/C ratio
of 2 and presents contact angles in the range 104–111◦ [25].
From the results presented in Fig. 13, the importance of the surface wet-
tability on the friction mechanism is clear. In this figure, the friction co-
efficients were plotted as function of the contact angle and show a nearly
linear decrease. This result is in good agreement with previous observation
of the influence of the surface hydrophobicity on the friction forces for a-C:H
films [60].
The importance of the nitrogen incorporation to surface hydrophobicity
and friction processes was also verified when the nitrogen incorporation into
fluorinated carbon films was studied [27]. In Fig. 14, the friction coefficients
and the contact angles were plotted as functions of the nitrogen content in the
films. It is clear from the figure that the friction coefficient and the contact
angle have an opposite tendency.
The observed correlation between the contact angle and the friction co-
efficient reinforce the importance of the capillary condensation kinetics in
the friction process that occurs at nanometer scale: the greater the surface
hydrophobicity, the less important is the water vapor condensation for the
nanoscale friction process. In a recent publication [61], it was suggested that
the sliding friction forces are determined by two competitive processes. De-
pending on the surface wettability, one of the processes can dominate. The
first one is the thermally activated stick and slip behavior, and friction is
due to the cohesive forces between the two surfaces in contact. The second
contribution comes from the formation of capillaries at the nanoasperities
contact points between tip and film surface due to water vapor condensa-
tion, which must be taken into consideration especially for measurements
performed in air. In fact, the role of the water condensation was investigated
80
)
o
Contact angle (
72
64
56
0.24
Friction coefficient
0.20
0.16
0.12
0 5 10 15 20
Nitrogen content (at.%)
Fig. 14. Contact angle (upper part) and friction coefficient (lower part) as functions
of the nitrogen content in the films. The a-C:H:F:N films are represented by black
dots, while the open dots correspond to values measured in a-C:H films deposited
in pure methane atmosphere
from the velocity dependence of the friction force, and the results revealed
the importance of the surface energy [25]. In this case, the kinetics of cap-
illary condensation of water vapor in the contact area causes a decrease of
friction with increasing velocity, since for higher scanning velocities there is
no time for capillary formation [61]. In the case of hydrophilic surfaces, this
mechanism can predominate.
Nano- and microscale wear behavior of a-C:H, a-C:H:N and a-C:H:F films
deposited by PECVD have been studied with the use of atomic force mi-
croscopy and the sliding sphere test [62]. At the nanometer scale, the wear
depth is observed to increase linearly with the increase of the tip surface nor-
mal force. An analysis of the minimum force necessary to scratch the surface,
obtained by extrapolating the wear versus load force curve, as well as the
wear rates and the scratch depth test results, show that the wear is strongly
related to the film hardness and microstructure. This fact has been observed
either in a very low force regime, in the range of µN using the AFM, or with
Fig. 15. Wear rate and critical force to scratch the films as functions of the
hardness. For sake of comparison, the wear rate obtained for the Si substrate is
20 × 103 mm3 /Nm. The lines are only eye-guides
forces in the range of mN with a sphere abrasion equipment, since a similar

trend is observed in both scales. Even though the range of forces and the
wear process in both tests are quite different, we are led to the conclusion
that the hardness of the films plays the dominant role in determining their
wear performance in both nano- and microscales [62]. From Fig. 15, it is seen
that the wear rate of the polymeric a-C:H:F film is higher than the values
obtained for the a-C:H film. The results show that when the film hardness
decreases, an increase in the wear rate, as determined by a sphere test, is
observed. At the micrometer scale, the hardness of the films appears to be
the dominant factor in determining the wear resistance. In this figure we also
plot the critical force to scratch the film surface using the AFM tip as a func-
tion of the film hardness. In this last set of data, the hardness also appears
as a good parameter to present the data.
5 Conclusion
We found that a-C:N:H films with up to 11 at. % of N retained their diamond-

like properties, despite the progressive graphitization induced by nitrogen in-
corporation. In this case, the reduction of the internal stress without substan-
tial changes in the film hardness was attributed to the reduction of network
connectivity. On the other hand, the increase of the fluorine content results
in polymer-like films, with smaller internal stress and hardness. A structural
transition from diamond-like to polymer-like films as a result of the incorpo-

ration of fluorine could be inferred. As a general rule, films deposited with
low values of self-bias voltage result in polymeric materials. In which concerns
the fiction properties, surface hydrophobicity plays an important role in the
friction properties. One can control the friction behavior of a-C:H films and
their alloys by changing the surface energy via the incorporation of fluorine
(increases the surface hydrophobicity) or nitrogen (reduces the surface hy-
drophobicity). In this way, the friction coefficient and the contact angle have
an opposite tendency. On the other hand, the wear rate is nearly independent
of the chemical composition of the film, but is directly proportional to the
film hardness.
Acknowledgements
The author is very grateful for the fruitful discussions with D.F. Frances-
chini, G. Mariotto, R. Prioli, L.G. Jacobsohn and M.E.H. Maia da Costa.
This work is partially supported by the Brazilian agencies: Fundação de Am-
paro à Pesquisa do Estado do Rio de Janeiro (FAPERJ), Conselho Nacional
de Desenvolvimento Cientı́fico e Tecnológico (CNPq) and Coordenação de
Aperfeiçoamento de Pessoal de Nı́vel Superior (CAPES).
References
[1] J. Robertson: Mater. Sci. Eng. R 37, 129 (2002) 217, 220
[2] Y. Lifshitz, S. R. Kasi, J. W. Rabalais: Phys. Rev. Lett. 62, 1290 (1989) 217
[3] J. Robertson: Diamond Relat. Mater. 14, 942 (2005) 217
[4] A. Grill: Diamond Relat. Mater. 12, 166 (2003) 218
[5] C. Casiraghi, A. C. Ferrari, J. Robertson, R. Ohr, M. V. Gradowski, D. Schnei-
der, H. Hikgers: Diamond Relat. Mater. 13, 1480 (2004) 218
[6] Y. Jun, J. Y. Choi, K. R. Lee, B. K. Jeong, S. K. Kwon, C. H. Hwang: Thin
Solid Films 233, 377–378 (2000) 218
[7] R. G. Lacerda, F. C. Marques: Appl. Phys. Lett. 73, 619 (1998) 218
[8] S. S. Camargo, A. L. B. Neto, R. A. Santos, F. L. Freire, Jr., R. Carius,
F. Finger: Diamond Relat. Mater. 7, 1155 (1998) 218
[9] G. A. J. Amaratunga, M. Chhowalla, C. J. Kiely, I. Alexandou, R. Aharonov,
R. M. Devenish: Nature 383, 321 (1996) 218
[10] S. Muhl, J. M. Mendez: Diamond Relat. Mater. 8, 1809 (1999) 218
[11] A. Y. Liu, M. L. Cohen: Science 245, 841 (1989) 218
[12] M.-L. Wu, M. U. Guruz, W. A. Chiou, V. P. Dravid, Y. W. Chung, S. Anders,
F. L. Freire, Jr., G. Mariotto: Appl. Phys. Lett. 76, 2692 (2000) 218
[13] F. L. Freire, Jr.: Jpn. J. Appl. Phys. 36, 4886 (1997) 218, 226
[14] D. F. Franceschini, C. A. Achete, F. L. Freire, Jr.: Appl. Phys. Lett. 60, 3229
(1992) 218, 230
[15] S. R. P. Silva, J. Robertson, G. A. J. Amaratunga, B. Raferty, L. M. Brown,
J. Schwan, D. F. Franceschini, G. Mariotto: J. Appl. Phys. 81, 2626 (1997)
218
[16] D. F. Franceschini, F. L. Freire, Jr., S. R. P. Silva: Appl. Phys. Lett. 68, 2645
(1996) 218, 224
[17] S. Rodil, N. A. Morrison, W. I. Milne, J. Robertson, V. Stolojan,
D. N. Jayawardane: Diamond Relat. Mater. 9, 524 (2000) 219
[18] A. Grill: Diamond Relat. Mater. 10, 234 (2001) 219
[19] J. A. Theil: J. Vac. Sci. Technol. B 17, 2397 (1999) 219
[20] A. Grill, V. Patel, C. Jahnes: J. Electrochem Soc. 145, 1649 (1998) 219
[21] R. d’Agostino, R. Lamendola, P. Favia, A. Giquel: J. Vac. Sci. Technol. A 12,
308 (1994) 219, 224, 227
[22] K. Endo, T. Tatsumi: J. Appl. Phys. 78, 1370 (1995) 219
[23] L. G. Jacobsohn, D. F. Franceschini, M. E. H. M. da Costa, F. L. F. Jr.: J.
Vac. Sci. Technol. A 18, 2230 (2000) 219, 220, 227, 231
[24] L. G. Jacobsohn, M. E. H. M. da Costa, V. J. Trava-Airoldi, F. L. Freire, Jr.:
[25] R. Prioli, L. G. Jacobsohn, M. E. H. M. da Costa, F. L. Freire, Jr.: Tribol.
Lett. 15, 177 (2003) 219, 232, 233
[26] F. L. Freire, Jr., M. E. H. M. da Costa, L. G. Jacobsohn, D. F. Franceschini:
[27] C. M. T. Sanchez, M. E. H. M. da Costa, R. R. M. Zamora, R. Prioli,
F. L. Freire, Jr.: Diamond Relat. Mater. 13, 1366 (2004) 219, 232
[28] L. Valentini, J. M. Kenny, R. M. Montereali, L. Lozzi, R. Santucci: J. Vac. Sci.
Technol. A 20, 1210 (2002) 219
[29] H. Yokomichi, A. Masuda: J. Non-Cryst. Solids 271, 147 (2000) 219
[30] M. E. H. M. da Costa, C. M. T. Sanchez, L. G. Jacobsohn, F. L. Freire, Jr.:
Thin Solid Films 482, 109 (2005) 219
[31] F. L. Freire, Jr., D. F. Franceschini, C. A. Achete: Nucl. Instrum. Meth. B 85,
268 (1994) 220, 224
[32] K. J. Clay, S. P. Speakman, G. A. J. Amaratunga, S. R. P. Silva: J. Appl.
Phys. 79, 7227 (1996) 221
[33] P. Hammer, W. Gissler: Diamond Relat. Mater. 5, 1152 (1996) 222
[34] F. D. A. Reis, D. F. Franceschini: Phys. Rev. E 61, 3417 (2000) 222
[35] W. Jacob: Thin Solid Films 326, 1 (1998) 222
[36] A. von Keudell, T. Schwarz-Selinger, M. Meier, W. Jacob: Appl. Phys. Lett.
76, 676 (2000) 222
[37] W. Schwarzenbach, G. Cunge, J. P. Booth: J. Appl. Phys. 85, 7562 (1999)
222
[38] M. E. H. M. da Costa, F. L. Freire, Jr., L. G. Jacobsohn, D. F. Franceschini,
G. Mariotto, I. J. R. Baumvol: Diamond Relat. Mater. 10, 910 (2001) 224,
227
[39] L. G. Jacobsohn, S. S. Camargo, F. L. Freire, Jr.: Diamond Relat. Mater. 11,
1831 (2002) 224, 228
[40] D. S. Tuinstra, J. L. Koenig: J. Chem. Phys. 53, 1126 (1970) 225
[41] R. Kalish, Y. Lifshitz, K. Nugent, S. Prawer: Appl. Phys. Lett. 74, 2936 (1999)
225
[42] R. O. Dillon, J. A. Woollam, V. Katkanant: Phys. Rev. B 74, 2936 (1999) 225
[43] A. C. Ferrari, S. E. Rodil, J. R. Robertson: Phys. Rev. B 67, 155306 (2003)
225
[44] G. Mariotto, C. A. Achete, F. L. Freire, Jr.: Thin Solid Films 241, 255 (1994)
226
[45] S. R. Salis, D. J. Gardiner, M. Bowden, J. Savage, D. Roway: Diamond Relat.
Mater. 5, 589 (1996) 226
[46] T. E. Karis, G. W. Tyndall, D. Fenzel-Alexander, M. S. Crowder: J. Vac. Sci.
Technol. A 15, 2382 (1997) 227
[47] P. Hammer, W. Gissler: Diamond Relat. Mater. 5, 1152 (1996) 227
[48] S. E. Rodil, W. Beyer, J. Robertson, W. I. Milne: Diamond Relat. Mater. 12,
921 (2003) 228
[49] F. L. Freire, Jr., C. A. Achete, G. Mariotto, R. Canteri: J. Vac. Sci. Technol.
A 12, 3048 (1994) 228
[50] U. Müller, R. Hauert, B. Oral, M. Tobler: Surf. Coat. Technol. 367, 76–77
(1995) 228
[51] G. Binning, S. F. Quate, C. Gerber: Phys. Rev. Lett. 56, 930 (1986) 229
[52] E. Meyer, R. M. Overney, K. Dransfeld, T. Gyalog: Nanoscience: Friction and
Rheology on the Nanometer Scale (World Scientific, Singapore 1996) 229
[53] F. L. Freire, Jr., R. Prioli: Nanotribology of carbon films, in H. Nalwa
(Ed.): Encyclopedia of Nanoscience and Nanotechnology (American Scientific,
Stevenson Ranch, California 2004) pp. 899–917 229
[54] J. M. Neumeister, W. A. Ducker: Rev. Sci. Instrum. 65, 2527 (1994) 229
[55] E. Liu, B. Blanpain, J. P. Celis, J. R. Roos: J. Appl. Phys. 84, 4859 (1998)
229
[56] M. M. Lacerda, F. L. Freire, Jr., R. Prioli, C. M. Lepinski, G. Mariotto: J.
Vac. Sci. Technol. A 17, 2811 (1999) 229
[57] W. C. Oliver, G. M. Pharr: J. Mater. Res. 7, 1564 (1992) 229
[58] J. C. Angus, F. Jansen: J. Vac. Sci. Technol. A 6, 1778 (1988) 230
[59] R. D. van de Grampel, W. Ming, A. Gildenpfennig, J. Laven,
H. H. Brongersma, G. de With, R. van der Linde: Langmuir 20, 145 (2004)
231
[60] S. S. Perry, C. M. Mate, R. L. White, G. A. Somorjai: IEEE Trans. Mag. 32,
115 (1996) 232
[61] E. Riedo, F. Lévy, H. Brune: Phys. Rev. Lett. 88, 185505–1 232, 233
[62] P. Ayala, M. E. H. M. da Costa, R. Prioli, F. L. Freire: Surf. Coat. Technol.
182, 335 (2004) 233, 234
Index
a-C:H, 218–220, 222–224, 227–230, atomic force microscopy (AFM), 229,
232–234 233, 234
a-C:H:F, 219–224, 226, 228–234
a-C:H:F:N, 219, 233 carbon coordination, 230
a-C:H:N, 218, 220–224, 226, 228–230, coatings
233 protective coatings, 218
sp2 -bonded clusters, 217
D band, 224, 225
cluster size, 226 D/G band intensity ratio, 225, 226
density, 224, 227–229
annealing, 228 diamond, 217
diamond-like carbon (DLC), 218, 224, nanotribology, 229

228–230, 234 nanotubes, 218
diamondlike carbon (DLC), 217 nitrogen incorporation, 217–219, 221,
224, 226, 228, 230, 232, 234
electrical properties, 218, 219
electron energy loss spectroscopy optical properties, 218
(EELS), 224, 226
fluorine incorporation, 217–219, 223, deposition (PECVD), 218, 219,
224, 226–228, 230, 231 228, 233
friction coefficient, 219, 229, 232, 233 polymer-like carbon (PLC), 224,
fullerene, 218 227–229, 234
polymers, 217
G band, 224, 225
graphene, 218 Raman spectroscopy, 224–226, 229
graphite, 217
silicon incorporation, 218
graphite-like carbon (GLC), 218
subplantation model, 217, 218
graphitization, 230, 234
thermal stability, 218, 219, 228
hardness, 218, 229–231, 233, 234 thermal-induced gas effusion, 228
hydrogen incorporation, 218 tribological properties, 217, 219, 229
hydrophobicity, 219, 231, 232, 235
ultra-large scale integrated circuits
infrared (IR) spectroscopy, 224, 227, (ULSI), 219, 229
229
internal stress, 217, 218, 229–231, 234 wear resistance, 229, 233, 234
ion beam analysis (IBA), 220, 224 wettability, 219, 232
mechanical properties, 217–219, 229 x-ray photoemission spectroscopy

microstructure, 217, 219, 224, 228, 233 (XPS), 224, 227
Ion Microscopy on Diamond
Claudio Manfredotti
Dipartimento di Fisica Sperimentale and Centro d’Eccellenza “Nanostructured

Interfaces and Surfaces” (NIS), Università di Torino, Via P. Giuria 1, I-10125
Torino, Italy
manfredotti@to.infn.it
Abstract. Because of its physical properties (strong radiation hardness, wide en-
ergy gap with a consequent extremely low dark current, very large electron and
hole mobility) diamond is a very good candidate for nuclear particle detection, par-
ticularly in harsh environments or in conditions of strong radiation damage. Being
commonly polycrystalline, diamond samples obtained by chemical vapour deposi-
tion (CVD) are not homogeneous, not only from the morphological point of view,
but also from the electronic one. As a consequence, as it was indicated quite early
starting from 1995, charge collection properties such as charge collection efficiency
(cce) are not uniform, but they are depending on the site hit by incoming particle.
Moreover, these properties are influenced by previous irradiations which are used in
order to improve them and, finally, they are also dependent on the thickness of the
sample, since the electronic non uniformity extends also in depth by affecting the
profile of the electrical field from top to bottom electrode of the nuclear detector
in the standard “sandwich” arrangement. By the use of focussed ion beams, it is
possible to investigate these non uniformities by the aid of techniques like IBIC
(Ion Beam Induced Charge) and IBIL (Ion Beam Induced Luminescence) with a
space resolution of the order of 1 µm. This relatively new kind of microscopy, which
is called “ion microscopy”, is capable not only to give 2D maps of cce, which can be
quite precisely compared with morphological images obtained by Scanning Electron
Microscopy (generally the grains display a much better cce than intergrain regions),
but also to give the electric field profile from one electrode to the other one in a
“lateral” arrangement of the ion beam. IBIL, by supplying 2D maps of luminescence
intensity at different wavelength, can give information about the presence of spe-
cific radiative recombination centers and their distribution in the material. Finally,
a new technique called XBIC (X-ray Beam Induced Charge), which makes use of
very collimated (to 0.1 µm) x-ray beams from high energy electron synchrotrons,
opens new ways to map cce with a less damaging radiation and with a better en-
ergy resolution. In this paper we resume recent and less recent work carried out by
our group by using these techniques, a work that has been undertaken afterwards
also by other research groups in the world. In particular, topics such as the better
homogeneity obtained by “priming” and the effects of “light priming”, together
with a certain “complementarity” between IBIC and IBIL maps, giving evidence
that radiative recombination centers along the grain boundaries or in damaged re-
gions are important in affecting cce, will be presented and discussed in some details.
The conclusion is that ion microscopy is a powerful and essentially unique method
for the investigation of diamond and other semiconductor materials proposed for
nuclear detection.
240 Claudio Manfredotti
1 Introduction
Ion microbeams [1] represent a powerful tool for investigating the physical
and chemical properties of solids. Even if the spatial resolution is still lower
than for electron microscopy (SEM, TEM), the types of measurement with
ions (protons, helium) are wider and their accuracy is much better. In effect,
ion microbeams in the energy range between 2 and 10 MeV or more allow
for: (1) trace elements analysis and chemical composition measurements down
to ppma range [2] and below (particle-induced X-ray emission, PIXE), due
to the absence of bremsstrahlung radiation background; (2) measurement of
transport properties [3] of semiconductors (charge collection distance or drift
length, mobility-trapping time product) and for the acquisition of the rele-
vant maps over regions as wide as 2 mm or more (ion beam-induced charge,
IBIC); (3) measurement and mapping of luminescence properties [4] down to
a thickness of 100 µm or more, with respect to 1 µm depth reached by CL (ion
beam-induced luminescence, IBIL); (4) measurement of doping levels down
to 1 ppma with a depth resolution of 1 nm (Rutherford, RBS backscattering)
or light elements distribution profiles with the same resolution (elastic recoil
diffusion, ERD). The precise definition of the range, the very low Coulomb
scattering and lateral straggling keep the spatial resolution at the same level
as the ion spot diameter, while the possibility of getting a trigger signal
allows for the realization of time-resolved measurements, which give infor-
mation about time behaviour of the measured parameter, such as radiative
recombination lifetime with IBIL. The number of charges released by ions is
so large to allow for the possibility of getting a map just in one pass over
the selected area without getting problems with radiation damage, which,
of course, is heavier with ions than with electrons. Even if this technique is
much younger than electron microscopy, there are now on the market ap-
paratuses which are technologically advanced and which are to some extent
as user-friendly as SEM or TEM. In this paper we shall concentrate only
on IBIC and IBIL measurements on diamond, both CVD and natural, be-
cause diamond is a very interesting material for the realization of nuclear and
other kinds of detectors (very large resistivity, complete depletion over thick
regions at low voltages, extremely high resistance to radiation damage). Still,
diamond has the problem of being mainly polycrystalline and, as it will be
shown, this structural property is reflected in material nonhomogeneity with
respect to radiation response or charge collection efficiency (CCE). Change
collection efficiency is defined as the fraction of generated charges which are
contributing to the current pulse released by the nuclear particle or by some
ionising radiation crossing the sample. IBIC [5, 6, 7, 8] represents, in fact, the
only simple way to map CCE over suitable region sizes, while with IBIL it
is possible to select the regions which limit CCE by giving rise to a strong
radiative recombination, which characterizes by its blue light emission the
luminescence properties of the purest, detector grade, diamond.
Ion Microscopy on Diamond 241
Fig. 1. The proton microbeam set up at AN-2000 electrostatic accelerator in Leg-

naro National Laboratories (LNL) of National Institute for Nuclear Physics (Italy)
2 IBIC
There are more than 50 ion microbeam apparatuses around the world and
many of them are in Europe. One such apparatus, that at the National Labo-
ratories of Legnaro, Italy, National Institute for Nuclear Physics, is illustrated
in Fig. 1, which shows the last part of AN-2000 electrostatic accelerator [6]
with the focusing system (three quadrupoles) and the scattering chamber.
The spot diameter can be reduced down to about 1 µm, and the scanned
area can be as large as 2 mm × 2 mm, but is typically much lower (400 µm
× 400 µm). With a dwell time of few ms, the time taken for a full image
258 × 258 pixels is on the order of few minutes. Beam intensities are kept
very low, much less than a fA. By limiting the counting rate to 100 cps, for
instance, it is possible to get one event per pixel. At larger counting rates
local trapping may create space charge in the pixel and reduce the charge
collection efficiency.
Some details of the scattering chamber are shown in Fig. 2. Several ports
at the periphery can be used for signal outputs related to the collected
charge (IBIC), induced luminescence (IBIL, with monochromator and photo-
tube), charateristic X-rays revealed by a Si(Li) detector suitably cooled by a
liquid nitrogen dewar (PIXE), ions backscattered from the sample as revealed
by a Si surface barrier detector (RBS). Downstream a similar detector may
be used for scanning transmission ion microscopy (STIM) and other similar
detectors may be used for other techniques (a Faraday cup can be used for
detection of secondary electrons to obtain secondary electrons ion microscopy
Fig. 2. Details of the scattering chamber used in LNL for IBIC/IBIL measurements
(SEIM) images. A scintillator is used to detect beam position and to align

the samples under investigation.
An IBIC setup is schematically shown in Fig. 3. The acquisition is carried
out event-by-event in a file (x, y, collected charge) giving the possibility to
replay the events in order to follow some specific behaviour (polarization, for
instance). IBIC could be used both in standard frontal geometry and also in
“lateral” geometry [9, 10], on a cross section of the sample (Fig. 4). There
are a number of important advantages for lateral IBIC. First, carriers are
immediately swept apart by the electrical field and plasma recombination is
avoided. In addition, CCE can be profiled as a function of depth, giving in-
formation concerning different regions in inhomogeneous samples, like CVD
diamond. Third, some CCE loss under the electrodes can be suitably mon-
itored. Finally, profiles of the electrical field could be obtained under some
reasonable assumptions. The drawbacks of obtaining a cross section of the
sample, of getting in general large dark currents and some fringing field ef-
fects, are overcome in CVD diamond by cleaving the sample (dark currents
remain very low) and by using protons of a suitable energy in order to place
the Bragg’s peak at a depth large enough to be sure to have a real bulk
electric field.
In frontal IBIC [7,11] it is possible to obtain CCE maps with the maximum
contribution given by carriers generated at the depth of Bragg’s peak, which
ranges from 25 µm (2 MeV protons) up to more than 100 µm for 9 MeV
protons (shorter thicknesses are reached with few MeV helium ions). Frontal
IBIC may present some problems in thicker CVD diamond samples with drift
lengths of both carriers much longer than thickness. In effect, with respect
Fig. 3. Layout of a typical IBIC experiment
Fig. 4. Description of the geometry of irradiation and signal collection in a lateral

IBIC measurement. In frontal IBIC the microbeam hits one of the two electrodes
to the minimum ionising particles (mips), as may be defined for almost all
the particles coming out from high-energy experiments, ionisation density is
more than 100 times larger (0.7 MeV/mm for mips and 80 MeV/mm for
2 MeV protons, on average). Taking into account the lateral diffusion of
charge, created by δ-rays, assuming carrier lifetimes in diamond on the order
of 0.1 ns, the ionisation given by low energy protons can give rise to an
Fig. 5. A frontal IBIC map carried out on a CVD diamond sample biased at
+600 V (front electrode) together with the relevant multichannel pulse spectrum.
The proton energy is 2 MeV. The electrode diameter is 2 mm. Previous X-ray
priming of the sample amounted to 10 Gy
instantaneous carrier density on the order of 1013 –1015 cm−3 , which is much
larger than the equilibrium one (about 105 cm−3 ) in diamond. Mips create
much more energetic δ-rays, and local ionisation density along the track is
lower by some orders of magnitude. With respect to mips, it is expected that
frontal IBIC maps could be affected by polarization effects, even if priming
with X-rays is used in order to fill the traps and make the response of the
sample more homogeneous.
One example is given in Fig. 5, which shows a CCE map together with the
relevant multichannel spectrum obtained on a CVD diamond detector 600 µm
thick in which a charge collection distance (CCD) of 75 µm was measured by
using mips. CCD, which is defined as the average distance between electrons
and holes, as averaged over the thickness of the sample, is obtained simply
by multiplying the average CCE by the thickness. It has been proved by the
“linear model” that it increases linearly from the back to the growth face of
the sample. In our case, even if the average CCE is between 10 and 15%,
in agreement with data obtained with mips, the spatial distribution is very
inhomogeneous, with large regions of the sample in which, because of strong
polarization, no events were recorded. In fact, after 2 hours of measurement,
the event rate dropped down to zero, while it started at more than 100 cps.
The homogeneity of the response [12] is quite important in tracking appli-
cations, since the precise event position is reconstructed as a center of mass
Fig. 6. A frontal IBIC map carried out on the same CVD diamond sample biased
at −600 V (front electrode) and under illumination, together with the relevant
multichannel pulse spectrum. The scan region is the same as in Fig. 5
by means of the collected charge in nearby strips or pixels and, if CCE is not
uniform, the position is badly reconstructed. In IBIC maps there are small
regions with CCE as large as 30–35%, which should correspond to CCD val-
ues of about 200 µm, the best it is possible to get in similar samples. As a
matter of fact, these regions are corresponding to almost perfect diamond
monocrystals, as evidenced in maps obtained on thin samples by X-rays.
By illuminating the sample with blue light (400 nm) or with white light
and by switching the polarity to negative bias (Fig. 6) the situation changes
drastically: high-efficiency spots are still there, but now the large “blind” re-
gion among these spots has got CCE values of about 10%, i.e., CCD values of
60 µm, close to those obtained by mips. If these regions are disordered regions
with grain boundaries and dislocations, with a lot of filled dangling bonds
and a distributed space charge with very low electric fields, now this opti-
cally induced detrapping eliminates, at least partially, this space charge and
the relevant polarization and thereby improves CCE. Since these data were
obtained by 2 MeV protons (25 µm penetration depth) and with a negative
bias voltage applied at the front electrode, we can immediately attribute this
beneficial effect to electrons, which are giving the main contribution to the
charge signal according to Ramo’s theorem. In fact, this theorem states that
a charge contributes with the fraction of the covered path with respect to the
extension of the electrical field (sample thickness in this case, as evidenced
by lateral IBIC and discussed in the next section).
Fig. 7. Lateral IBIC CCE profile on a cross section of a CVD diamond sample
200 µm thick. Bias voltage (+400 V) is applied at the growth side (left). The
profile was averaged over a region 450 mm wide. Black squares are the experimental
data, the continuous line is the fit, empty and filled circles are holes and electrons
contributions, respectively
Fig. 8. Lateral IBIC CCE profile on a cross section of the same CVD diamond
sample 200 µm thick. Bias voltage (−400 V) is applied at the growth side (left).
For symbols see Fig. 7
3 Lateral IBIC
Lateral IBIC [9,10,13,14,15] has been used in order to test the “linear model”
and to throw some light on the electric field depth distribution, which is gen-
erally assumed to be uniform. Moreover, from CCE depth profiles obtained
now without polarization effects and therefore in conditions not different

from the mips case, it is possible to evaluate CCD from experimental data
and compare it with data obtained with mips. In order to do this, charge
collection distance or charge collection length should be better defined as
drift length µτ E, i.e., the product of (mobility)×(trapping time)×(electric
field), which is a local value and can be diversified for the two carriers. Then
the average value of the sum of µτ E for electrons and holes over the sam-
ple thickness, taking into account that ionisation for mips is uniform, can
be compared with CCD. This is what will be done in what follows. We can
start from CCE distributions (Figs. 7 and 8) for positive and negative bias
voltages, respectively (the voltage is applied on the left, which is also the
growth side) in the case of a sample 200 µm thick. It can be observed that
these distributions do depend on the bias voltage polarity, essentially because
mobility lifetimes are different for the two carrier types and also because the
electric field distribution is different due to the electric contacts. In effect, by
looking at both curves, one observes that peak CCE is closer to the negative
contact. This means, by Ramo’s theorem, that the main contribution in this
case is given by electrons, which have at their disposal a much longer path
towards the anode. It is possible to express CCE by Ramo’s theorem as a
double path integral for each carrier as a function of hit point of the beam and
to obtain CCE distribution in terms of µτ E depth distribution separately for
each carrier. In order to reduce the number of parameters, one can make a
more physical “linear model” assumption in terms of a linear dependence of
µτ on depth, increasing from the back (substrate) side, a constant ratio of
µτ between the two carriers and a suitable parametrization of the electrical
field with as few parameters as possible [3]. Apart from electrical field, all
the other parameters are kept constant for all the bias voltages investigated
(seven in total).
From Figs. 7 and 8 it can be deduced that effectively the largest con-
tribution to CCE (about ten times) is given by electrons, which display an
almost uniform value of µτ of about 5 × 10−7 cm2 V−1 , with a resulting fit
to experimental data that can be considered quite good (not only in this,
but also in all the other cases). The electric field distribution is, however,
different, with a peak at the positive front side and a much more penetrating
behaviour in the case of a negative front side.
The effect of the electric field is that of giving a similar behaviour to
drift lengths of both carriers (Fig. 9), with an average value of 100 µm for
electrons at −400 V and with a decay at both electrodes. The final result is
to get a combined drift or mean collection length, averaged over the sample
thickness and defined here as Le + Lh , similar to CCD and symmetric with
respect to the bias voltage (Fig. 10). The same behaviour resulted from mips
data, with a quite close value of 50 µm for CCD at 200 V, i.e., at an ohmic
electrical field of 1 V/µm and of 100 µm at 2 V/µm. Apart from the details
of the deconvolution of CCE in terms of different contributions of electrons
Fig. 9. Electron and hole drift lengths (L) as a function of depth derived from
previous figure, together with the profile of the electric field
Fig. 10. Behaviour of mean collection length L for electrons (filled circles), holes
(open circles) and both carriers (filled squares) as a function of bias voltage obtained
by simulating the behaviour of a mip
and holes, this result is given directly from figures like Figs. 7 and 8, and it
proves the validity of lateral IBIC in deriving CCE depth distributions.
In the same way, it is possible to observe the effect of priming or pumping
(see [16] at p. 93), i.e., with irradiations of the sample with X-rays or high-
energy electrons, in order to stabilize the performances of the detector. After
a priming dose of 46 Gy (Fig. 11), the CCE profile has a higher and broader
peak and it is much improved in the right part at lower values of CCE.
This is an average over a cross section region of 450 µm. By considering now
the single paths connecting the two electrodes (called here “columns”), one
Fig. 11. Lateral IBIC CCE profiles on a cross section of the same CVD diamond
sample 200 µm thick, as obtained before (top) and after (bottom) a priming dose
of 46 Gy. Bias voltage (−300 V) is applied at the growth side (left)
obtains distributions of CCE like those reported in Fig. 12, which measures
the average increase of CCE or CCD given by priming.
Since peak efficiency increases from 0.153 to 0.283 after priming, the im-
provement amounts to a factor 1.8, which is exactly the same as that obtained
with mips [16]. What can be observed in more detail is that this improvement
is really “local”, as clearly shown in Fig. 13, which refers to the detailed ef-
ficiency profile obtained along the border of the top (or front) electrode and
derived from the averages along the “columns”, as it was for Fig. 12.
The exact comparison with the results concerning uniformity of CCE
obtained by mips, which have been here simulated by starting from CCE
profiles across the sample, will be reported in what follows. It is clear that
in our case we got uniformity along a line, while with mips the uniformity
refers to an area, which is, of course, impossible to obtain by our approach.
However, it has to be noted that the nonuniformity, obtained as a standard
deviation from fits reported in Fig. 12, is decreased by priming from 32.7% to
Fig. 12. Distributions of CCE as obtained from the same data as Fig. 11 and
averaged over the different straight lines connecting the two electrodes (called
“columns”): these distributions simulate CCE as viewed from mips with hit points
along a straight line (the border of top electrode). The fit has been obtained with
two Gaussian curves for the nonirradiated or unprimed case (mean 0.153, FWHM
0.059 for the first peak, mean 0.283 and FWHM 0.035 for the second one) and one
Gaussian line for the primed case (mean 0.292, FWHM 0.077)
Fig. 13. Profile of CCE as viewed by mips along a segment of top electrode, for
the primed and unprimed case
22.2%. This result confirms those reported more qualitatively in Figs. 5 and 6:
the improvement is due to the strong increase of the number of pixels showing
CCE values around 10–15%, but also to the analogous increase around 30%
of CCE. It is confirmed that, as observed in Figs. 5 and 6, those regions were
already present in a smaller number in the nonirradiated sample.
Fig. 14. Layout of IBIL apparatus, with ellipsoidal mirror, the monochromator
and the phototube
4 IBIL
With the same proton microbeam, it is also possible to carry out IBIL mea-
surements [3, 4, 17, 18, 19] and to map the luminescence efficiency produced
by the beam itself. When, as occurs in diamond, the radiative recombinations
are important and at least comparable with the nonradiative ones, IBIL maps
are indicative of regions with lower CCE. As a consequence, by the examina-
tion of the relevant wavelength spectra, it is possible to infer which are the
recombination centers responsible for the low values of CCE.
A IBIL schematic setup is given in Fig. 14. The microbeam enters a
small hole (diameter of 1 mm) of an ellipsoidal metal mirror, which allows
for the beam scanning, and the light is collected by a vacuum light guide
without dispersions or absorptions (two photos of mirror and monochromator
in Fig. 15).
With a 0.25 m focal length monochromator it is possible to reach a good
wavelength resolution in the interval 300–900 nm, which is also the region of
the maximum quantum efficiency of the phototube. Two retractable mirrors
give the possibility of also carrying out panchromatic images. With respect
to standard cathodoluminescence (CL), the electronic chain works in pulse
conditions: the signal from the phototube is amplified by a gated amplifier
(gate of 70 ms at each microbeam position), and the pulses are collected and
simply counted (no pulse height analysis is carried out) until the beam moves
to the next position. Taking into account the microbeam intensities, which
are here much higher than for IBIC (about 700 pA), the luminescence effi-
ciency of the sample, the solid angle viewed by the mirror, its reflectivity, the
optical band pass, etc., we are led to the conclusion that we are just counting
the photons, i.e., one event per pixel and no pile-up problems. As a conse-
quence, the maps are intended to be maps in photon number, instead of CCE
or electron number as in IBIC. In any case, the map signals are not to be
quantitatively related by a direct proportionality to the luminescence inten-
Fig. 15. Photographs of parts of IBIL apparatus
sity coming from the relevant pixels. The advantages with respect to CL are
that the luminescence is a real bulk luminescence, i.e., it comes mainly from
a defined depth due to the Bragg’s peak and time-resolved measurements can
be carried out by this approach.
An example of the complementarity between IBIC and IBIL maps is given
in Fig. 16. The maps were carried out on the same sample cross section (and
therefore are to be quoted as lateral). As expected, IBIC CCE is higher in the
upper part of the map, which is on the growth side and contains larger and
better quality grains, while IBIL is higher in the bottom region, in which the
grains are smaller and the defective regions are wider. Of course, there are
exceptions, such as the big luminescent region at the center-left extending
towards the upper contact, as there are regions completely white, in which
Fig. 16. Top: Lateral IBIC maps on a cross section of a CVD diamond sample
600 mm thick. The growth side is on the top, the region is about 400 mm wide.
Proton energy 2 MeV. Bottom: Lateral IBIL map on the same cross section. The
substrate side is more visible at the bottom. Proton energy 2 MeV
Fig. 17. Frontal IBIL spectra on a CVD diamond sample of good electronic quality.
The spectra are decreasing in intensity at increasing proton doses. Proton energy
2 MeV
both CCE and luminescence are low or absent. It could also be that the two
maps, which were carried out without moving the sample, the scan size and
the beam position (but, of course, focusing or collimation had to be changed
because of higher intensity needed by IBIL) are not exactly coincident and
that the complementarity could be better than what “rigidly” observed.
It is suggested that IBIL maps are indicative of the “non-detector quality”
regions present in the sample; what therefore are the defects responsible for
the worsening of CCE? The answer is clearly given in Fig. 17, which reports
Fig. 18. Frontal IBIL spectrum on a CVD diamond sample of bad electronic qual-
ity. The spectra are decreasing in intensity at increasing proton doses. Proton energy
2 MeV
a wavelength spectrum obtained on a CVD diamond sample of good quality:

the presence of the blue A-band at about 440 nm (2.8 eV), attributed to
dislocation networks, is indicative of the facts that the main contribution
to the decrease in CCE is given by the disordered regions in which a lot of
dislocations are present and that these regions occupy the space between the
“good grains” (the other band at 515 nm, 2.41 eV is the H3 center, probably
created by irradiation itself). The IBIL spectra were taken subsequently and
it is seen that radiation damage probably converts radiative regions to non-
radiative ones (or also creates vacancies in N–N groups that decorate A-band
regions by converting them in H3 centers, which are believed to be N–V–N
centers).
Figure 18 shows the same spectra for a non-detector quality CVD diamond
sample. It appears that there is another broad band at about 600 nm, which
also strongly decreases as a function of radiation damage.
The damage-induced decay of the A-band is shown in Fig. 19, which
reports the area under the peak as a function of proton dose. The decay is
observable after doses on the order of 1015 p/cm2 , which are obtained in few
minutes of measurement. The decay seems to be exponential with the dose,
which could suggest some kind of “positive reaction”, i.e., the presence of
damage itself seems to facilitate the production of damage.
The problem of attributing the space distribution of the luminescent cen-
ters can be solved also by CL maps, which display a better spatial resolution,
as shown in Fig. 20. There are – for the same cross section of the sample to
which Fig. 17 refers – respectively a SEM secondary electron image, and three
CL images (panchromatic and monochromatic at 440 nm and 600 nm). Apart
from the consideration that CL, which is more sensitive to dislocations, gives
a brighter image of the columnar structure of CVD diamond cross section, it
Fig. 19. Decay of intensity of blue A-band at 433 nm as a function of proton dose
Fig. 20. SEM micrograph (secondary electrons) and three CL maps of a cross
section of a CVD diamond sample of good electronic quality (monochromatic at
440 nm, at 600 nm and panchromatic)
Fig. 21. SEM micrographs (secondary and backscattered electrons) and two CL
frontal maps on a sample of bad electronic quality
can be clearly observed that luminescence at 440 nm (blue A-band) gives the
main contribution also in this case, while the luminescence at 600 nm, apart
from some spots, is not spatially well correlated with dislocated regions.
A better view is given in Fig. 21, in which single grains have a sharper
appearance. Here a micrograph by backscattered electrons is also included,
since it gives a sharper identification of single-crystal grain dimensions in
depth. It can be observed that while the luminescence at 440 nm repro-
duces exactly the borders of the two main squared grains at the center of
the image, the luminescence at 600 nm is much broadly distributed in the
volume of the sample. According to our interpretation, this band is due N–
V complexes distributed in the sample and created by lower temperatures
at which the mobility of the species at the surface is lower. These centers
should be responsible for the short CCD of holes in non-detector quality dia-
mond: when filled during priming, they could allow for a longer CCD of holes
themselves. These should be therefore mainly in-grain centers. Centers re-
sponsible for A-band luminescence are certainly intergrain centers, and they
could be responsible for electron recombination, which is mainly radiative.
By illuminating the sample at around 400–440 nm, these centers are emptied
and the space-charge, due to electron capture, disappears, strongly improving
CCD of electrons, which are now able to reach CCD values above the grain
dimensions and to overcome grain boundaries.
Fig. 22. The ID21 beam line of ESRF and the scanning X-ray microscope, with a
layout of XBIC
5 XBIC
The recent availability of X-ray microbeams has opened the possibility of

carrying out IBIC measurements (referred here as XBIC) with a much bet-
ter spatial resolution [20, 21]. This turned out to be possible at ID21 line
at ESRF of Grenoble (see Fig. 22, which shows in some detail the scanning
X-ray microscope (SXM)). The nominal diameter of the X-ray spot is well
below 1 µm in this case, the measurements are carried out in air, pixel size is
1 µm and the dwell time of the beam in each pixel can be varied from 100 to
400 ms. Time-resolved measurements were not possible due to the very tight
time structure of the beam, and therefore the measurements were carried
out in current by using a picoamperometer. The nominal photon intensity at
5.5 KeV was about 108 photons/s. The synchronous acquisition of the pho-
tocurrent signal and of the signal coming from the sample micropositioning
system allowed for image formation on a pixel-by-pixel basis as in proton
IBIC. There are two main differences, however, with respect to IBIC: one is
related to the different energy loss of electrons and protons and the other to
the output signal.
With respect to 2 MeV protons, 5.5 KeV X-rays display a much lower en-
ergy loss (Fig. 23) and an exponential absorption, which lets them penetrate
thicknesses of much more than 100 µm. As a consequence, carrier generation
may be assumed to be roughly uniform across the sample for thicknesses on
the order of 100 µm. Space charge and polarization problems are experimen-
tally absent in XBIC. The photocurrent is a DC signal which may include
gain, and carrier generation represents a strong perturbation with respect to
Fig. 23. Energy absorption coefficient for 5.5 KeV X-rays and percent stopping
power of 2 MeV protons as a function of depth in diamond
Fig. 24. XBIC map of a diamond sample (the same as in Figs. 5 and 6): X-ray
energy, 5.5 KeV, bias voltage +200 V applied at the back side (substrate). The
small square indicates a 100 × 100 µm2 region over which the maps and spectra of
Figs. 25 and 26 were obtained
the IBIC proton case, even if also with protons the local perturbation is not
small. In any case, the photocurrent is another way to get CCE values, and it
will reported in effect as photoconductive gain with respect to dark current
in arbitrary units, since quantitative estimates of CCE are not possible.
An example of an XBIC map is given in Fig. 24, which refers to the same
sample as Figs. 5 and 6. The circular electrode is clearly visible, and the black
spot at the center is due to a drop of silver paste used for gluing the wire to
Fig. 25. XBIC photocurrent maps obtained at different (positive and negative)
bias voltages over the region indicated by a square in Fig. 24
the electrode. The nonuniformity of the current is due to the nonhomogeneity

of the electronic properties of the sample, being the electrode, as observed
by an optical microscope, perfectly smooth.
An idea of the much better spatial resolution achievable is given in Fig. 25,
which shows the maps at different negative and positive bias voltages obtained
in the small square indicated in the previous figure. “Grains” of dimensions
much smaller than 20 µm are visible in particular at low bias voltages, while
the overall uniformity and the average photocurrent increases with bias volt-
age.
It seems that uniformity is better at positive bias; in effect this observation
is confirmed by Fig. 26, which reports the “multichannel spectra” of the
photocurrent obtained by sampling during the dwell time. These spectra are
displayed here in terms of numbers of pixels instead of counts, in order to
make a better comparison in terms of uniformity. Since the bias is applied on
the back side with respect to the incoming beam and since X-ray absorption
is much larger in the first 100 µm (Fig. 23), it can be assumed that the
main contribution is given by electrons, which have at their disposal a much
longer path. This conclusion is in agreement with what was reported for the
same sample in the discussion concerning lateral IBIC. In effect, the spectra
of Fig. 26 are narrower at positive bias voltages with respect to negative
ones, and this means that spatial uniformity is better, even if the mean
photocurrent is larger at negative bias. The logarithmic grey scale used in
these figures is used to improve the contrast between nearby regions in order
to observe “structures” given by the photocurrent signal and to obtain a
better correspondence with the morphology of the material. Since the sample
is quite thick (0.6 mm) and since also since the penetration depth of X-rays is
quite large, this correspondence is not easily seen (and in effect it can be and
it has been observed in much thinner samples). However, with respect to a
colour-based IBIC map, a map graphically obtained with a suitably selected
logarithmic grey scale can give a detailed and more visible structure (Fig. 27,
Fig. 26. XBIC photocurrent spectra of the maps reported in Fig. 25
which has been obtained with the same data of Fig. 6, in which grains of a
clear geometrical structure make their appearance).
One question could be given by the different “grain” size observed in XBIC
with respect to IBIC maps. This could be related to the shorter penetration
depths of protons in IBIC with respect to X-rays in XBIC and to the larger
grain sizes at the growth side of the sample.
As already stated, one of the main applications of IBIC/XBIC is the
investigation of the uniformity of the electric response in terms of CCE. It
Fig. 27. IBIC map obtained from the same data as Fig. 6 by a suitably selected
grey scale
is possible to quantitatively define the uniformity vj by dividing the selected

area in bins of square size and by calculating the fluctuations between the
average values in the bins and the average value over the total area, according
to the following equation:

i (sij − s)
1 2
vj = 1 − , (1)
s Nj
where s is the overall average signal, sj is the average signal of the ith
bin of bin size j and Nj is the number of bins of bin size j (j indicates the
square root of the bin area).
This exercise is presented in Fig. 28 for the central bottom region, 480 ×
480 µm2 wide, of Fig. 27. The results, which are compared with the results
obtained by a similar exercise carried out on an XBIC map of the same size
and on the same sample, are reported in Fig. 29. It is seen that there is a
relatively good agreement between the two, even if the selected regions are
not the same.
It has to be noted that in the case of IBIC the homogeneity has been
improved by blue light illumination (Fig. 6 and the relevant part of the dis-
cussion in Sect. 3). These results are also in agreement with those obtained
directly by mips in a much more cumbersome way (see [16] at p. 162) by
placing two position-sensitive detectors (two couples of silicon strip detector
with crossed strips), by triggering the event and by reconstructing the track
across the diamond detector by a suitable analysis of the event. It has also
Fig. 28. IBIC maps of a region 480 × 480 mm2 wide, in which CCE has been
averaged over bins of increasing surface areas
Fig. 29. Efficiency uniformity, as defined by (1), as a function of bin size, obtained
in the same way from XBIC and IBIC maps
been noticed that X-ray priming improves homogeneity: this was not possible
to obtain by frontal IBIC, because of the polarization effect, but it has been
observed by lateral IBIC by a suitable “simulation” of a mip.
This result, obtained by following the lines depicted in the illustration of
Fig. 12, is shown in Fig. 30. These results prove that, at low size bins at least,
effectively priming improves CCE uniformity.
Fig. 30. Efficiency uniformity, as defined by (1), as a function of a linear bin size,
obtained from lateral IBIC maps as reported in Fig. 12, both for a primed and an
unprimed state
6 Conclusions
A new kind of ion microscopy, applied in this case to qualify CVD diamond
as a nuclear detector, or even simply as an electronic material, has been
described both in terms of capabilities and of specific results obtained on
samples of different quality. IBIC and IBIL can be used together with other
beam techniques like PIXE, ERD or RBS in order to make a correlation with
chemical impurities. The direct comparison between IBIC and IBIL maps
leads to the conclusion that the morphologically polycrystalline CVD dia-
mond is in effect also “electronically” polycrystalline, i.e., the largest values
of CCE and CCD are found in the central part of specifically “good” single
crystal grains or columns and that the main defects responsible for the low
CCE values found in CVD diamond are due to the radiative recombination
defects, which are distributed along the grain boundaries. The specific con-
tribution of the two carriers types to CCE can be separated by lateral IBIC
and the profile of the electrical field across the sample can be suitably calcu-
lated, by putting in evidence the still unsatisfactory quality of the electrical
contacts. The effects both of standard priming and of blue light illumina-
tion have been also evidenced in terms of improvement of the homogeneity
of the response, both at the surface and along the thickness of the sample.
Time-resolved IBIC and IBIL may supply in the future even more powerful
methods of investigation of CVD diamond, both in terms of material and of
a specific detector. Improvements are under way in order to obtain a value of
lateral resolution better than 1 mm and may be to reach a value of the order
of 100 nm, by a more accurate design of focusing quadrupoles. If this will be
the case, ion microscopy could enter the class of more standard and much
more widely used electron microscopies, even if the experimental apparatus is
much more cumbersome. In any case, it must be emphasized that the amount
of information given by this kind of microscopy and the variety of techniques
which can be applied is much wider, as exemplified in this work.
Acknowledgements
The author acknowledges the important and direct contributions given by

E. Vittone, A. Lo Giudice, F. Fizzotti, P. Olivero and C. Paolini, during
various periods of IBIC, IBIL and XBIC measurements from 1995 to 2004,
concerning data acquisition, analysis and interpretation. Many IBIC measure-
ments reported in this review were carried out at Ruder Boskovic Institute
in Zagreb (Croatia), with the professional help of M. Jaksic, S. Fazinic and
I. Bogdanovic.
References
[1] M. B. H. Breese, D. N. Jamieson, P. J. C. King: Materials Analysis Using a
Nuclear Microprobe (Wiley, New York 1996) 240
[2] N. P. O. Homman, C. Yang, K. J. Malmquist: Nucl. Instr. Meth. A 353, 610
(1994) 240
[3] C. Manfredotti, F. Fizzotti, P. Polesello, E. Vittone, M. Truccato, A. L. Giu-
dice, M. Jaksic, P. Rossi: Nucl. Instr. Meth. B 1333, 136–138 (1998) 240,
247, 251
[4] C. Manfredotti, G. Apostolo, G. Cinque, F. Fizzotti, P. Polesello, E. Vittone,
M. Truccato, A. L. Giudice, G. Egeni, V. Rudello, P. Rossi: Diamond Relat.
Mater. 7, 742 (1998) 240, 251
[5] C. Manfredotti, E. Vittone, P. Polesello, F. Fizzotti, M. Jaksic, I. Bodganovic,
V. Valkovic: Scanning ion beam microscopy – a new tool for mapping transport
properties of semiconductor nuclear detectors, in J. Duggan, I. Morgan (Eds.):
CP392, Application of Accelerators in Research and Industry (AIP, New York
1997) pp. 705–708 240
[6] F. Cervellera, C. Donolato, G. P. Egeni, G. Fortuna, R. Nipoti, P. Polesello,
P. Rossi, V. Rudello, E. Vittone, M. Viviani: Nucl. Instr. Meth. B 130, 25
(1997) 240, 241
[7] C. Manfredotti, F. Fizzotti, E. Vittone, M. Boero, P. Polesello, S. Galassini,
M. Jaksic, S. Fazinic, I. Bogdanovic, M. Valkovic: Nucl. Instr. Meth. B 100,
133 (1995) 240, 242
[8] P. J. Sellin, M. B. H. Breese, A. P. Knights, L. C. Alves, R. S. Sussmann,
A. J. Whitehead: Appl. Phys. Lett. 77, 913 (2000) 240
[9] C. Manfredotti, F. Fizzotti, P. Polesello, E. Vittone, P. Rossi, G. Egeni,
V. Rudello, I. Bogdanovic, M. Jaksic, V. Valkovic: Nucl. Instr. Meth. B 130,
491 (1997) 242, 246
[10] C. Manfredotti, F. Fizzotti, A. L. Giudice, P. Polesello, E. Vittone, R. Lu,
M. Jaksic: Diamond Relat. Mater. 8, 1597 (1999) 242, 246
[11] C. Manfredotti, F. Fizzotti, E. Vittone, P. Polesello, F. Wang: Phys. Stat.
Solidi A 54, 327 (1996) 242
[12] C. Manfredotti, F. Fizzotti, A. L. Giudice, C. Paolini, E. Vittone, R. Lu: Nucl.

Instr. Meth. B 187, 566 (2002) 244
[13] E. Vittone, F. Fizzotti, A. L. Giudice, C. Paolini, C. Manfredotti: Nucl. Instr.
Meth. B 446, 161–163 (2000) 246
[14] C. Manfredotti, F. Fizzotti, P. Polesello, E. Vittone: Nucl. Instr. Meth. A 426,
156 (1999) 246
[15] R. Lu, C. Manfredotti, F. Fizzotti, E. Vittone, A. L. Giudice: Mater. Sci. Eng.
B 90, 191 (2002) 246
[16] D. Meier: Diamond Detectors for Particle Detection and Tracking, Phd thesis,
University of Heidelberg, Heidelberg (1999) 248, 249, 261
[17] C. Manfredotti, F. Fizzotti, A. L. Giudice, P. Polesello, E. Vittone, M. Truc-
cato, P. Rossi: Diamond Relat. Mater. 8, 1592 (1999) 251
[18] C. Manfredotti, E. Vittone, A. L. Giudice, C. Paolini, F. Fizzotti, G. Dinca,
V. Ralchenko, S. V. Nistor: Diamond Relat. Mater. 10, 568 (2001) 251
[19] A. L. Giudice, G. Pratesi, P. Olivero, C. Paolini, E. Vittone, C. Manfredotti,
F. Sammiceli, V. Rigato: Nucl. Instr. Meth. B 210, 429 (2003) 251
[20] E. Vittone, A. L. Giudice, C. Paolini, F. Fizzotti, C. Manfredotti, R. Barrett:
[21] E. Vittone, A. L. Giudice, C. Paolini, P. Olivero, C. Manfredotti, R. Barrett,
V. Rigato: Nucl. Instr. Meth. B 210, 159 (2003) 257
Index
cathodoluminescence (CL), 251 lateral IBIC, 242, 243, 245–250, 253,
charge collection distance (CCD), 244, 259, 262, 263
245, 247, 249, 256, 263 ion beam induced luminescence (IBIL),
charge collection efficiency (CCE), 240, 240–242, 251–256, 263
242, 244–254, 258, 260, 262, 263 frontal IBIL, 253, 254, 256
CCE maps, 242, 244 IBIL maps, 251–253, 255, 256, 263
CVD diamond, 242, 244, 254, 263 lateral IBIL, 252, 253
ion beam-induced charge (IBIC), 240
detectors, 240, 241, 244, 248, 253, 254, ion beam-induced luminescence (IBIL),
256, 261, 263 240
ion beam induced charge (IBIC), Ramo’s theorem, 245–247

241–243, 245–252, 257, 258, 260,
261, 263 X-ray beam induced charge (XBIC),
frontal IBIC, 242–244, 262 257–263
IBIC maps, 244, 245, 252, 259–263 XBIC maps, 258–262
Measurements of Defect Density Inside CVD
Diamond Films Through Nuclear Particle
Penetration
Renato Potenza and Cristina Tuvé
INFN and Dipartimento di Fisica e Astronomia, Università di Catania,

Via S. Sofia 64, I-95123, Catania, Italy
Cristina.Tuve@ct.infn.it
Abstract. Fast nuclear particles (light or heavy ions), passing through a diamond
film, deposit energy by ionization. It is possible to collect the developed charges by
applying an electric voltage through both sides of the film. Experimental results
show that the positive hole charges and the negative electron charges have very
different mean free paths inside a real diamond film. Holes travel more deeply than
electrons, so that the charges collected under a given voltage and under the oppo-
site one are very different, as a suitable “two fluid” microscopic model of electrical
conduction of the film can show. Furthermore, using fast particles of different pene-
tration produced by a Tandem accelerator, it is possible to obtain a “photograph” of
the differential charge collection, which can show the distribution of charge produc-
tion (through the Bragg’s ionization curve) and destruction (through the defects,
and therefore on defect density along the particle path). Minimization of this defect
density allows us to optimize the deposition procedure of CVD diamonds.
1 Diamond as Radiation Detector

1.1 Properties of Diamond
Diamond is a solid with a density ρ = 3.51 g/cm3 . It shows a gap of 5.47 eV

between valence and conduction bands, compared to that of 1.12 eV of sil-
icon (Fig. 1). So, while silicon can be called a semiconductor, diamond is
intrinsically an insulator. Its resistivity at room temperature is in fact ρD >
1012 Ω cm, more than 4 · 106 times that of pure silicon (ρSi = 2.3 · 105 Ω cm [1])
and more than 5 · 1017 times that of a good conductor such as copper with
ρCu = 1.7 · 10−6 Ω cm. Diamond does not need doping to act as a nuclear
detector. There is no need of an artificial depletion region; it is always totally
depleted. It simply does not conduct in both directions, unless an ionizing
particle crosses it. In this case electron–hole pairs are formed in the bulk of
the material, and they can become carriers of electric charge for a short time
and can give rise to a pulse like in other solid state detectors, but with the
characteristic property of an insulator. That is, a pulse can be produced both
for positive or negative polarization of the crystal. The mean energy required
to create an electron–hole pair in diamond is = 13.3 eV, while in silicon it
is only 3.6 eV (Fig. 1). The minimum energy resolution width of a detector is
268 Renato Potenza and Cristina Tuvé
Fig. 1. Gap and pair production energies in diamond and silicon
given by the statistical fluctuations in the number of carrier pairs N c created

by the passage of the ionizing particle. That is:
√
∆E ∆N c 1
= = √ =√ , (1)
E Nc Nc E
where N c = E/ and assuming valid Poisson’s statistics. E means here the
energy released by the particle into the detector. Since is rather higher for
the diamond, it is certainly worse than the silicon as a radiation detector
from the point of view of the energy resolution. Diamond has, however, a
great advantage and conversely a great disadvantage with respect to silicon.
The advantage consists in its much stronger hardness to radiation damage.
The disadvantage is given by the fact that good natural diamonds cost a great
deal and that synthetic ones have an energy resolution much worse than fore-
seen by (1) because lattice defects cause independently both an increase of
the dispersion ∆N c and a decrease of the value of N c with respect to the the-
oretical ones. The radiation hardness of diamond has been extensively tested
through charged particle bombardment [2] and in some cases also through
neutron bombardment [3]. The results obtained with charged particles are all
in agreement and show that diamond, instead of being damaged, is in many
cases improved by the bombardment. This effect is called pumping and is
obtained by irradiation for a given time with ionizing particles, frequently
β-particles. It is discussed in another paper in this volume [4] and seems to
Defect Density Measurements Inside CVD Diamond Films 269
graphite diffuse crystal defects
a)
b)
-1
2.4 cm
c) 1320.0 1332.5 1345.0
0 2000 4000 6000

-1
Wavenumber shift (cm )
Fig. 2. Raman spectra of different diamond specimens in order of increasing purity

of the CH4 –H2 vapour mixture
be due to the saturation of some type of lattice defects by means of the many
electrons and holes produced by a strong flux of ionizing particles. Neutron
effects are still under study, but also for them diamond seems to be sufficiently
hard. In effect recent measurements [5] suggest that diamond detectors could
advantageously replace silicon detectors as neutron flux monitors in the core
of the experimental nuclear fusion reactor at JET, Culham, U.K., due to the
much longer life of diamond with respect to silicon detectors of comparable
volumes.
1.2 Lattice Defects in Synthetic Diamond Crystals
Lattice defects in diamonds can be detected using Raman scattering tech-

niques. Figure 2 shows some results obtained on synthetic diamonds grown
by chemical vapour deposition (CVD), that is, condensation of carbon from
a CH4 (1%)–H2 (99%) gas mixture [4] onto silicon crystal substrates. Let this
technique be called CVD heteroepitaxial growth (polycristalline, pCVD). As
one can see in the figure, the Raman spectra clearly detect the presence of
graphite, the allotropic crystalline state of the carbon, the presence of diffuse
lattice defects and impurities, and, of course, the sharp peak produced by
diamond at ∆k = ∆(1/λ) = 1332.7 ± 1.2 cm−1 .
Both graphite and diffuse crystal defects can be decreased, improving the
purity of the vapour mixture so as to avoid as much as possible nitrogen and
oxygen. The spectra in Fig. 2 are taken in conditions of increasing purity.
350
300 Negative
bias
250
Counts
200 Positive
bias
150
100
50
0
0 20 40 60 80
h (%)
Fig. 3. Pulse height efficiency η = Qc /Q 0 of a “pure” diamond specimen for α-
particles of 5.5 MeV. Qc is the collected charge and Q0 is the charge really produced
inside the diamond. This has been subject to β-irradiation before being used.
Fig. 4. Scheme of the circuit in which the diamond sensor is inserted. The reported
case is that of positive polarization of the diamond (Au electrode is positive)
One can note that in the specimen in the spectrum (c) there are no more
defects or impurities detectable with Raman technique, that is, the specimen
can be considered “pure”. Only these specimens will be considered as nuclear
detectors in this paper. However, they contain still many defects which in-
fluence their quality as nuclear radiation detectors, as Fig. 3 shows for the
spectra produced by 241 Am α-particles in a similar diamond specimen for
both polarizations.
The spectra are reported in Fig. 3 as functions of the charge collection
efficiency:
Qc ph meas
η= = , (2)
Q0 ph max
where Qc is the charge detected in the external charge-sensitive electrical

circuit, and Q0 is the charge produced inside the diamond by the passage
of the ionizing particle, so that phmeas and phmax refer to the measured
and maximum obtainable pulse height, respectively. To detect the spectra in
Fig. 3 and all the other ones with which this paper will deal, it is necessary
to adapt the diamond crystals to an external electrical circuit. To do this one
must distinguish between the substrate side of the diamond crystal, where it
is in contact with the semiconductive silicon substrate, and the growth side,
where the free CVD diamond surface appears. Figure 4 shows the schematic
drawing of a pCVD diamond inserted in the circuit which allows us to record
the electrical signals produced by the passage of a nuclear particle. A thin
(500 Å) gold deposit, sometimes done above a first, also thin (100 Å) titanium
deposit, assures electrical contact on the growth side.
The silicon substrate itself assures the contact on the substrate side.
A charge-sensitive amplification chain allows us to record the signals from
the diamond. The bias voltage is applied to the growth side, so that pos-
itive or negative polarization means that the growth surface, that is, the
gold electrode, is respectively positive or negative with respect to the silicon
substrate. Figure 4 reports the case of positive polarization, as one can see
from the indicated directions of electron and hole motion. Referring again to
Fig. 3, remembering that for a silicon detector η ≈ 1 with very good approx-
imation, one clearly sees that the spectra produced by the diamond detector
not only show mean values of the pulse heights significantly less than the
maximum possible and in any case much less then those from equivalent sil-
icon detectors, but also that the widths of the spectra are very large, though
the specimen could be considered “pure”, that is, detector grade, from the
point of view of a Raman spectrum. The resolution widths are much larger
than the minimum foreseen by (1). So one must conclude that the influence
of crystal defects is still large as it concerns the qualities of a detector and,
since the Raman spectrum is already flat, these defects can be detected by
now essentially using other probes. The proposal is to use as probes the same
nuclear particles one would like to detect.
1.3 Application of Diamond to Beam and Beam Profile

Monitoring
Before discussing in more detail how to use nuclear radiation to measure the
defect distribution in “pure” diamond crystals, it is worthwhile to discuss
how these diamonds behave when bombarded with very high nuclear particle
intensities, say I > 1 pnA = 6.25 · 109 p/s, that is, intensities for which the
crystal cannot resolve each single particle from the others within its intrinsic
resolving time τ r ≈ 1 ns. In this case the response of the detector is a direct
current in the external circuit, where the charge-amplifying chain is replaced
by a direct current amplifier. If the gold contacts (see Fig. 4) are designed
in the form of a network of pixels, the insulating properties of the diamond
assure that each pixel collects, without detectalbe cross-over among pixels,
the current produced in the cylindrical column traced along electric field lines
between the gold pixel on one side and its projection on the substrate on the
other side. So such a pixel detector bombarded by a beam of particles from
an accelerator impinging along the electrical field lines inside the diamond
can give very good information about the beam profile after an easy internal
calibration made by moving the detector so that the same part of the beam
can impinge on different pixels. Using such a direct current amplification cir-
cuit, we obtained the results shown in Fig. 5 [6]. In that figure the response
of a diamond detector is compared with the measurement of the intensity
of a 26 MeV proton beam in the Faraday cup at the end of the pipe of the
15 MV Tandem accelerator of the National Southern Laboratories (LNS)
of National Institute of Nuclear Physics (INFN) in Catania (Italy). As can
be seen in Fig. 5, the diamond follows instantaneously, in the time scale of
the figure, the changes of beam intensity. Moreover, the fluctuations in the
measurement are much less for the diamond, as seen from the fact that the
electron–hole current is much larger than the original proton current. Taking
into account the amplification factor of the circuit, it was estimated that the
ratio between proton and related electron–hole currents was 1/23 000 [6], so
that possible influence from secondary electron emission, which is a cause
of detectable error in the measurements of currents with Faraday cup, is ab-
solutely negligible. The diamond used in these measurements was only 50 µm
thick, so that a simple calculation, taking into account that 26 MeV protons
lose 0.33 MeV in passing through and remembering the energy required per
carrier pair, tells us that the charge collection efficiency at the used polariza-
tion voltage, 150 V, is about 60%. Other less rough evaluations through the
so-called restricted energy loss [7] or the computation through the standard
electron–hole pair number produced by minimum ionizing particles (MIP) [8]
give the same values within a 10% variation. As will be seen below, because
the protons pass over the diamond, there is almost no difference between the
efficiency for positive or negative polarization. This is again confirmation that
pCVD diamonds have efficiency significantly less than 1, which is different
from that of silicon detectors. The diamond detector we used is transparent
to low-energy protons and also to other low-energy ions of intermediate mass.
This means that if one removes by lapping the substrate and makes a second
gold–titanium contact on the substrate side of the diamond, the new detec-
tor offers a very good measurement of beam intensity and profile without
appreciably perturbing the beam itself, which can be monitored continuously
during the applications.
This is very important, e.g., in medical applications of nuclear radiation or
in monitoring fast neutron fluxes in the cores of nuclear reactors, particularly
fusion reactors. And this quality is well exploited by the high hardness to
radiation, proved by the fact that the diamond specimen used in [6] lasted
several hours under a current of 10 pnA of 26 MeV protons, that is, a fluence
Diamond
sensor
TANDEM
26 MeV protons Accelerator
5 mm Collimator
Faraday
cup
A
to pump
0.06
Diamond detector
current (µA)
0.04
0.02
0.00
0.2 Faraday cup

current (pA)
0.1
0.0
560 630 700 770 840 910

t (sec)
Fig. 5. Intensity vs. time (bottom) for a 27 MeV proton beam from the 15 MV
Tandem accelerator of LNS in Catania (Italy), measured simultaneously through a
diamond detector and the Faraday cup shown at the top
φ ≈ 1015 p/cm2 , without any sign of deterioration, whereas a good silicon

detector would have been damaged by a fluence one hundredth as large [2, 3].
2 Mechanism of Conduction in Circuits Including

Diamond
Let us consider the electrical circuit of Fig. 4 containing a diamond. Call V
−
the applied voltage, and let 2Q0 = Q+ 0 − Q0 = 2N c e be the charge produced
at a point x by the passage of a particle through the diamond. Let dU =
−Q+ 0 dV = −Q0 Edx be the energy gained by the charge Q0 during its
+ +
movement along a length dx parallel to the field E. The work needed to supply
this energy is done by the electromotive force of the generator, which moves
the induced charge of opposite sign dQ− ind in the external circuit, maintaining
a constant applied voltage V = EL, where L is the thickness of the diamond.
This work is given by dW = −V dQ− ind = −ELdQind .
−
Taking the x-axis going from the substrate to the growth side of the
diamond, one has:
−
−dU = Q+
0 Edx = −dW = ELdQind . (3)
That is,
dx v+ µ+
dQ− +
ind = Q0 = −Q+
0 dt = −Q+
0 Edt, (3a)
L L L
from which:

d+ E + +
−Q−
ind = − dQ− +
ind = Q0
c
≤ Q+
0 µ τ c ≤ Q+
0, (3b)
L L
and three similar equations with the appropriate signs for the negative charge
Q− −
0 , where Qi → −Qi and all the superscripts + → −. In (3b) and its
+
equivalent d± ± ± ±
c ≤ λc = µ τ c E ≤ L are the mean actual distances, in a
crystal of finite thickness, covered by the charges Q± 0 travelling inside under
the action of the applied field and subject to: (i) absorption from donor
or acceptor defects along the path, which is responsible for the finite λ± c ,
the charge collection distances of the carriers in that crystal; (ii) abrupt
interruption of the path for the surviving carriers able to reach the contacts,
which makes d± ± ±
c ≡ dc (x) < λc near the contacts themselves. Strictly related
to the λ±c are the mobility µ+
= v + /E and the charge collection time τ + c .
Equations (3), (3a) and (3b) are expression of the so-called Ramo’s theorem,
demonstrated by S. Ramo in 1939 [9].
Noting that when an ionizing particle passes through a solid state detector
−
it is always, at a given x, Q+0 = −Q0 ≡ Q0 ≡ Q0 (x), one can now define the
corresponding collected charge as
− − d−
c (x) d+
c (x)
ind − Qind = −Q0
Qc (x) = Q+ + Q+
0 (4)
L L
(d− +
c + dc )
= Q0 ≤ Q0 (x),
L
and the charge collection efficiency at the same x (see (2)) as:
Qc (d− + d+
c ) (λ− + λ+
c )
η(x) = = c < c . (5)
Q0 L L
It is clear that if the mean free paths λ±

c are both equal to or bigger
than L, taking into account the interruptions of carrier paths at contact
surfaces, where moving and induced charges neutralize each other, one has:
+x
−Q−
ind = Q0 ,
L
and
− (L − x)
ind = −Q0
Q+ .
L
As a result, Qc = Q0 , from (4), as happens in the depletion barriers of all
good semiconductors such as silicon. When, on the other hand, acceptor or
donor defects present in the material forbid long free paths for charge carriers,
−
then not only is Qc < Q0 , but also it can happen that λ+ c = λc . This is just
the case of the pCVD diamonds, as various recent results show [4, 10, 11]. Of
course, the charge collection efficiency η for a whole crystal is a mean value
of η(x), as said below.
2.1 Modified Hecht’s Model for Charge Transport Inside

Diamond
This section follows, with modifications, the treatment given in [12].
2.1.1 Absorption of Carriers During Their Transport Along

the Electric Field Lines
Because of the finiteness of the specimen, the values of the d± c (x)’s which
appear in (5) are less than those of the mean free paths λ±c . If the specimen
can be considered infinite, that is, L λ±
c , then edge effects can be neglected
and the d± ± ±
c ’s reduce to the λc . To compute the values of the dc ’s in a
finite specimen it is, however, possible to start from the situation in the
infinite specimen. The general expression of a mean free path for particles in
a medium is given by
1
λ±
c = (6)
N ± σ±
where σ is the cross section of the absorption centres and N is their con-
centration (volume−1 ). There is no need that the values of σ or the N be
the same for electrons and holes, so it will be assumed that N + = N − and
σ + = σ − . There is also no need that they be constant with position inside the
specimen, so that in general λ± ±
c ≡ λc (x, y, z). However, it is useful to assume
±
for the moment that λc = const. Now, referring to Fig. 4 with the direction
of the x-axis going from bottom to top, as shown also in Fig. 6, that is, from
substrate to growth side of the diamond along the lines of the applied field,
−
assume that an incident particle creates at the level x, N + c (x, 0) = N c (x, 0)
carriers of both signs. Following the definition of the absorption cross sec-
tion σ, one has that after a displacement
of a given type of carriers
± ± −
N± ±
= N±
(−N d σ ) ±
c (x,
) = N c (x, 0)e c (x, 0)e , (7)
λc
because at every
dN ± 1
c
= − ± N± ± ±
c (x,
) = −p (x,
)N c (x,
), (7a)
d
λc
where 1/λ± c is just the constant probability of absorption per unit length
p± (x,
) for positive or negative carriers, constant here since one is speaking
of infinite homogeneous specimens. Note that N ± c (x,
) is just the number
of surviving carriers at a depth
away from the point at coordinate x. To
compute the mean free path of the carriers inside an infinite and homogeneous
medium, one can write:
∞ ± ∞ − λ±

N (x,
)d

e c d
λ = 0∞ ± = λ±
c
= 0 c . (7b)
∞ − λ±

0 N c (x,
)d
e c d
2.1.2 Edge Effects
Trying now to compute the charge collection distances d± c for finite specimens,
one must note that carriers that succeed in reaching the edges of the specimen,
that is, the surfaces at
= x moving towards the substrate side, or
= L − x
moving in the other direction, do not proceed anymore since they are all
absorbed at these values of
. So the mean value of the path length becomes:
ap±
± ∞ ±
0

N ±
c (x,
)d
+ ap ap± N c (x,
)d
d±
c = ∞ ±
0 N c (x,
)d
ap± −
± ∞ −
±
0
e λc
d
+ a±
p ap± e λc
d
= ∞ − ±
0 e λc
d
±
ap

−
= λ± < λ±
±
c 1−e λc
c , (8)
where a± p can take the value of x or L − x depending on both the type

−
+ = a− = L − x and
of carriers and the polarization p. More precisely, a+
+ −
a− = a+ = x, where subscripts refer to polarizations and superscripts to
carriers. Because of the dependence of the a± ±
p on x, the dc depend on x also
in this simple model of finite homogeneous specimens. The total number of
carriers that succeed in escaping absorption in diamond is just:

1 ∞ ± ±
N c (x,
)d
≡ Np> .
λ±
c ap ±
That is,
a±
± 1 p
Np> = N ± (x, 0) − N±
c (x,
)d
Λ± 0
∞ a±
1 1 p
≡ N±
c (x,
)d
− N±
c (x,
)d
. (8a)
Λ± 0 Λ± 0
Λ± is a normalization constant, equal to λ± c only for infinite homogeneous

media; its value depends on the definition of p± (x,
) outside the edges of the
specimen. p± (x,
) has, in fact, physical meaning only inside the specimen,
but it can be convenient to extrapolate it outside. How to do the extrapolation
does not matter, because of the appearance in (8) of Λ± , which depends on
p± (x,
) in order to compensate for different arbitrary choices of the nonphys-
ical part. If the charge Q0 were all produced at the point x then substitution
of (8) in (5) would give the charge collection efficiency η. But when an ioniz-
ing particle impinges on the detector, charges are produced all along a path
which begins at the entrance edge and goes through up to the end of the
range R, or even passes over the detector if the particle energy is sufficiently
high. So, in the simple case actually being discussed, the charge produced
between x and x + dx inside the detector, neglecting the nonuniform deposit
due to Bragg’s distribution, is given by
dQ0 (x) Q0,tot Q0,tot

= = = const, (9)
dx min[Rcosθ, L] min[G, L]
where G is the penetration depth of the particle along the lines of the electric
field and θ is the angle of incidence on the growth surface. One obtains, using
(5), (8) and (9):
min[G,L]
Qc,tot 0
η(x) dQ
dx dx
η= = min[G,L] dQ
Q0,tot
0 dx dx
min[G,L] (d+ −
c (x)+dc (x)) Q
0,tot
0 L · min[G,L] dx
=
Q0,tot

min[G,L] + (L−x)
− + min[G,L] − − x−
0
λc 1 − e λc
dx + 0 λc 1 − e λc
dx
=
L · min[G, L]
L−min[G,L] min[G,L]

− − L+ − −
− 2
−e + (λc ) 1 − e
+
− (λ+
c )
2
e λc λc λc
(λ+
c + λc )
= − ,
L L · min[G, L]
(10)
for positive polarization (the case reported in Fig. 4); for negative polarization
−
λ+c and λc interchange between each other to give η− . When it happens that
−
− (λ+
c + λc ) δ
λ+
c = λc = ≡ ,
2 2
(10) reduces to the original Hecht formula [13]:

δ δ −
2min[G,L]
2(min[G,L]−L)
η= 1− 1−e δ 1+e δ . (10a)
L 4G
However, there is an enormous difference between (10) and (10a). Equa-
tion (10), in fact, foresees different efficiencies of the same detector for the
same particles when the polarization is reversed, that is, when the growth
side is positive or negative with respect to the substrate side, due to the
− −
different values of λ+ +
c and λc . If one supposes for a moment that λc > λc ,
the efficiency for low penetrating particles in negative polarization is defi-
nitely less than that in positive polarization. In fact, when the particles have
G λ± c < L, either because they have low energy or because they impinge
with high incident angle, in positive polarization one has

− L+
+ −
λc e λc
+ λc
(λ+ + λ−
c ) λ+ − L+
η+ ≈ c − = c 1 − e λc ,
L L L
and in negative one has

− L−
− +
λc e λc
+ λc
(λ− + λ+c ) λ− − L−
η− ≈ c − = c 1 − e λc < η+ ,
L L L
−
provided that L > λ+
c > λc .
2.2 Effect of Nonuniform Bragg’s Deposit of Charge into Diamond

The nonuniform production of charge carriers inside the diamond due to the
Bragg’s curve B(s) of specific energy loss along the trajectory of the particle
is easily taken into account by changing (9) to
dQ0 (x) x
= B[x(s)] = B , (9a)
dx cosθ
where θ the incidence angle as above. Equation (10) becomes
min[G,L]
Qc,tot η(x) dQ dx
η− = = 0 min[G,L] dx
Q0,tot dQ
dx
0 dx
min[G,L] (d+ −
c (x) + dc (x)) dQ0,tot
0 L dx dx
= , (10b)
Q0,tot
and η+ and η− can be computed numerically.
Fig. 6. Schematic representation of in-grain and grain-boundary defects in diamond
2.3 Nonuniform Distributions of Charge and Defects
In finite and inhomogeneous diamond crystals the measurement of charge

collection efficiency for both polarizations of the crystal is a very powerful tool
to obtain the distribution of defects responsible for the quality of the crystal,
and consequently to allow the optimization of the growth parameters of the
crystals themselves. In such specimens the probabilities of disappearance of
carriers p± (x,
) in (7) is no longer constant. So the solution to (7a) becomes
the more general:

− p(x,)d
N±c (x,
) = N ±
c (x, 0)e 0 , (11)
and the collection distances become:

a±
± ∞ ±
0
p

N ±
c (x,
)d
+ ap a± N c (x,
)d
±
dc = ∞ ± p
0 N c (x,
)d
a± ∞
− p± (x,)d − p± (x,)d
0
p

e 0 d
+ a±
p a± e 0 d
= ∞ p
, (12)
− p± (x,)d
0
e 0 d
formally identical to (8).

The probability of disappearance p± (x,
)d
of a given carrier in the path
between
and
+ d
is, of course, proportional to the product N ± σ ± of (6).
Assuming that the cross sections for the various processes that give rise to
the absorption of a given type of carriers are constant at every point inside
the diamond, p± (x,
) is directly proportional only to the local concentrations
of defects N ± . Now, these last can be divided into two types [4], as shown
in Fig. 6: those distributed all along the diamond (the so-called in-grain
defects), which depend on the impurities and on random accumulations of
vacancies, and those which depend on the mismatch between the reticular
constants of diamond and of the substrate, which are located preferentially
in the columnar structure [4], that is, in the first part of the deposit, near the
substrate surface (border-grain defects). As said below, the in-grain defects

seem to affect the two types of carriers differently, with a greater influence on
the electrons, while the border-grain ones, being essentially structural defects,
less dependent on impurities in the crystal, seem to affect both carriers in
a similar way. So a suitable way to write the probability of disappearance
seems to be:
1 1 − (x∓)
p± (x,
) = + e Λ p = ±1, (13)
λ±
g λb
where p = ±1 is the polarization given to the crystal, assuming that the con-
centration of border-grain defects goes down exponentially from substrate to
growth surface. Now, substituting N ± (x, 0) with B(x) (9a) and using p± (x,
)
given by (13) in (11), one can finally compute numerically the correct charge
collection efficiency η± for positive and negative polarizations of a diamond
specimen bombarded with different penetrating particles. The integration
must be done numerically because (11), with the mentioned substitutions,
contains combinations of Γ and γ incomplete functions [14]. One can expect
that also in specimens considered pure from the point of view of Raman
scattering the two curves η± vs. G should be significantly different since the
defects still influence the quality of the diamond, as shown, e.g., by energy
resolution. So, the comparison of the results of the measurements of the η’s
in both polarization states can give much better information about the dis-
tribution of defects inside the diamond than Raman effects, as shown in the
next section.
3 Charge Collection Efficiencies of Diamond Detectors

Under Ion Bombardment
The results of typical measurements of mean charge collection efficiencies η±
(for positive and negative polarization) in thin pCVD diamond detectors are
reported in Figs. 7 and 8. In the first case the employed method shows a spec-
imen that is sufficiently good. In fact, the theoretical curves computed with
the simple model of (10) with uniform distribution of defects (mirrored by
the constant values of p(x,
) = (λ±c )
−1
) are very near to the curve computed
with the complete model; the length Λ in the exponent of (13) is large, so that
the defects are almost uniformly distributed. This specimen is practically the
best that one can hope to obtain from the technique of pCVD, though it is
still not satisfactory as a detector, because of the big difference between λ+ c
and λ− c and the fact that the resolution width is still on the order of 35% in
both states of polarization (compare with Fig. 3). The efficiencies reported in
Fig. 8 denote a worse specimen. The efficiency for G > L is low (on the order
of 5%), the curves given by (10) are very different from those given by the
complete model and the defect density is high in the uniform terms as in the
Fig. 7. Charge collection efficiency of a first 55µm thick diamond detector bom-
barded with 12 C at various energies. The penetration depths of 12 C particles are
also shown. Upward pointing triangles: positive polarization; downward pointing
triangles: negative polarization. Theoretical curves: (a) simple model (10) with
−
λ+c = 17 µm and λc = 0.5 µm; (b) same model with account of Bragg’s curve; (c)
−
complete model (13) with λ+ g = 33 µm, λg = 0.8 µm, λb = 25 µm and Λ = 37 µm
exponential one. As is said below, these different results can be associated

with different values of the parameters of the reactor in which the CVD dia-
monds are produced and can be used to optimize these parameters. However,
the best results for thin specimens (L < 300 µm) are not much better than
those of the diamond represented in Fig. 7. At this point it seems useful to
pay attention to some interesting characteristics of the complete model that
can be seen in Figs. 7 and 8. First, the simple model gives efficiencies that
change smoothly from those at G ≈ 0 to those at G > L, where both η’s
become equal.
The complete model instead gives a cross between η+ and η− at G ≈ L,
after which both η’s tend to the same asymptotic value as well. This value is
about the mean between those at low penetrations. The figures clearly show
this effect, which is produced by the nonuniform production of charge given
by Bragg’s curve. It is essentially an edge effect due to the fact that the
Bragg’s peak moves towards the substrate edge, so that the efficiencies again
feel the influence of the different values of the λ±
c . Second, we must note that
the points at the lowest measured values of the penetration G for positive
polarization are systematically higher than foreseen by the best fitted curves.
As said below, this depends on the fact that, for crystals with a large number
of defects, though not visible in Raman spectroscopy, a non-negligible pulse
penetration depth (µm)

10 20 30 40 50 60
20
positive bias
negative bias
simple model
15 Bragg curve
defects
c
η(%)
10
5 a
b
0
0 10 20 30 40 50 60 70 80 90 100
Beam Energy (MeV)

Fig. 8. Same as Fig. 7 for a second 55µm thick diamond detector. Theoretical
−
curves: (a) simple model (10) with λ+c = 3.7 µm and λc = 0.05 µm; (b) same
model with account of Bragg’s curve; (c) complete model (13) with λ+
g = 13 µm,
λ−
g = 0.3 µm, λb = 0.33 µm and Λ = 8 µm
height defect is to be foreseen and a correction must be introduced into these

data, which can be considered only rough data.
3.1 Experimental Procedure to Measure the Charge Collection

Efficiencies
The experimental setup for bombardment with 12 C or other accelerated ions

is similar to that shown in Fig. 5 and reported in Fig. 9. A collimated beam of
accelerated ions (typically I ≈ 1–3 pnA) from the 15 MV Tandem accelerator
of the LNS in Catania impinged on a ≈ 300µg/cm2 thick gold target. The
machine voltage varied between 6.8 MV and 14 MV. These values allowed
the energy of the ions to be varied between a minimum of 18 MeV for 6 Li
to a maximum of 110 MeV for 16 O; for 12 C in particular the energy was
varied between 22 MeV and 91 MeV. Scattered ions were detected either
through a silicon detector placed at θ = −20◦ with respect to the beam or by
diamond detectors placed at +20◦ ≤ θ < +30◦ at the other side of the beam.
The calibrations were done with respect to Si detector. Target and detectors
were contained in a high-vacuum scattering chamber. Pulses from Si and
diamond detectors were amplified through conventional charge preamplifiers
followed by amplifiers, converted to digital signals and registered on disc to
be analyzed off-line afterwards. The setup used for irradiation with the 5.4
MeV α-particles from an 241 Am source was, of course, simpler in that the
Fig. 9. Experimental setup for bombardment of di-

amond detectors with ions accelerated by the Tan-
dem accelerator at LNS in Catania (Italy)
experiment was done in air and the energy of the α-particles was varied
by simply changing and carefully measuring the distance between the Am
source and the detectors, that is, using air as the absorber. The electronics
were essentially the same.
3.2 Corrections to the Efficiencies η± Needed to Into Take

Account the Pulse Height Defect of Diamond for Detectors Heavy
Ionizing Particles
Figure 10 reports the charge collection efficiencies measured for the same
diamond specimen with α-particles and with 12 C of various penetrations G
in their lower range. The efficiencies measured with α-particles are definitely
higher than those obtained with carbon at the same penetration. Moreover,
there is a difference ∆η asymp ≈ 15% between the asymptotic values of efficien-
cies for α-particles and 12 C. The best hypothesis seems that this difference is
a measure of the pulse height defect in diamond between differently ionizing
particles. As is well known [15, 16], every detector is subject to the phenom-
enon of pulse height defects. It consists essentially in the fact that when the
local charge density produced by the ionizing particle in its passage is too
high, a spatial charge forms, which lowers the local electric field. This last
becomes less effective in separating rapidly carriers of opposite signs, so that
they can recombine, with a consequent reduction of the output pulse. The
reduction of the local field depends on the angle θ between the incident parti-
cle and the applied electric field, because when the electric field is just along
the line of production of carriers, these last tend to pack together, making
recombination easier.
The pulse height defect ∆E shown by a given diamond detector for given
incident ions of energy E0 can be defined empirically as the absolute value of
the difference between the mean signal actually given by the detector hit by
40
Pos Neg
35
α 0°
60°
30 75°
α
25
Efficiency (%)
α asymptotic
20 asymptotic
12C pulse height defect
15
10 12C asymptotic
α
5 12C
0
0 5 10 15 20 25 30
Penetration depth (µm)
Fig. 10. Efficiencies at low penetrations of a 130µm thick diamond in positive (full
symbols) and negative (empty symbols) polarization for α-particles and 12 C ions at
various angles of incidence
those ions and the mean signal expected for α-particles that can create the
same charge inside the detector, that is:
Q0
∆E = |η ion − η α |, (14)
e
where Q0 can be computed from the Bethe–Bloch energy loss formula [7] or
obtained experimentally from measurements with a silicon detector, making
allowance for the different values of in (1). The value of ηα can be ob-
tained by fitting appropriate curves from (5) (using (12) and (13) in it) to
the points taken for positive and negative polarizations at low penetration
for α-particles and allowing these curves to cross just at G = L. The implicit
assumption under this definition is that the pulse height defect for light par-
ticles is negligible. A preliminary study of this phenomenon has shown [17]
that the pulse height defect ∆E defined as above can be well represented by
an empirical formula with the form of a power law in the energy E ≤ E 0 lost
inside the specimen:
∆E = AE b (1 − e−c · cosθ ), (15)
where b = 1.2 (as in silicon detectors [15]) is the same for all particles and
diamond detectors, while A and c depend on both the detector and the inci-
dent particle. A can be usefully derived from the ∆η asymp between η±,α and
η ±,ion ion at about the energy for which G = L; that of c from, e.g., η +,ion
vs. θ at constant E.
4 Conclusions
With the application of the corrections foreseen in Sect. 3.2, the model is com-
plete and is applicable to various types of detector-grade pCVD diamond, as
defined in Sect. 1.2. Using this model means: (i) bombard diamond detectors
under test in both states of polarization with α-particles in order to know
their maximum charge collection efficiencies. (ii) If the maximum available
energy of α-particles gives G > L it is enough to do measurements of η±,α ,
changing either the incident energy or the incident angle: where ∆E = 0
changing the incidence angle does not affect the ηα ’s. (iii) If α-particles can-
not penetrate the whole thickness L of the diamond, use penetrating ions from
an accelerator; measure always both rough η± , compute the pulse height de-
fect, if any, and correct the rough data as said above. At this point fitting
to the points giving the η ±,corr ’s vs. G the curves computed from (5), (12)
and (13) will give a reliable measurement of the defect distribution inside the
specimen, allowing one to compare diamonds obtained by different methods.
That is: pCVD without and with lapping from the substrate side; as-grown or
pumped CVDHG diamonds [4] (pumping improved the values of η+ increas-
ing λ+g in (13)); eventually also some new CVD diamonds grown on natural
diamond (homoepitaxial growth). It remains to explore the consequences of a
given defect distribution on the behaviour of diamond as a beam profile mon-
itor, mainly to see how it is possible, from a priori knowledge of the defect
density, to foresee the current amplification factor that one must expect.
References
[1] M. Friedl: Diamond Detectors for Ionizing Radiations, Diploma thesis, Uni-
versity of Technology, Vienna (1999) 267
[2] D. Meier, (RD42 Collaboration): Nucl. Instr. Meth. Phys. Res. A 426, 173
(1999) 268, 273
[3] D. Husson, (RD42 Collaboration): Nucl. Instr. Meth. Phys. Res. A 388, 421
(1997) 268, 273
[4] M. Marinelli, E. Milani, A. Tucciarone, G. V. Rinati: in Chap. 6 268, 269,
275, 279, 285
[5] M. Angelone, M. Pillon: EFDA Fusion Newsletter 2004/1, 8 (2004) 269
[6] M. Marinelli, E. Milani, A. Paoletti, A. Tucciarone, G. V. Rinati, S. Albergo,
V. Bellini, V. Campagna, C. Marchetta, A. Pennisi, G. Poli, R. Potenza, F. Si-
mone, L. Sperduto, C. Sutera: Diamond Relat. Mater. 10, 706 (2001) 272
[7] P. D. Group: Review of particle physics, Phys. Lett. B 592, 1 (2004) 272, 284
[8] O. Madelung, et al. (Eds.): Kristall und Festkörperphysik, Band 22, Halbleiter,
Landolt–Bornstein and Zahlenwerte und Funktionen Halbleiter (Springer,
Berlin, Heidelberg 1987) 272
[9] S. Ramo: Proc. of the I.R.E. 27, 584 (1939) 274
M. Pillon: Diamond Relat. Mater. 10, 645 (2001) 275
[11] C. Tuvé, V. Bellini, R. Potenza, C. Randieri, C. Sutera, M. Marinelli, E. Milani,

A. Paoletti, G. Pucella, A. Tucciarone, G. V. Rinati: Diamond Relat. Mater.
12, 499 (2003) 275
[12] G. Pucella: Phd thesis, Rome (2003) 275
[13] K. Hecht: Z. Phys. 77, 235 (1932) 278
[14] I. S. Gradshteyn, I. M. Ryzhik: Table of Integrals, Series and Products (Aca-
demic, New York 2000) 280
[15] E. C. Finch, M. Asghar, M. Forte, G. Siegert, J. Greif, R. Decker: Nucl. Instr.
Meth. Phys. Res. 142, 539 (1977) 283, 284
[16] M. Ogihara, Y. Nagashima, W. Galster, T. Mikumo: Nucl. Instr. Meth. Phys.
Res. A 251, 313 (1986) 283
[17] G. Pucella, et al.: private communication 284
Index
charge collection efficiency (CCE), defects in diamond, 269, 270
280–285 diamond detectors, 267–273, 277, 278
chemical vapour deposition (CVD), 269 transport in diamond, 273–279
CVD heteroepitaxial growth Hecht’s model, 275
(pCVD), 269
defects Hecht theory, 275, 278
defects distribution, 279, 280
diamond Ramo’s theorem, 274
Laser Ablation-Deposited CNx Thin Films
Enza Fazio1 , Enrico Barletta1 , Francesco Barreca1, Guglielmo Mondio1 ,

Fortunato Neri1 , and Sebastiano Trusso2
1
Dipartimento di Fisica della Materia e Tecnologie Fisiche Avanzate, Universitá
di Messina, Salita Sperone 31, I-98166 Messina, Italy
fazio@ortica.unime.it
2
CNR – Istituto per i Processi Chimico-Fisici sezione di Messina, Via La
Farina 237, I-98123, Messina, Italy
Abstract. Amorphous CNx thin films were deposited at room temperature by

pulsed laser ablation of graphite targets in a controlled atmosphere using both N2
and N2 /Ar gas mixtures. The structural and electronic properties of the samples
were investigated by means of X-ray photoelectron, reflection electron energy loss
and Raman spectroscopy. Nitrogen contents as high as 32% have been estimated.
The overall results show a progressive transformation of the C–C sp3 bonds into
sp2 hybridized C–N ones with the development of sp2 carbon clusters upon increas-
ing the gas pressure values. By means of fast photography measurements, different
regimes of the laser-induced plasma expansion dynamics have been evidenced. The
results indicate that the electronic and structural properties of the laser ablation-
deposited CNx films depend strongly on the regime under which the plasma expan-
sion takes place, not simply on the overall nitrogen content. In this respect the role
played by the gas composition and pressure are of fundamental importance.
1 Introduction
Carbon nitride materials have been the focus of considerable experimental
and theoretical attention since Liu and Cohen proposed that β-C3 N4 should
have a hardness comparable to that of diamond [1]. Subsequent calculations
have shown that other crystalline C3 N4 structures should have stability com-
parable to or greater than that of β-C3 N4 , and that many of these struc-
tures should be very hard (e.g., the cubic form). The only exception is repre-
sented by the energetically most stable form, rhombohedral C3 N4 , which has
a graphite-like structure and is expected to be quite soft [2, 3]. From the ex-
perimental point of view, most of the thin films deposited with very different
preparation techniques have been shown to have an amorphous or disordered
phase although there are several reports of crystalline phases dispersed in
the amorphous matrix [4, 5]. Nevertheless, nonstoichiometric carbon nitride
thin films have also shown promising technological applications in the fields
of optical and electronic engineering due to their peculiar mechanical, optical
and electronic properties, which can be tailored as a function of the nitrogen
content x [6]. The effect of nitrogen insertion on the structure of amorphous
carbon thin films is not yet completely and easily interpreted due to the va-
riety of different local chemical environments allowed for nitrogen atoms in
288 Enza Fazio et al.
such a structural context. Usually, five possibilities based on different nitro-

gen bonding configurations are considered: incorporation in aromatic rings,
sp with one carbon neighbour, sp2 with two or three carbon neighbours and
sp3 with three carbon neighbours. The identification of such different local
chemical environments has been generally performed by means of different
spectroscopic techniques such as X-ray photoemission spectroscopy (XPS),
reflection electron energy loss (REELS), Fourier transform infrared (FTIR)
and Raman spectroscopy [7, 8, 9, 10, 11, 12, 13, 14, 15].
Many different preparation techniques can be adopted to deposit CNx
thin films [16, 17, 18, 19]. Their properties vary considerably, depending both
on the technique adopted and the deposition parameters for a given tech-
nique. Among others, pulsed laser ablation (PLA) has been widely used due
to its particular characteristics: the production of high-energy species and
great flexibility in the control of several deposition parameters (laser fluence
and wavelength, target composition, substrate temperature, target–substrate
distance and the possibility of performing the process in an inert or a reactive
background-controlled atmosphere) [20, 21, 22].
Since the PLA technique involves high-energy species, one can expect to
observe interesting, from an application point of view, C–N stoichiometries
and structural configurations, even for materials deposited at room temper-
ature (RT). For this reason, in this work we report the studies on CNx thin
films deposited at RT by means of pulsed laser ablation of graphite targets in
a controlled nitrogen atmosphere. The effects of the gas pressure and compo-
sition were also taken into account by depositing some samples in the presence
of a N2 /Ar gas mixture. Indeed, generally, there is a heavy correlation be-
tween plasma dynamics and the structural properties of the deposited film,
thus the study of characteristics of the plasma can contribute to a better un-
derstanding and control of the deposition process itself. The dynamics of the
expansion laser-induced plasma were investigated by means of fast photogra-
phy measurements, while the stoichiometry and the structure of the samples
were obtained by means of optical and electronic spectroscopy.
2 Experiments
Carbon nitride thin films were grown in a high vacuum chamber with a resid-
ual pressure down to approximately 1 × 10−4 Pa. High-purity graphite targets
were ablated by focusing the beam of a KrF excimer laser (wavelength 248 nm,
pulse width 25 ns, repetition rate 10 Hz) onto the surface with an incident an-
gle of 45◦ , and the estimated energy density was 2 J/cm2 . The samples were
deposited onto crystalline silicon and Corning 7059 glass substrates kept at
room temperature and placed at 40 mm from the target, which is positioned
on a rotating holder in order to avoid craterization phenomena. A mass flow
controller was used to introduce high-purity nitrogen gas during the depo-
sition process, reaching partial pressure values up to 66.7 Pa. Moreover, two
Laser Ablation-Deposited CNx Thin Films 289
samples were grown in a N2 /Ar gas mixture with a total pressure of 13.3 and
66.7 Pa and a N2 /Ar partial pressure ratio of 1/9 and 1/49, respectively.
The investigation of both the stoichiometry and the structural proper-
ties of the samples were performed by means of XPS, REELS and Raman
spectroscopy. In particular, REELS and XPS measurements were carried out
using a VG ESCALAB vacuum system. REEL spectra were obtained by a
VG LEG61 electron gun in conjunction with a VG CLAM 100 hemispherical
analyzer. The primary electron energy was Ep = 2.5 keV, and the incidence
angle α was about 40◦ from the normal to the sample surface. The gun cur-
rent was about 10 µA, and the pass energy of the analyzer was set to 7.5 eV.
The acceptance angle β of the analyzer was about 3◦ . The same analyzer,
with the same pass energy, was used to measure X-ray photoelectron spec-
tra excited by the Al Kα radiation (1486.6 eV) of a conventional twin-anode
Al/Mg Kα source. To take into account any possible charging effects, a tiny
silver paint droplet has been deposited on the sample surface for referencing
the binding energy scale to the Ag 3d5/2 line. The relative concentrations of
nitrogen and carbon atoms were estimated by the areas under the N1s and
C1s peaks weighted by the relative sensitivity factors. Raman scattering mea-
surements were performed by means of an U1000 Jobin Yvon monochromator
coupled with an Olympus BX-40 microscope. The 514.5 nm line of an Ar+
laser was focused on the sample surface through the 100× objective of the
microscope. The backscattered radiation was collected by the same micro-
scope optics and dispersed by a 1000 mm double monochromator, equipped
with two holographic gratings (1800 lines/mm). The dispersed radiation was
detected by means of a LN2 -cooled CCD sensor. During the film deposition
process, the light emitted from the laser-induced plasma was collected at a
right angle with respect to the plume expansion direction by means of an
optical system. A fast intensified charge-coupled device (Andor Technology
iStar iCCD) with a variable gate (∆t ≥ 2 ns) was used to detect the optical
emissions from the plasma and then to acquire the images of the expanding
plasma. A time gate of 2 ns was adopted in the initial stage of the expansion
and then progressively increased up to 50 ns to acquire the images of the
expanding plasma at different time delays after the laser pulse arrival.

The chemical bonding states and the relative atomic contents in CNx thin
films were investigated by means of X-ray photoelectron spectroscopy and
low-energy reflection electron energy loss spectroscopy. In fact, in the case
of CNx materials, these spectroscopic techniques are thoroughly sensitive to
local coordinations of carbon and nitrogen atoms. In particular, changes in
the carbon microstructure can be well evidenced from the modifications oc-
curring in both the subband structures and the chemical shifts of the C1s
Fig. 1. Behavior of the nitrogen N/C ratio as a function of the gas pressure for
samples grown in pure nitrogen atmosphere. Line is a guide for the eye
and N1s core level photoemission peaks. Taking into account both the ex-
perimentally determined and the literature-reported sensitivity factors [23],
a compositional analysis was performed and the N/C atomic ratio x was ob-
tained from the integrated areas of the carbon and nitrogen photoelectron
peaks. In Fig. 1 is shown the compositional parameter x as a function of the
nitrogen partial pressure. The results (see also Table 1) show that, under
the ablation conditions chosen, increasing P N2 up to 13.3 Pa, the nitrogen
content reached a value close to x = 32 % and then decreased down to about
17% at P N2 = 66.7 Pa.
In Fig. 2a are shown the C1s XPS lineshapes of some of the investi-
gated samples grown at different pure nitrogen partial pressures: a clear
modification is evident. The effects of the increasing gas pressure are an
asymmetric broadening and a shifting towards higher binding energies of the
C1s peak, with respect to the position observed for pure carbon systems
(284.5 eV). These effects, also evident for the lowest nitrogen partial pressure
(P N2 = 0.1 Pa), indicate the occurrence of changes in the carbon chemical
bonding structure due to the formation of nitrogen–carbon bonds. It is not
possible to unambiguously deconvolve the C1s core level photoelectron peak
since an accurate determination of the different contributions is difficult to
give. Nevertheless, it is quite well established that the main contribution de-
tected at 284.5 eV is attributed to pure carbon (C–C bonds in a-C network),
while the asymmetry on the high binding energy side depends on the lo-
cal environment of the carbon atoms into C–N sp3 and C–N sp2 bonding
coordinations located at 286 eV and 287 eV, respectively [24].
Fig. 2. Experimental C1s and N1s XPS spectra of CNx films grown at different
nitrogen partial pressures
In Fig. 2b are shown both the experimental and the numerically fitted N1s
XPS spectra of CNx films. To check the bonding states of carbon in the sam-
ples, the N1s peak was deconvolved by using two Gauss–Lorentzian bands
positioned around 398.5 eV and 400 eV [21]. These two bands are assigned
to different nitrogen bonding states in C–N bonds and, more precisely, to
N bonded to sp3 -hybridized C and to N bonded to sp2 -hybridized C, respec-
tively. Nitrogen–carbon triple bonds have been excluded due to the absence
of their characteristic resonance at about 2200 cm−1 in the infrared trans-
mission spectra. Overall, the results of the fitting procedure show that the
400 eV N1s photoemission peak exhibits a slight, but definite, growth upon
increasing the nitrogen partial pressure.
The XPS results, by themselves, do not give a complete information about
the carbon–carbon bond reorganization, nevertheless, combining them with
REELS data, an estimate of the various C–N and C–C bonding types con-
centration can be obtained. In the following, according to the literature [25],
the short notations [C–Csp2 ], [C–Csp3 ], [N–Csp2 ] and [N–Csp3 ] have been
used to indicate the concentrations of different bonds type in a-CNx .
From the analysis of energy loss spectra [10, 12], the low-energy plasmon
Eπ is known, and a shift of the main plasmon peak towards lower energies
Table 1. Results of XPS and REELS spectra analysis: x(%) is the relative nitrogen
content with respect to carbon, [N–Csp2 ] (%) and [N–Csp3 ] (%) are the fractional
contributions to the C–N bond, Ep (eV) and Eπ (eV) are the high- and low-energy
plasmons, ρ is the calculated mass density, and sp2 and sp3 are the relative fraction
of sp2 and sp3 bonds
Sample P N2 /P A x [N–Csp2 ] [N–Csp3 ] Ep Eπ ρ sp2 sp3

# Pa % % % eV eV g cm−3 % %
22 0.0/– 0.0 0.0 0.0 26.9 5.1 2.61 50.5 49.5
40 0.1/– 2.7 1.4 1.3 27.1 5.2 2.65 53.7 46.3
37 1.3 8.6 5.1 3.5 26.7 4.9 2.57 53.8 46.2
47 1.3/12.0 9.0 6.0 2.9 – – – – –
49 1.3/65.0 3.0 1.8 1.3 21.8 5.3 1.71 75.6 24.4
42 6.5 7.3 5.1 2.2 26.2 5.2 2.46 61.5 38.5
24 12.8/– 25.0 12.0 13.0 24.0 4.2 2.05 58.8 41.2
45 13.3/– 32.0 24.4 6.7 23.7 4.4 1.99 67.7 25.6
39 17.0/– 11.9 9.2 2.7 26.1 5.0 2.44 60.6 39.4
38 26.6/– 12.2 8.5 3.7 25.5 4.8 2.32 58.0 42.0
48 66.7/– 16.0 14.1 3.5 23.4 5.4 1.96 90.1 9.9
(from 26.9 to 24.0 eV) is also evident, increasing the x values (see Table 1).
Applying the Drude formula, this trend can be related to a reduction of
the valence electron density and interpreted in terms of a reduction of mass
density ρ of the samples. Moreover, to simplify, we assume that our samples
are essentially built up by only threefold and fourfold coordinated carbon
atoms bonded to other carbon or nitrogen atoms. So, we can state:
[N–Csp2 ] + [C–Csp2 ] + [N–Csp3 ] + [C–Csp3 ] = 1, (1)
where the concentrations of N–Csp2 and N–Csp3 bonds are known from the
XPS quantitative analysis of the N1s photoemission peaks, while the total
concentration of the C–Csp2 bonds can be calculated by applying the Drude
formula to the low-energy plasmon resonance Eπ ,
Eπ2
[Csp2 ] = [C-Csp2 ] + [N-Csp2 ] = 7.25 × 1020 , (2)
sNC
where NC is the density of carbon atoms and s = 0.28 is a static screening
factor which takes into account the screening effects on the π electrons due
to the remaining σ electrons [10, 26, 27]. Thus, using the previous relations
(1) and (2), it is possible to calculate the remaining bond fraction [C–Csp3 ].
All the results (see Table 1) show the continuous growth of the sp2 bond
fraction and the contemporary decrease of the sp3 one with an increase of
the nitrogen partial pressure (see Fig. 3a). Moreover, both [N–Csp3 ] and [N–
Csp2 ] slightly increase at the same rate. This behavior indicates a progressive
material graphitization which proceeds at expense of the [C–Csp3 ] concen-
tration. It seems that nitrogen could preferentially substitute carbon atoms
Fig. 3. Behavior of the partial contributions [N–Csp2 ], [N–Csp3 ], [C–Csp2 ] and

[C–Csp3 ] as a function of the gas pressure
bonded in a fourfold coordination, inducing an increase of the graphitic is-

lands of the samples. From Fig. 3b is also evident a slow and smooth variation
of the estimated mass density ρ. The density values of our samples are not
comparable to those reported in literature for CNx samples with low nitro-
gen content [28, 29]. Nevertheless, in our opinion, the bonding configurations
present in the CNx films are strongly dependent not only on the nitrogen
content, but also on the starting sp3 /sp2 bonding ratio, which, in turn, is
determined by the samples preparation conditions.
Additional details about the change in the carbon bonding structure were
also evidenced by the evolution of the D and G bands in vibrational Raman
spectra. In fact, visible Raman spectroscopy is sensitive to the carbon sp2
phase due to the resonance with π–π ∗ transitions which occurs in sp2 do-
mains while σ–σ ∗ transitions, occurring in both sp2 and sp3 domains, are
in resonance with UV light. Then, visible Raman spectroscopy can only in-
directly probe the evolution of an sp3 phase through the evolution of the
sp2 one. In particular, the sp3 /sp2 bonding ratio can be indirectly evaluated
from the ratio between the intensity of the two contributions (the so-called
D and G bands). Indeed, the G band is connected with the relative motion of
sp2 carbon atoms, while the D peak is related to the breathing mode of the
aromatic rings [30]. The evolution of the D band is indicative in microcrys-
talline graphite of a disordering process which produces smaller and smaller
crystalline domains, while in diamondlike carbon system its appearance is re-
lated to the development of an sp2 phase with the creation of sixfold aromatic
rings.
PN =0.1 Pa PN =6.5 Pa
2 2
Intensity (arb. un.)
P =66.7 Pa
N
PN =17 Pa 2
2
1100 1300 1500 1700 1100 1300 1500 1700

Raman Shift (cm−1)
Fig. 4. Raman spectra of CNx films grown at different nitrogen partial pressures.
The dashed lines represent the fitting of the D and G peaks
The Raman spectra of some samples grown at different nitrogen partial

pressures are shown in Fig. 4. The spectra are characterized by a broad asym-
metric band near 1550 cm−1 (G band), with a shoulder at 1300–1400 cm−1
(D band), typical of diamond-like carbon systems. Raman spectra were de-
convoluted using a Breit–Wigner–Fano (BWF) lineshape for the G peak plus
an additional Gaussian line for the D peak [31]. The BWF lineshape is de-
scribed by:
[1 + 2(ω − ω0 )/QΓ ]2
I(ω) = I0 , (3)
1 + [2(ω − ω0 )/Γ ]2
where I0 is the peak intensity, ω0 is the peak position, Γ is the full width half
maximum (FWHM) and Q−1 is the BWF coupling coefficient. The maximum
of the BWF line used to account for Raman contribution at around 1300 cm−1
due to its asymmetric line-shape lies at:
Γ
ωmax = ω0 + , (4)
2Q
where the peak position ωmax is lower than the line position of the undamped
mode peaked at ω0 if Q value is negative.
The positions, the linewidths and the intensities of the G and D bands
were deduced from the fitting procedure (see Table 2). It is evident that
Fig. 5. Variation of I D /I G ratio of the Raman spectra vs. the gas pressure for
samples grown in pure nitrogen atmosphere. The line is a guide to the eye
Table 2. Results of the deconvolution procedure performed on the Raman spectra

and sp3 /sp2 bonding ratio as determined by REELS analysis
Sample P N2 /PA x I D /I G ωG γG sp3 /sp2

# Pa % cm−1 cm−1
40 0.1/– 2.7 0.21 1548 250.0 0.86
37 1.3/– 8.6 0.32 1545 230.0 0.86
47 1.3/12.0 9.0 0.32 1574 188.3 –
49 1.3/65.0 3.0 0.30 1548 187.8 0.32
42 6.5/– 7.3 0.34 1555 217.3 0.63
45 13.3/– 32.0 0.48 1563 181.7 0.38
39 17.0/– 11.9 0.49 1575 162.8 0.65
48 66.7/– 16.0 0.57 1583 144.7 0.11
with increasing nitrogen partial pressure, the D band intensity increases and
the G band position shifts towards higher frequencies, while its linewidth
decreases. In Fig. 5 the I D /I G ratio increasing behaviour is evident, even at
relatively low nitrogen gas pressures.
As reported above, the I D /I G values can be used only to indirectly esti-
mate the sp3 content. However, the values obtained from REELS and XPS
results are in good agreement with Raman ones. In fact, at the lowest nitro-
gen partial pressure, the films present a higher sp3 /sp2 bonding ratio and,
correspondingly, low I D /I G values (the structure is characterized by olefinic
groups [32]). Upon increasing the nitrogen partial pressure, the sp3 /sp2 ratio
decreases and, correspondingly, the I D /I G ratio increases (see Table 2). Such
behavior is consistent with a transformation of sp3 bonded carbon atoms into

sp2 ones, induced by the higher nitrogen partial pressure values. Clustering
of sp2 sites gives rise to ordering and aromatic ring development, which in
turn increase the D band intensity.
As can be seen from Fig. 1, the nitrogen content steadily increases with
the nitrogen pressure, but above a certain value, it drops drastically. Nev-
ertheless, the I D /I G value shows a monotonic increase with the pressure.
It is well known that in PLA experiments the plasma expansion dynam-
ics is strongly affected by the pressure and by the chemical nature of the
gas through which the expansion takes place. Thus, the observed structural
properties of the films cannot be simply dependent on the nitrogen gas pres-
sure but, more generally, on the expansion dynamics also. To investigate this
important aspect, we prepared a set of samples in a nitrogen/argon gaseous
mixture, keeping fixed the nitrogen partial pressure.
Just after the arrival of the laser pulse on the target surface, a plasma
ball develops and starts to expand along the normal to the target surface.
When such an expansion takes place in a background gas, as happens in the
deposition of CNx thin films, a shock wave can develop if some conditions
are fulfilled. Indeed a shock wave, due to the supersonic expansion of the
plasma in the gas, develops if the mass of the removed material is lower than
the mass of the gas surrounding the plasma and if the pressure driving the
moving front of the plasma is greater than the pressure of the gas at rest.
Useful information about the expansion dynamics can be obtained by means
of time-resolved fast photography of the expanding plasma. The appearance
of a bright edge at the contact front between the plasma and the surrounding
gas is, in fact, indicative of the development of the shock wave. Moreover, the
position R of the front edge of the plasma as function of the time t, after the
arrival of the laser pulse, can be obtained from these images. The expansion
of the shock wave front as a function of time is given by the following relation:
1/5
E
R = ξ0 t2/5 , (5)
ρ0
where ξ is a factor related to both geometrical and thermodynamical quanti-

ties, E is the plume energy, and ρ0 the density of the gas at rest [33]. Usually,
three expansion regimes are observed: a first stage, where interaction with
the gas weakly affects the expansion, in which the plasma expands linearly
with time, then a shock wave develops and finally the plasma expands as a
sound wave. Eventually the plume stopping can also be observed. In Fig. 6
the position of the moving front of the plasma is reported as a function of
time for different ambient gas and pressure conditions, as obtained from the
time-resolved images of the expanding plasma, not shown here. In particular,
experiments were performed at three pure nitrogen pressures (i.e., 1.3, 13.3
and 66.7 Pa) and two N2 /Ar mixtures (i.e., 13.3 and 66.7 Pa with P N2 /P A ra-
tios of 1/9 and 1/49, respectively). As can be seen from Fig. 6, at a pressure
Fig. 6. Variation of the position R of the plume front edge as a function of time.
Open symbols refer to expansion in different pure nitrogen pressure, while solid ones
to expansion in mixed N2 /Ar atmosphere
of P N2 = 1.3 Pa the expansion dynamics is weakly influenced by the back-

ground gas: initially a linear behavior holds and only at longer delay times
deviation from this behavior can be envisaged. Interaction with the gas be-
comes more and more effective as the nitrogen pressure is increased, and a
slowing down of the plume can be clearly observed. Similar results have been
obtained for the two experiments performed in the mixed N2 /Ar atmosphere.
In this case the slowing down of the plume is evident also at the total pressure
of 13.3 Pa. Argon atoms thus are more efficient in limiting the expansion of
the plasma with respect to the nitrogen molecules. When data reported in
Fig. 6 were fitted with relation (5) a good agreement could be obtained only
in a limited time range. Arnold et al. [34] proposed an analytical model that
could describe the whole expansion process in terms of the above reported
temporal stages. The transitions from a temporal stage to the following one
are also taken into account. The experimental data are expressed in terms of
the following dimensionless variables:
−1/3 −1/3
2E 2E
R̃ = R t̃ = tvs , (6)
P P
where E is the plume energy, P the pressure of the gas and vs the velocity
of the sound in the undisturbed gas. When the R − t experimental data are
transformed using relations (6), all the curves showed in Fig. 6 collapse onto
Fig. 7. Plot of the distance vs. time in dimensionless variables for pure nitrogen
atmosphere and for the nitrogen/argon gaseous mixture case (inset)
a single master curve as shown in Fig. 7. The experimental data describing

the expansion in pure nitrogen gas show two distinct slopes as a function
of t̃ : (i) a linear behavior typical of a free expansion regime and (ii) a region
where a R ∝ t−1/3 relation holds. In the inset is shown the same plot, but
for the experiments performed in the N2 /Ar mixtures. It can be clearly seen
that the shock wave regime is present for all the mixtures.
In Fig. 8 we report the XPS (Fig. 8a) and Raman (Fig. 8b) spectra of
the samples deposited in nitrogen/argon gaseous mixture. Comparing these
results with those obtained for the corresponding samples deposited at the
same pressure but in an all nitrogen atmosphere, the following effects are
evident: (i) a growth of the concentration of the N atoms bonded to sp2 -
hybridized C and (ii) a blue-shift of the G band position with a decrease
of its linewidth, while the D band intensity remains almost constant. The
relevant nitrogen content x and the I D /I G ratio are reported in the tables.
A clear correlation between the stoichiometry and the structure of the
growing films and the plasma expansion regimes emerges from these results.
In fact, concerning the samples grown in a pure nitrogen atmosphere, increas-
ing P N2 up to 13.3 Pa, the N/C content reaches a value close to 31.2% and the
I D /I G ratio increases: a progressive development of the sp2 carbon clusters
is induced. Nevertheless, the further increase of P N2 up to 66.7 Pa does not
induce an increase in nitrogen content, since the N/C ratio is less than 20%.
A further transformation of sp3 bonded carbon atoms into sp2 ones, as well
Fig. 8. XPS (a) and Raman (b) spectra of the samples deposited in nitrogen/argon
gaseous mixture
as a clustering process of the sp2 domains are observed. In such a case the
dynamics of the plasma expansion plays a crucial role. As observed from the
fast photography, the dynamics of the expanding plasma changed from a free
expansion behavior at low pressure to a shock wave formation at the higher
pressures and, in the late stage, plume stopping has also been observed. In
such a case the reduced nitrogen content in the films upon raising the pres-
sure up to 66.7 Pa is a consequence of the reduced flux of species reaching
the substrates, due to the increased collision rate between the plasma species
and the nitrogen molecules. The observed increase of the I D /I G ratio can be
understood if the reduction of the kinetic energy of the species impinging on
the substrates is considered.
Concerning the sample deposited in the nitrogen/argon atmosphere, at
13.3 Pa with 1/9 N2 /Ar, the dynamics of the expansion plasma are different
from the ones observed at the same total pressure, but in pure nitrogen. In our
case, in the mixed gaseous mixture the nitrogen content drops rapidly from
27% down to 3% as argon atoms are introduced in the chamber, even though
the total nitrogen gas content is kept constant. Thus, the reduction of the
nitrogen content has been explained considering the plasma dynamics, and
it is expected that the reduction of the kinetic energy of the species should
produce an increase of the I D /I G ratio. On the contrary, it remains constant
for both the samples with values comparable to the one estimated for sample
grown at 1.33 Pa in a pure nitrogen atmosphere. Really, it is important to
remark that the nitrogen content of the two films is very low (from 9% down
to 3%) so that the increase of sp2 carbon phase due to the nitrogen content
can be expected to be very low. Thus, the structures of the films are more
dependent on the dynamics of the plume expansion rather than on the total
pressure at which they are deposited. In this respect the nature of the gas
mixture must be taken into account.
4 Conclusion
In summary, CNx thin films have been deposited by means of pulsed laser
ablation at different N2 and N2 /Ar mixed partial pressures. In particular, the
results obtained can be summarized as follows. Concerning the samples grown
in pure nitrogen atmosphere, the nitrogen content increases upon increasing
the nitrogen partial pressure up to 13.3 Pa, but when it is raised to 66.7 Pa,
its value decreases. The nitrogen introduction in the amorphous carbon ma-
trix has been found to induce an increase in the total threefold-coordination
fraction with a contemporaneous lowering of the C–C fourfold-coordination
concentration, as evidenced by the XPS and REELS measurements.
Nevertheless, as shown by the Raman spectroscopy results, the growth
of graphitic domains has also been observed when nitrogen content in the
films decreases, as happens for the samples deposited at the highest partial
pressures. This behavior is explained in terms of the reduction of the kinetic
energy of the depositing species, i.e., it involves the plasma expansion dynam-
ics. Changes of the expansion dynamics from a linear expansion to a shock
wave formation and, finally, to the stopping of the expansion were observed.
The decrease of the nitrogen content in the films above a certain nitrogen
pressure have been related to the transition from the free expansion to the
shock wave formation transition.
Samples grown in N2 /Ar gaseous mixture show a nitrogen content as
low as that of the sample deposited in 1.33 Pa of pure nitrogen gas. Also
the trend of the structural properties, as evaluated from the I D /I G ratio,
is similar, even though, considering the total pressure at which they were
deposited, higher values should be expected. All these results reveal that
the structure of laser ablation-deposited carbon–nitrogen thin films depends
strongly on the dynamics of the expansion plasma regime, not simply on
the overall nitrogen content, and in this respect the role played by the gas
composition and pressure is of paramount importance.
References
[1] A. Y. Liu, M. L. Cohen: Phys. Rev. B 41, 10727 (1990) 287
[2] D. M. Teter, R. J. Hemley: Science 271, 53 (1996) 287
[3] S.-D. Mo, L. Ouyang, et al.: Phys. Rev. Lett. 83, 5046 (1999) 287
[4] K. M. Yu, M. L. Cohen, E. E. Haller, et al.: Phys. Rev. B 49, 5034 (1994) 287
[5] A. Y. Liu, R. M. Wentzcovitch: Phys. Rev. B 50, 10362 (1994) 287
[6] M. N. Semeria, J. Baylet, et al.: Diamond Relat. Mater. 8, 801 (1999) 287
[7] A. G. Filtzgerald, L. Jiang, M. J. Rose, et al.: Appl. Surf. Sci. 175–176, 525
(2001) 288
[8] F. L. Normand, J. Hommet, T. Szorenyi, et al.: Phys. Rev. B 64, 235416 (2001)
288
[9] S. Widemann, M. Knupfer, J. Fink, B. Kleinsorge, J. Robertson: J. Appl. Phys.
89, 3783 (2001) 288
[10] A. M. Mezzasalma, G. Mondio, F. Neri, S. Trusso: Appl. Phys. Lett. 78, 326
(2001) 288, 291, 292
[11] F. Barreca, A. M. Mezzasalma, G. Mondio, F. Neri, S. Trusso, C. Vasi: Thin
Solid Films 377–378, 631 (2000) 288
[12] F. Barreca, A. M. Mezzasalma, G. Mondio, F. Neri, S. Trusso, C. Vasi: Phys.
Rev. B 62, 16893 (2000) 288, 291
[13] N. Tsubouchi, B. Enders, A. Chayahara, A. Kinomura, C. Heck, Y. Horino: J.
Vac. Sci. Technol. A 17, 2384 (1999) 288
[14] P. Papakonstantinou, D. A. Zeze, A. Klini, J. M. Laughlin: Diamond Relat.
Mater. 10, 1109 (2001) 288
[15] E. Fazio, F. Barreca, F. Neri, S. Trusso: GNSR 2001 – State of Art and Future
Development in Raman Spectroscopy and Related Techniques (IOS, Amster-
dam 2002) pp. 61–67 288
[16] H. Sjostrom, L. Hultman, et al.: J. Vac. Sci. Technol. A 14, 56 (1996) 288
[17] B. C. Hollowaya, O. Kraft, et al.: J. Vac. Sci. Technol. A 18, 2964 (2000) 288
[18] P. Hammer, N. M. Victoria, F. Alvarez: J. Vac. Sci. Technol. A 18, 2277 (2000)
288
[19] B. Kleinsorge, A. C. Ferrari, J. Robertson, W. I. Milne: J. Appl. Phys. A 88,
1149 (2000) 288
[20] D. B. Crisey, G. K. Hubler (Eds.): Pulsed Laser Deposition of Thin Films
(Wiley, New York 1994) 288
[21] E. Riedo, F. Comin, J. Chevrier, A. M. Bonnot: J. Appl. Phys. 88, 4365 (2000)
288, 291
[22] G. M. Fuge, C. J. Rennick, et al.: Diamond Relat. Mater. 12, 1049 (2003) 288
[23] J. F. Moulder, W. F. Stickle, P. E. Sobol, K. D. Bomben: Handbook of X-Ray
Photoelectron Spectroscopy (Perkin-Elmer, Eden Prairie 1992) 290
[24] M. Tabbal, P. Merel, S. Moisa: Appl. Phys. Lett. 69, 1689 (1998) 290
[25] T. Kohler, G. Jungnickel, T. Frauenheim: Phys. Rev. B 60, 10864 (1999) 291
[26] S. Bhattacharyya, C. Valleé, et al.: J. Appl. Phys. 85, 2162 (1999) 292
[27] G. Curró, F. Neri, G. Mondio, et al.: J. Electron. Spectr. Related Phen. 72,
89 (1995) 292
[28] S. Rodil, N. A. Morrison, W. I. Milne, J. Robertson, et al.: Diamond Relat.
Mater. 9, 524 (2000) 293
[29] J. Hu, P. Yang, C. M. Lieber: Phys. Rev. B 57, 3185 (1998) 293
[30] K. W. R. Gilkes, H. S. Sands, D. N. Batchelder, J. Robertson, W. I. Milne:

Appl. Phys. Lett. 70, 1908 (1997) 293
[32] J. Schwan, S. Ulrich, V. Baton, H. Ehrhardt: J. Appl. Phys. A 80, 440 (1996)
295
[33] Y. B. Zel’dowich, Y. P. Raizer: Physics of Shock Waves and High Temperature
Hydrodynamic Phenomena (Academic, New York 1966) 296
[34] N. Arnold, N. J. Gruber, J. Heitz: Appl. Phys. A 69, S87 (1999) 297
Index
sp3 /sp2 bonding ratio, 293, 295 G band, 293, 294, 298
carbon chemical bonding, 290 plasma

carbon coordination, 291 plasma expansion, 296, 299
threefold and fourfold coordination, plasma plume
292, 296 plume stopping, 297
chemical shift, 290 pulsed laser ablation (PLA), 288, 296,
CNx thin films, 289, 290, 296, 298 299
core level photoelectron peak, 290
Raman spectroscopy, 288, 289, 293–295,
D/G band intensity ratio, 295 298–300
D/G band intensity ratio, 295, 298–300 reflection electron energy loss spec-
density, 293 troscopy (REELS), 288–292, 295,
diamond-like carbon (DLC), 294 300
diamondlike carbon (DLC), 293
D band, 293, 294, 296, 298 shock wave, 296, 298–300
fast photography, 287–289, 296, 299 X-ray photoemission spectroscopy

free expansion regime, 297 (XPS), 288, 291, 292, 295,
298–300
graphite-like carbon (GLC), 287 x-ray photoemission spectroscopy
graphitization, 292, 293, 299 (XPS), 289–291
Modeling of the Transport Properties
of Diamond Radiation Sensors
Stefano Lagomarsino and Silvio Sciortino
Dipartimento di Energetica, Università di Firenze, Via S. Marta 3, I-50139

Firenze, Italy, and INFM, Unità di Ricerca di Firenze
sciortino@fi.infn.it
Abstract. A fully quantitative model of electronic transport in polycrystalline

chemical vapour deposited (pCVD) diamond sensors is presented, predicting the
conductivity behavior of diamond devices during and after exposure to ionizing
radiation. The model takes into account a widely adopted qualitative picture of the
diamond band gap, based on two distributions of defect levels: a mid-gap group
of recombination centers and a distribution of traps closer to one of the band
edges. Analytical expressions for the radiation-induced currents (RIC) and persis-
tent radiation-induced currents (PIC) are derived from the solutions of a complete
set of rate equations, and the experimental data are well fitted by assuming the
distribution of the trap centers to be formed from the superposition of several uni-
form bands, with different cross sections, energies and concentrations. The model is
validated against experimental data from a set of diamond detectors whose charge
collection distance ranges over an order of magnitude (from 15 µm to 250 µm),
i.e., from highly defective to the state-of-the-art material. A rationale is then pro-
posed for the relationship between material quality and trap parameters, also with
regard to the changes in the material properties caused by high irradiations of fast
neutrons.
1 Models of the Polycrystalline Diamond Band Gap

The simplest model for carrier kinetics in a wide band-gap material which
can explain and reproduce the experimental data involves an active trap and
a recombination center [1]. Deeper trap centers can also be introduced which
are not active at the temperature of interest, i.e., whose activation energy is
far higher than the Boltzmann factor kT . These centers can play a role only
in the neutrality condition of the material, which is an underlying assumption
of almost all the works found in literature, and in determining the position
of the Fermi level at thermodynamic equilibrium.
This picture is obviously oversimplified in the case of crystalline diamond
with a band gap as high as 5.5 eV and a considerable variety of possible
defect centers. The effort to take into account a competition between different
trapping centers in an insulator material is due to Chen et al. [2], whilst
McKeever et al. [3] have reviewed the main models proposed to describe
thermally and optically stimulated luminescence phenomena.
304 Stefano Lagomarsino and Silvio Sciortino
Nevertheless, a two-level model allows us to introduce basic concepts

and describe processes which will be generalized in the following sections
to a more satisfying model. This generalization is based on the experimental
data collected for several years on polycrystalline chemical vapour deposited
(pCVD) diamond. In this section, we give a short review of some relevant
models presented so far, resulting in a commonly accepted qualitative de-
scription of the band-gap structure of the highest quality pCVD diamond. In
Sects. 1.4 and 1.5 we will introduce experimental facts which will be useful
to develop a mathematical model of the carrier kinetics. The assumptions
on which the model is based will be reviewed in Sect. 2. The model will be
tested on pCVD diamond devices, specially prepared as ionizing radiation
detectors. Radiation-induced conductivity will be considered in Sect. 3, and
thermal relaxation of trapped carriers in Sects. 4 and 4.1. In Sect. 4.2 persis-
tent (radiation-induced) current will be treated in detail. The quality of the
samples, their performances and the effect of neutron irradiation will be ex-
plained in terms of the obtained values of the trap and recombination center
parameters.
1.1 Band A
Donor–acceptor pair recombination (DAP) has been extensively studied in

natural and synthetic high-pressure high-temperature (HPHT) diamond [4]
as well as in thin diamond films [5]. In those works DAP is ascribed to
recombination between levels due to boron and nitrogen, i.e., the extrinsic
impurities more easily incorporated in the diamond lattice. DAP is related to
well-known luminescence broad bands [6], from 1.8 eV to about 3.2 eV. Con-
versely, a broad band peaking at 2.8 eV, commonly referred to as band A, is
believed to be originated by dislocations [5], possibly including sp2 -like struc-
ture [7]. Band A is generally prominent in the cathodoluminescence spectra in
high-quality pCVD diamond [8]. Moreover, sharp photoluminescence peaks
are less observed in higher quality samples, due to the absence of nitrogen
and other impurities [9].
As can be expected, the band A emission disappears in some single-crystal
CVD (scCVD) diamond films, also if it is observed in homoepitaxial diamond
films with nonepitaxial crystallites [7]. Manfredotti et al. [10] have suggested,
by photoconductivity (PC) and electroluminescence measurements, that the
defect levels related to band A act as a recombination center inside the dia-
mond band gap.
1.2 Trapping Centers
Many conductivity investigations [11, 12] support the existence of acceptor

states close to the valence band, probably related to grain boundaries. Gonon
et al. [12] proposed a trapping/recombination mechanism governing the ther-
mally stimulated conductivity (TSC) and UV photoconductivity. According
Modeling of the Transport Properties of Diamond Radiation Sensors 305
to their scheme, acceptor states at 1 eV from the valence band (VB), partially
filled in their initial state, are responsible for the conduction of unirradiated
samples. As electron–hole pairs are generated by UV illumination, electrons
are trapped on levels located at 1.9 eV below the conduction band (CB).
During thermal annealing carriers are released from the deep electron trap
to the CB and return to the initial electron state. A similar mechanism has
been suggested by Vittone et al. [13], relying on thermoluminescence (TL)
characterization. Diamond samples are exposed here to α-radiation and then
illuminated by visible light. Finally, the TL glow curve is recorded. Optical
fading of the TL signal starts at about 550 nm and is nearly full at 500 nm. In
order to explain this effect, the authors assume the existence of an electron
trapping band at half band gap and a hole trapping level with activation
energy in the range 0.98–1.13 eV above the VB. These traps are first satu-
rated (primed) by exposure to an α-source, then illuminated. On reaching the
wavelength of 550 nm, light photons have enough energy to release trapped
electrons, which tend to annihilate the still trapped holes. Souw et al. [14] in-
dependently propose a qualitative model with a defect band close to the VB,
and a defect band close to half band gap, negatively and positively charged,
respectively, at thermal equilibrium. Under these hypotheses they explain
their photoconductive (PC) measurements from red to over band-gap UV
and backwards. They observe that only holes contribute to the PC signal in
their samples.
Conversely, Nebel et al. [15, 16] report that nominally undoped pCVD
diamond is an n-type semiconductor, but in this case the investigated samples
yield a high concentration of nitrogen (∼ 1018 cm−3 ). They also propose a
qualitative model of amorphous carbon-like density of states present at grain
boundaries.
In conclusion, there is plenty of experimental evidence of distributions of
deep and shallow traps close to the band edge and recombination bands close
to the mid-gap. Whether the traps are close to the valence or conduction
band edge it is not well assessed. In the following we will try to give a math-
ematical treatment of the carrier kinetics independent from this choice. We
will consider hole traps close to the VB edge, but our conclusions will hold
as well for the opposite situation.
1.3 Carrier Lifetimes and Charge Collection Distance
A theoretical model on natural diamond, based on transient conductivity

measurements, has been proposed by Pan et al. [17, 18], to relate the defect
levels inside the band gap to the carrier lifetimes. They introduce a deep
donor level, acting as a recombination center for holes and electrons, which
they tentatively attribute to an impurity concentration of 20 ppm of nitrogen
in different aggregate forms. Two additional trap mechanisms for holes and
electrons are considered, by means of two decay times τ p and τ n . This model
fits very well to the experimental data, yielding the concentration, N r ∼
1018 cm3 , and cross-section values of the recombination center σ p = 1.4 ×

10−16 cm2 and σ n = 3.1 × 10−14 cm2 , for holes and electrons, respectively.
The lifetime of the carriers is on the order of hundreds of picoseconds.
Such a high density of defects is believed to limit the charge collection
distance (CCD) in natural IIa diamond to values of 20 to 50 µm [19]. The
CCD is defined as the average distance electrons and holes move apart un-
der the influence of the external electric field, before they are trapped in the
diamond material or reach the electrodes, and it is used as a major figure
of merit of pCVD diamond detectors and of the material itself. In pCVD
diamond the content of impurities can be strongly reduced with respect to
natural or HPHT synthetic diamond. Indeed, pCVD diamond from produc-
tion reactors now “regularly” exhibits 300 µm charge collection distance [20],
a value which is believed to be mainly limited by grain boundaries.
1.4 Location of the Trapping and Recombination Centers

in the Diamond Band Gap
In order to test the outlined picture of two distributions of defect levels, and
to locate their position inside the diamond band gap, we performed a set
of CCD measurements on a state-of-the-art pCVD diamond after exposure
to light [21]. We illuminated with photons of different energies a diamond
device produced by De Beers Industrial Diamonds (DEBID) for the RD42
collaboration (sample CDS88) [22]. The diamonds were metallized with a
single electrode on one side and 256 narrow strips on the other side. The
strips had a width of 25 µm and a pitch of 50 µm, which allowed the radiation
to penetrate the material. All the strips were shorted to form a single electric
contact. The sample was exposed to light starting from two different initial
states:
– (P) “pumped” state, obtained by irradiating the sample with a 10 mCi
β-radiation source to an absorbed dose of about 1 Gy. A rise of the ra-
diation-induced current of the sample occurs due to passivation of deep
traps, a phenomenon which is commonly referred to as “pumping”. No-
tably, a dose one tenth as large is high enough to obtain an increase in
CCD, measured off-line, from about 150 µm to 230 µm. This effect can
endure for months in a dark environment at room temperature.
– (D) “depumped” state, obtained by exposing the sample to a continuous
light source such as a halogen lamp. This procedure reset the CCD of the
sample to the initial value, about 150 µm.
We divided the band gap into twelve intervals ∆Ei = [Ei , Ei+1 ], and exposed
the sample to photons with energy spread ∆Ei . Then, we measured the CCD
under exposure to relativistic β-particles, which is the standard method to
test the collection efficiency for minimum ionizing particle (MIP) detection.
The measurements were performed according to the following scheme for each
energy interval ∆Ei :
Fig. 1. Measured CCD in the depumped (dashed line) and pumped state (solid
line) as a function of the photon energy of the applied luminous exposure. Model of
the trapping–detrapping mechanisms. The fat arrows indicate the main transitions
during the pumping and depumping processes. A tentative evaluation of the extent
and position of the defects bands is also indicated
1. First, we set the sample in the P state, then we expose it to photons of

energies in ∆Ei , finally we measure the CCD.
2. First, we set the sample in the D state, then we expose to photons of
energies in ∆Ei , then we measure the CCD.
The outcome of these measurements is presented in Fig. 1.
The band gap appears to be divided into four regions (I–IV in the figure):
– Starting from the D state, photons of energy in II and IV are able to
pump (at least partially) the sample while photons of energy in I and III
do not change the state of the sample.
– On the other hand, starting from the P state, photons corresponding to
region III completely depump the sample.
Notably, the energies of region I and III are complementary, in the band
gap, to those of region IV and II, respectively, in the sense that, if photons
of energy E tend to pump diamond, photons of energy (Egap − E) tend to
depump it, and vice versa. This is true except for region I, whose photons
does not affect the P state of diamond. We ascribe this fact to a small capture
cross section for these photons of the centers responsible for the P state.
These considerations lead to the conclusion that pumping/depumping
processes are ruled by the filling or emptying of two bands of levels, one for
each pair of regions: zone II and III relative to the band closer to mid-gap,
zone I and IV to the other one. Photons with pumping energies fill directly
one of the bands, leaving a free carrier (hole or electron) which is trapped
by the other one. Depumping photons, on the other hand, empty one of the
bands, and the released carriers recombine in the other. We will assume, in
the following, that the shallowest trap distribution is closer to the VB, but all
our considerations hold as well when the positions are reversed with respect
to half band gap.
In order to assure neutrality, the two bands have to be either neutral or

oppositely charged. The depumped state most likely corresponds to charged
bands, because the capture cross sections of charged centers tend to be higher
than those of the neutral ones [23].
In conclusion, there is experimental evidence supporting the picture of
two bands, a deep donor and a deep acceptor center distributions, with the
Fermi level situated somewhere in between, so that the two distributions are
charged at thermal equilibrium (depumped state).
We will refer to the deep donor distribution as “band A” because it has
a maximum density in the region where the luminescence band A is peaked.
This assumption is in agreement with the suggestion of Manfredotti et al. [10],
that this band acts as a recombination center. We also denote the deep ac-
ceptor center as “band B”.
The boundaries of the regions I–IV suggest that band A ranges from about
1.7 to 2.7 eV, where it has a maximum concentration [21], while band B is
extended up to 1.7 eV. These values of energy are confirmed by photocon-
ductive responsivity measurements performed on the same sample. The lower
limit of band B is not detected by this method, but it can be studied in detail
with thermal spectroscopy methods (TSC or TL) or with the combined analy-
sis of radiation-induced currents (RIC) [24] and persistent radiation-induced
currents (PIC) [25], developed by the authors and presented in Sects. 4.1
and 4.2.
1.5 Unipolar Conduction
As discussed before, it has been reported by several authors that the transport
properties in the pCVD diamond material are due mainly to one type of
carrier. This consideration, which is essential for the subsequent development
of the model, has to be treated in more detail. It has been found by the
RD42 CERN collaboration [26] that the collection efficiency of CVD diamond
detectors can be strongly incremented by removing the material from the
substrate side (linear model). This is due to the poorer quality of the CVD
diamond film near the substrate. As the removed thickness increases, the
charge collection distance, after reaching a maximum, begins to decrease,
being limited by the film thickness. As the thickness L tends to zero, the plot
of the CCD vs. L tends to approach a straight line crossing the origin.
If conduction were due to both types of carriers, the straight line would
have a form CCD = L (all the carriers reaching the contacts). On the contrary,
if only one type of carrier contributes, then CCD = L/2, i.e., only holes or
electrons reach the contacts, whilst the carriers of the other type are trapped
in the vicinity where the pair is created. In intermediate cases CCD ≈ L for
small L values, and its derivative approaches 1/2 for higher L values.
We fitted the data reported by the RD42 collaboration [26], as shown in
Fig. 2, making use of the expression for the CCD calculated by assuming that
the CCD is linearly increasing from the substrate to the growth side. The best
Fig. 2. Simulation of the CCD (d) vs. material thinning at different dn /d values.
A two-carrier model yields values very distant from the experimental points. Pure
single-carrier conduction is quite better. The best fit is given when the electrons
contribute for only 5% to the collected charge. Modified from a cited reference [26]
fit of the experimental data, as shown in Fig. 2, is obtained assuming that

one carrier contributes to conductivity for only 5%. Hence, in the following,
we will assume conductivity to be proportional to the hole concentration,
which is the choice anticipated in the previous sections.
2 A General Model for Transport Properties of pCVD

Diamond: Underlying Assumptions
Any theoretical modeling of transport properties of a semiconductor material
requires the solution of a complete set of rate equations for the populations of
both the defect levels and the conduction/valence bands. Unfortunately, the
presence of two distributions of levels in the diamond band gap, both with a
spread in energy and, presumably, capture cross section, can make it impos-
sible to find a simple, analytical solution for the rate equations of the system.
Nevertheless, a few assumptions listed below make the system analytically
solvable and permit us to relate the radiation induced conductivity (RIC) to
the structure of the diamond band gap.
1. The centers of band B are purely hole traps, as discussed in Sects. 1.4 and
1.5, and their capture cross sections for electrons are negligible.
2. Band B is decomposable in several (m) components with different capture
cross sections for holes (σi , with 1 ≤ i ≤ m) and distinct trap concentra-
tions per unit volume (Ni , 1 ≤ i ≤ m).
3. The capture cross sections for electrons (σnr ) and for holes (σpr ) take con-
stant values over the entire band A.
These three assumptions alone make it possible to solve analytically the rate
equations of diamond under irradiation.
A further assumption, the unipolar conduction approximation, whose va-
lidity is assessed by the linear model (Sect. 1.5), makes it possible to relate
directly the hole density in the VB to the conductivity of diamond. This
hypothesis is expressed as:
4. The capture cross section for electrons of the A centers is much higher
than that for holes: σnr σpr ≡ σr .
The validity of the inequality above is based on the assumption, discussed
in Sect. 1.4, that A centers are positively charged when empty of electrons.
Moreover, assumption 4 is strongly suggested by comparison with the re-
sults reported by Pan et al. [17, 18] on conductivity of natural IIa diamond,
discussed in Sect. 1. The authors of the cited work report two orders of mag-
nitude of difference between σnr and σpr = σr for the recombination centers.
In Sect. 3 a model of RIC is developed, based on assumptions 1–4.
Moreover, the decay of the current after irradiation (persistent radiation-
induced conductivity or PIC) is also related to the presence of several com-
ponents of band B. This phenomenon is accounted for by hypotheses 1 and 2
and by the further assumptions:
5. Each component of band B has a spread in energy, and the density of
levels per unit energy di /dE is assumed to be constant in the energy
interval of interest.
6. The energy limits of each component of band B are different and range
from E1i to E2i (i = 1, . . . , m).
7. Retrapping terms are considered to be negligible with respect to recom-
bination during thermal release of carriers from band B (first-order ki-
netics [3]).
We will sometimes refer to the slow carrier release from the traps caused
by thermal energy as thermal fading. Then, the validity of assumption 5 is
based on the fact that, in a range of 1–105 s (the range of our measured
fading times) the width of the energy interval emptied by thermal fading is
only about 5 × kT log 10, i.e., 0.25 eV. It is quite reasonable to assume that
the density of levels is constant on this interval.
In the following sections, we will develop a model of PIC based on as-
sumptioins 5–7 and also relate these hypotheses to the partial depumping of
RIC after thermal fading. In this context, assumption 7 will be justified by
an a posteriori consideration (Sect. 4.1). The picture of the diamond band
gap derived from assumptions 1–7 is represented in Fig. 3, by the schematics
on the left.
Fig. 3. Schematic representation of the observed radiation-induced current be-

haviour and of the trap states model of the pumping process, from the depumped
state (left) to the pumped state (right)
3 Conductivity Under Exposure to Ionizing Radiation

We consider here a pCVD diamond detector with an externally applied volt-
age bias and assume that it is depumped, i.e., it has been reset by irradiation
with visible light of proper wavelength [13] or heated [11]. By exposing the
depumped device to a continuous source of ionizing radiation (e.g., X- or β-
radiation), we can observe a typical rise of the induced current from an initial
level to a saturation level, with a time response dependent on the dose rate.
The increase of the induced current is believed to be due to a progressive
filling of traps that are responsible for the degradation of carrier lifetimes.
Different samples not only show a different rise time or saturation level, pre-
sumably dependent on the concentration of defects, but also a quite different
shape of the rise current, probably related to the details of the defect dis-
tributions. As an example, Fig. 4 shows a comparison among the responses
of five samples of different quality, exposed to β-radiation. The figure plots
the sensitivity of the samples as a function of absorbed dose, defined as the
induced current density divided by the product of the dose rate and the elec-
tric field. This quantity is proportional to the product of the lifetime and
the mobility of the carriers, and it is directly related to the quality of the
material. The considerable difference in sensitivity and time response of the
five devices is clearly observable.
Starting from the hypotheses introduced at the beginning of Sect. 2, we
can explain the details of the conductivity increase in terms of differences in
concentration and cross section of the components of band A and B. We
will denote with Nr and σr the concentrations and the capture cross sections
for holes of the A recombination levels, and with N1 , . . . , Nm , σ1 , . . . , σm ,
the same quantities relative to the m components of band B. The terms
{q1 (t), . . . , qm (t)} and qr (t) represent the numeric concentrations of charged
centers, i.e., the concentrations of empty centers (both A and B centers are
neutral when filled with electrons and holes, respectively). The generating
factor of the betas is denoted by g. The concentration of holes and their
Fig. 4. The sensitivity of five samples as a function of absorbed β-radiation dose.

CDS92, P13 and P16 are DEBID detectors; F56 and F58 were produced in Florence
by pulsed DC glow-discharge under similar deposition conditions. P13 and P16
were cut from the same wafer. P16 as well as F56 were irradiated at a fluence of
2 × 1015 MeV equivalent n/cm2
thermal velocity are indicated by p and vp , respectively. The rate equations

for the VB and for the components of band B are the following:
m
dp
=g−p qi (t)vp σi + (Nr − qr (t)) vp σr . (1)
dt
i=1
dqi
= −pqi (t)vp σi . (2)
dt
In pCVD diamond, even under irradiation, the free carrier concentration is
very much lower than the population of the defect centers (∼ 104 cm−3 free
carriers generated by a dose rate of 1 Gy/min against no less than 1012 cm−3
capture centers). As a consequence:
dp dqi
. (3)
dt dt
Inequality (3) justifies the “quasi-equilibrium (QE) approximation” [1], from
which we can set dp/dt = 0 in (1). Accordingly, the neutrality condition can
be written:

m
qi (t) = qr (t) + p − n ∼
= qr (t). (4)
i=1
In the single-carrier hypothesis 4 (Sect. 2, p. 310), the hole concentration p

is proportional to the current I:
I
p= , (5)
eµSE
where µ is the hole mobility, e the elementary charge, S the metallization

area and E the electric field strength. If we denote the total charge collected
by:
t
Q= I(t ) dt ,
0
we can solve each (2) in terms of Q, as follows:

vp
qi (Q) = Ni × exp − σi Q . (6)
eµSE
Substituting (4), (5) and (6) in the rate (1), and setting dp/dt = 0 (QE
approximation), we obtain for I the expression:
−1
eµSEg m
vp
I(Q) = × Nr σr + Ni (σi − σr ) exp − σi Q , (7)
vp i=1
eµSE
with the neutrality condition derived from (4) calculated at t = 0:

m
Ni = Nr . (8)
i=1
Equation (7) is a simple, analytical expression of the instantaneous cur-

rent I in terms of the overall collected charge Q. This theoretical expression
can be fitted to the measured current I, by varying the trap parameters Ni
and σi , together with Nr and σr , which are the corresponding quantities for
the recombination center.
A qualitative interpretation of the behaviour of the induced current is
shown in Fig. 3. At the beginning of the process, electrons and holes are
trapped by centers A and B , respectively. As a stationary condition is at-
tained, some A levels remain charged, while holes are mainly trapped in
neutral A levels, thus providing charged recombination states for the gen-
erated electrons. A small amount of charged B levels, necessary to satisfy
the neutrality condition, are mantained by thermal relaxation, which is not
taken into account by our model at this stage. Analysis of (7) reveals that,
if, most likely, the charged traps have a cross section for holes greater than
that of the neutral recombination center, then the induced current increases
monotonically with the absorbed dose.
Expression (7), with a suitable choice of the number m of components of
band B, accounts for the current rise of a number of pCVD diamond samples
of different qualities, grown with different deposition methods. It also allows
the interpretation of the response degradation due to high fluences of neutron
irradiation, in terms of structural modifications of bands A and B.
As an example, in Table 1 and Fig. 5 we report the results of the analysis
of RIC curves for five samples, three (CDS92, P13 and P16) grown by DEBID,
Fig. 5. Schematic plot of the trap parameters of the samples under study, deduced
from the analysis of the radiation-induced currents
presumably by microwave plasma-enhanced CVD, two (F56 and F58) grown

by the authors at the University of Florence, by pulsed DC glow-discharge
CVD [27]. The deposition conditions of F56 and F58 were very similar. Two
of the samples (P16 and F56) were irradiated with fast neutrons at a fluence
of 2 × 1015 /cm2 (1 MeV equivalent). Samples P13 and P16 were cut from
the same wafer, the difference between them being most likely due to the
irradiation of P16.
Table 1. Values of CCD, concentration and capture cross section of the

recombination centers, obtained for the samples under study
Manufacturer Sample CCD (µm) σr (10−16 cm2 ) Nr (1015 cm−3 )

DEBID CDS92 245 5.5 2.3
P13 140 5 5
P16 851 5 1 5 1
Univ. of Florence F58 14 0.6 18
F56 192 25 43
1
Measured after neutron irradiation
2
Measured before neutron irradiation
Figure 6 shows the best fit of the RIC curve of sample P16. The measure-
ments fit very well to the theoretical expression, with a χ2 on the order of
some units. From Table 1 and Fig. 5 it is possible to recognize some interest-
ing features of pCVD diamond. As far as the commercial (DEBID) samples
are concerned:
Fig. 6. Best fit of the model to the radiation-induced current of sample P16
(a) The capture cross section of the mid-gap A centers has about the same
value for all samples, i.e., there is a substantial identity of this type of
center over all the DEBID diamonds investigated.
(b) On the other hand, in the more recent sample (CDS92) the concentration
is about half that of the others (P# samples). Interestingly, the CCD is
about doubled passing from the P# batch to sample CDS92.
(c) Neutron irradiation affects neither the cross section nor the concentration
of the recombination states (as can be seen by comparing the parameter
values of P13 and P16). Thus, the variation in time response is likely
related to a change in structure of the more shallow band B.
(d) Indeed, neutron irradiation produces a slight change in cross-section and
concentration of the low cross section tail of band B.
Concerning the local samples:
(e) The capture cross section of the recombination levels is very different
from that of the DEBID samples, and it is strongly affected by neutron
damage;
(f) Neutron damage also produces a strong variation in the concentration of
both recombination and trap states.
With our analysis, we succeeded in separating band B in several components
of different cross sections. It is intrinsic to the method that no information
can be drawn about the activation energy of the levels.
In particular, we are not able to determine the position of the several com-
ponents of band B in the band gap nor can we evaluate the density of levels
per unit energy of the various components. This problem, left unresolved in
this section, will be answered in Sects. 4.1 and 4.2 by means of independent
measurements, interpreted in the framework of the same model.
4 Thermal Relaxation of Uniform Trap Level

Distributions
As the radiation source is removed, each trap level releases carriers with a
constant probability per unit time:

1 E
= s × exp − , (9)
τ kT
where E is measured by taking the VB edge as a reference, τ is the mean
life of the level and s is a frequency factor [1] equal to the product of the
effective density of states Nv , the thermal velocity vp and the capture cross
section of the trap:
s = Nv σvp . (10)
In the first-order approximation (assumption 7, p. 310), i.e., if the charge

released has much more probability to recombine in the band A rather than
to be trapped again, then the probability to find a single trap of band B
occupied after a time t from the removal of the radiation source is:

P (E, t) = exp −Nv σvp e−E/kT t . (11)
Now, let us adopt the hypotheses 5 and 6 of Sect. 2 (p. 310), stating the
presence of an almost uniform distribution of trap levels in the interval E1 <
E < E2 with a density of levels per unit energy given by d/ dE. In this case,
the rate of change of the distribution population is straightforwardly:

dq d d E2 d exp − τt2 − exp − τt1
=− P (E, t) dE = kT , (12)
dt dE dt E1 dE t
where:

1 E1 1 E2
τ1 = × exp , τ2 = × exp . (13)
s kT s kT
If E2 − E1 kT , well inside the interval [τ1 , τ2 ] the function (12) can be
approximated as:
dq d kT
≈ . (14)
dt dE t
Thus, the total charge q(t) of the distribution depends on time as:
d
q(t) = C + kT log t. (15)
dE
Equations (12) and (15) allow the interpretation of two different kinds of
measurements:
-10
7 10
-10
5 10
I (A)
-10
3 10
1 10 -10
0
0 2 4 6 8 10 12 14
t (10 s)
3
Fig. 7. Radiation-induced current after different (exponentially increasing) fading

times. The sample is fully depumped at the start of the measurement run
– RIC measured on a sample after different time intervals between consec-

utive irradiations (Sect. 4.1)
– PIC after removal of the radiation source, measured over five decades in
time (Sect. 4.2)
In the following sections we describe in detail the information that can be
extracted by these type of measurements.
4.1 Radiation Induced Conductivity Transient at Different Fading

Times
If we remove the ionizing radiation source after the RIC has attained its
stationary value, and then expose the sample to the same radiation source
after a variable time delay, the current does not reach immediately the stable
current level. A partial emptying of the band B traps, i.e., a thermal fading,
occurs, such that the radiation takes a certain time to fill the traps again,
dependent on the fading time between exposures. In this case, we say that the
sample is partially depumped ; then, the longer is the fading time, the higher is
the time necessary to reach the RIC saturation value, i.e., the pumped state.
Figure 7 shows the induced current behavior during a single measurement
run, under consecutive exposures to the same level of β-irradiation. The first
current step represents the response of the sample in the fully depumped state,
the following exposures alternate between exponentially increasing fading
times (0.33 min, 1 min, 3 min, and so on, up to 12 h). The stationary levels
are the same within the limits of reproducibility of the dose rate. In Fig. 8 we
report the conductivity dependence on the absorbed dose for several values
of fading time. It is worthy of note that the spacing between curves is equal.
Since the measurements were taken at exponentially increasing fading times,
this fact suggests a logarithmic dependence on time of the trap population
levels, as assessed by (15). As a matter of fact, if we perform the analysis
Fig. 8. Radiation-induced current vs. dose behaviour at different fading times
Fig. 9. Evolution with thermal fading of the numeric concentration of empty

(charged) traps qi for two trap components of different cross section
of the RIC at each fading time, in order to determine the dependence of the
population of the traps on time, we obtain the behavior shown in Fig. 9. The
concentration of empty traps tends to follow a logarithmic behavior of type:
q(t) = A + B log t. (16)
A comparison of (16) with the theoretical expression (15), allows us to

evaluate the density of levels per unit energy by simply dividing the slope of
the curves of Fig. 9 by the term kT . Figure 10 shows the results of this analysis
performed on three samples for the traps whose uncertainties in concentration
is small enough to recognize a definite time behavior. A comparison of the
overall trap concentration with the density of levels per unit energy shows
that they are consistent with level distributions whose width ranges over a
few eV.
Fig. 10. Comparison between concentration of defects (cm−3 ) and density of defect
states per unit energy (cm−3 eV−1 ) for three samples under study: F56, F58, P16
The analysis of RIC at gradually increasing fading times allows us to de-

duce the density of levels per unit energy, but does not yield any information
about the effective position of the level distribution in the band gap. The
study of the currents immediately after the removal of the radiation source
overcomes this difficulty, and gives a complementary method to evaluate the
density d/dE, also providing a consistency test of our model.
At this stage, we can verify “a posteriori” the first-order kinetics assump-
tion 7, introduced in Sect. 2. In order to demonstrate that recombination
play a prominent role with respect to retrapping, it suffices to verify the
statement:

qi σi (Nr − qr ) σr . (17)
i
This can be easily done by making use of the data of Table 1 and Fig. 9.
It appears that the inequality (17) is satisfied in the whole range of fading
times under consideration. In fact, in the worst case, after 15 h of thermal
fading, the left side of (17) is one six the value of the recombination term.
4.2 Persistent Radiation-Induced Conductivity
After removing the radiation source, the electric current approaches the initial
value prior to any irradiation. The time involved in reaching the original
(depumped) state depends on how fast the populations of the defect levels
in the band gap return to the thermal equilibrium value, and it is quite long
for polycrystalline CVD diamond. As an example, Fig. 11 reports the time
Fig. 11. Persistent radiation-induced current of each sample as a function of time.

The curves are normalized to the current level under irradiation I0
evolution of the ratio of the current levels after and immediately before the
removal of the radiation source, for the samples under study. A very slow
evolution towards a stationary level is hardly observed, in a log–log scale,
even after two days of thermal fading at room temperature. It goes without
saying that the shape of each relaxation curve is related to the structure of the
defect levels. In the following, we propose a method of trap characterization
of CVD diamond sensors based on the analysis of this time evolution of
electrical current at constant temperature (PIC).
The analysis of electrical currents induced by light photons after switching
off the light source (persistent photocurrent, PPC), has been studied [15, 16]
by means of a fitting function called stretched exponential, previously intro-
duced to describe effects related to disordered phases in solid state physics.
The stretched exponential, in the form
γ
t
i = i0 exp − , (18)
τ
has been tentatively related to the presence of continuous distributions of

defect levels.
Chen and Leung [28] fitted with a stretched exponential some numeric
simulations of a luminescence signal obtained in the framework of a two-level
model of the band gap, but did not obtain a correlation between relevant
trap parameters, such as energy or concentration or capture cross section,
and the parameters of the phenomenological function.
In our case, the stretched exponential appears to be a quite good approx-
imation both of experimental data and of the simulations over one or at most
two decades in time variation, but we verified that it does not fit PIC data
over longer periods of time. On the other hand, expression (12), giving the
rate of change of the charge stored in a continuous level distribution, permits
to study the PIC in terms of a sum of “limited hyperbolic” functions of the
type:
1 − e− τ
t
i = i0 τ , (19)
t
whose parameters, i0 and τ , are directly related to the characteristics of a
continuous distribution of trap levels: capture cross section, spread in energy
of the trap and density of levels per unit energy.
The expression of PIC as a sum of limited hyperbolic functions follows
from (12). In fact, assuming the QE approximation, the rate of change of
the overall trap population is equal to the recombination rate, i.e., if qi is
the concentration of the empty traps of the distribution i and p the hole
concentration in the VB, then:

m
dqi
= pvp σr Nr , (20)
i=1
dt
where vp is the thermal velocity of holes, σr and Nr are the capture cross
section and the concentration of the recombination center, respectively. By
expressing (20) in terms of Nr instead of the population of the recombination
centers (Nr − qr ), we implicitly assume that during thermal relaxation of
band B the recombination centers are almost entirely occupied by electrons.
This implies, from (4), that during the whole process also the number density
of free traps is very small compared to the overall trap population. During
the previous exposure to the ionizing radiation, as the stationary state was
reached, the generation rate g had become equal to the recombination rate,
and therefore proportional to the population p0 of the valence band in the
steady state:
g = p0 vp σr Nr . (21)
Equations (20) and (21) can be used to find a relationship between the PIC
(I(t)) and the steady current I0 just before the removal of radiation source.
Since the ratio I(t)/I0 is equal to the ratio p(t)/p0 of the hole concentrations,
then:
I0 dqi
m
I= . (22)
g i=1 dt
Hence, from [12]:

t t
di exp − τ2i − exp − τ1i
m
I0
I = kT . (23)
g i=1
dE t
Fig. 12. Best fit of the persistent current for sample CDS92 to a function composed
of three limited hyperbolic components (19). The PIC slowly relaxes to the initial
dark current level
If we assume that the upper limit of each trap distribution is too deep in the
band gap to be affected by thermal fading, setting:

1 E2
τ2 = × exp → ∞, (24)
s kT
we find:
t
di 1 − exp − τ1i
m
I0
I = kT , (25)
g i=1
dE t
In this way, we express the PIC as a sum of limited hyperbolic terms, from
which we can determine di / dE and, once the capture cross section σi is
known from the analysis of the RIC (Sect. 3), we can determine E1i too.
As an example, in Fig. 12 we report the fit of the current curve of sample
CDS92, whose behaviour has been well described by means of three com-
ponents, each one corresponding to the traps represented in Fig. 5. Fig-
ures 13 and 14 complete the information of Table 1 ad Fig. 5 reporting the
relationhips between d/ dE, σ and E1 for each sample under study.
In conclusion, the analysis of the results of Sect. 4 yields the following
conclusions. Concerning the commercial (DEBID) samples:
(a) All samples have a very similar structure of band B, with a cross section
decreasing as the level position becomes deeper and deeper in the band
gap, starting from minimum distance above the VB of about 0.77 eV.
(b) The sample with the highest CCD (CDS92) has the lowest trap concen-
tration.
Band lower energy E1 (eV) Modeling of the Transport Properties of Diamond Radiation Sensors 323
1.00
0.95
F58 F56 (n-damaged)
0.90
0.85
P16 (n-dam.)
0.80
CDS92 P13
0.75 -16
10 10 -15 10 -14 10 -13 10 -12 10 -11
cross section (cm 2)
Fig. 13. Dependence of the lower energy limit of each trap distribution as a
function of the capture cross section
(c) Neutron irradiation causes an increase of the trap densities, without ap-
preciably affecting the cross sections.
Concerning the local samples:
(d) Band B begins at about 0.82 eV from the VB, and has a very large spread
in cross section, compared to that of DEBID samples.
(e) After neutron damage the energies shift upward about 0.1 eV, together
with an increase of the trap density.
5 Conclusions
We have briefly reviewed some relevant works on modeling the diamond band
gap, aimed to correlate the defect levels with the electric transport proper-
ties of pCVD diamond devices. From the present literature a qualitative or
semiquantitative description can be extracted which we have used as the
starting point for a quantitative model of the trapping and recombination
processes. We have introduced two defect bands: a trapping center, with dif-
ferent components, each with distinct trap parameters, and a recombination
center, described by a single value of capture cross section and concentration.
The number of components of the trapping centers increases with increasing
numbers of defects in the sample. We have chosen as the figure of merit of
material quality the charge collection distance, since the diamond devices we
have tested are specially prepared for ionizing radiation detectors. Our cal-
culated expression for radiation-induced current, thermal relaxation between
Fig. 14. Dependence of the trap density per unit volume and unit energy of each
trap distribution as a function of the capture cross section
irradiations, persistent current after removal of the generation source, fits

very well to our experimental data. This has allowed us to determine defect
concentrations and capture cross sections for both the recombination and
trapping centers. The capture cross section of the recombination center in
the state-of-the-art diamonds is in good agreement with the one reported for
natural IIa diamond, suggesting that the dramatic improvement in the per-
formances of the diamond devices during the last eight years is more due to a
decreasing in the concentration of defects than to a change in their structure.
This has been confirmed by the results obtained on medium to high-quality
pCVD diamonds where an increase of a factor two of the CCD (from about
100 µm to more than 200 µm) corresponds to a decrease of the same factor
of the recombination and trapping centers.
Neutron irradiation (∼ 1015 /cm2 ) seems to have a very limited influence
on traps in medium quality samples whilst has a dramatic influence on low
quality samples. This is consistent with the limited decrease of the CCD
(about 30%) measured for the medium-quality samples and the reported
strong improvement of the dynamic response in low-quality samples [29]. We
have also been able to calculate the position in energy of the lower edge
of the different components of the trap center. Our analysis is limited to a
narrow interval in energy, nevertheless this is the range of energy of the traps
which are active at room temperature. A further investigation is required of
the shallower traps, i.e., the traps within 0.75 eV from the band edge, which
cannot be filled at room temperature but can, however, limit the response of
the material in transient conductivity experiments or single-particle events,
like in the case of charge collection measurements.
Finally, homoepitaxial CVD diamond can overcome these difficulties since
there are no grain boundaries or structural defects originating at the mid-gap
centers in ideally single-crystal diamond (sCVD). However, the reproducibil-

ity of defect-free sCVD material does not seem within reach in the short
term [30], and its radiation hardness is at present a subject of investigation.
Moreover, further improvements in the pCVD diamond devices are still to be
expected [30]. Therefore, defect modelization in CVD diamond will be most
likely useful in the future.
Acknowledgements
The authors are deeply indepted to Stefano Mersi (INFN, Florence) who
helped to perform the experiments as a diploma student and Ph.D. student,
to Prof. Emilio Borchi (University of Florence) for stimulating discussions, to
Dr. Fred Hartjes (NIKHEF, Amsterdam) and Prof. Raffaello D’Alessandro
(INFN, Florence) for their support with the experimental setup. We also
wish to thank the RD42 CERN collaboration for lending us their samples.
The neutron irradiation of sample F56 has been performed in the framework
of the experiment CONRAD, of the Italian National Institute for Nuclear
Physics (INFN).
References
[1] A. Lewandowski, S. S. McKeever: Phys. Rev. B 43, 8163 (1991) 303, 312, 316
[2] R. Chen, A. Lewandowski, S. Durrani: Phys. Rev. B 24, 4931 (1981) 303
[3] S. S. McKeever, R. Chen: Radiat. Meas. B 27, 625 (1998) 303, 310
[4] P. Klein, M. Crossfield, J. A. Freitas Jr., A. Collins: Phys. Rev. B 51, 9634
(1995) 304
[5] J. Ruan, K. Kobashi, W. Choyke: Appl. Phys. Lett. 60, 3138 (1992) 304
[6] L. Robins, L. Cook, E. Farabaugh, A. Feldman: Phys. Rev. B 39, 13367 (1989)
304
[7] D. Takeuchi, H. Watanabe, S. Yamanaka, H. Okushi, H. Sawada, H. Ichinose,
T. Sekiguchi, K. Kajimura: Phys. Rev. B 63, 245328 (2001) 304
[8] T. Sharda, A. Sikder, D. Misra, A. Collins, S. Bhargava, H. Bist, P. Veluchamy,
H. Minoura, D. Kabiraj, D. Awasthi, P. Selvam: Diamond Relat. Mater. 7, 250
(1998) 304
[9] K. Iakoubovskii, G. Adriaenssens: Phys. Rev. B 61, 10174 (2000) 304
[10] C. Manfredotti, F. Wang, P. Polesello, E. Vittone, F. Fizzotti, A. Scacco: Appl.
Phys. Lett. 67, 3376 (1995) 304, 308
[11] P. Gonon, A. Deneuville, E. Gheeraert: J. Appl. Phys. 78, 6633 (1995) 304,
311
[12] P. Gonon, S. Prawer, D. Jamieson: Appl. Phys. Lett. 68, 1238 (1996) 304,
321
[13] E. Vittone, C. Manfredotti, F. Fizzotti, A. L. Giudice, P. Polesello,
V. Ralchenko: Diamond Relat. Mater. 8, 1234 (1999) 305, 311
[14] E. K. Souw, R. Meilunas, C. Szeles, N. Ravindra, F. M. Tong: Diamond Relat.
Mater. 6, 1157 (1997) 305
[15] C. Nebel, A. Waltenspiel, M. Stutzmann, M. Paul, L. Schäfer: Diamond Relat.

Mater. 9, 404 (2000) 305, 320
[16] C. Nebel: Semicond. Sci. Technol. 18, S1 (2003) 305, 320
[17] L. Pan, D. Kania, P. Pianetta, O. Landen: Appl. Phys. Lett. 57, 623 (1990)
305, 310
[18] L. Pan, D. Kania, P. Pianetta, J. W. Ager III, M. Landstrass, S. Han: J. Appl.
Phys. Lett. 73, 2888 (2001) 305, 310
[19] D. Kania: Diamond radiation detectors, in A. Paoletti, A. Tucciarone (Eds.):
The Physics of Diamond (IOS, Amsterdam 1997) pp. 555–564 306
[20] H. Kagan: Nucl. Instr. and Meth. A 541, 221 (2005) 306
[21] M. Bruzzi, F. Hartjes, S. Lagomarsino, D. Menichelli, S. Mersi, S. Miglio,
M. Scaringella, S. Sciortino: Phys. Stat. Sol. A 199, 138 (2003) 306, 308
[22] The RD42 collaboration, IEEE Trns. Nucl. Sci. 49, 1857 (2002) 306
[23] A. Rose: RCA Rev. 12, 362 (1951) 308
[24] E. Borchi, S. Lagomarsino, S. Mersi, S. Sciortino: Phys. Rev. B 71, 104103
(2005) 308
[25] S. Lagomarsino, S. Sciortino, E. Borchi: Model of persistent radiation induced
current in CVD diamond detectors, Phys. Rev. B submitted 308
[26] The RD42 collaboration, Nucl. Instr. and Meth. A 514, 79 (2003) 308, 309
[27] S. Sciortino, S. Lagomarsino, F. Pieralli, E. Borchi, E. Galvanetto: Diamond
Relat. Mater. 11, 573 (2002) 314
[28] R. Chen, P. Leung: Radiat. Meas. 37, 519 (2003) 320
[29] M. Bruzzi, D. Menichelli, S. Pini, M. Bucciolini, J. Mólnar, A. Fenyvesi: Appl.
Phys. Lett. 81, 298 (2002) 324
[30] H. Kagan for the RD42 collaboration: Diamond (Radiation) Detectors Are
Forever, IWORID 2004, Glasgow, UK (2004) 325
Index
acceptor band B, 308–311, 313, 321–323
acceptor states, 304 neutron irradiation, 313, 315, 324
donor acceptor pair recombination, donor
304 deep acceptor centers, 308
deep donor level, 305, 308
cathodoluminescence (CL), 304 donor acceptor pair recombination,
charge collection distance (CCD), 304
306–309, 315
current electroluminescence, 304
persistent, 304, 308, 317, 319–324
radiation induced, 304, 306, 308, 311, fading
313 optical fading, 305
sensitivity, 311 thermal fading, 310, 317, 319
first order kinetics, 310, 316, 319
defects frequency, 316
in diamond
band A, 304 limited hyperbolic function, 321
band A, 308, 310, 311, 315, 316 linear model, 308, 310
photoconductivity, 304, 305, 308 trap

density per unit energy, 310, 315
quasi equilibrium (QE) approximation, depumping, 307, 310
321 pumping, 306, 307, 311
quasi-equilibrium (QE) approximation, trap filling, 307, 311
312
rate equations, 303, 309, 312, 313 unipolar conduction, 308–310
Nucleation Process of CVD Diamond
on Molybdenum Substrates
Giuliana Faggio1, Maria G. Donato1 , Stefano Lagomarsino2,

Giacomo Messina1 , Saveria Santangelo1, and Silvio Sciortino2
1
INFM, Dipartimento di Meccanica e Materiali, Università “Mediterranea” di
Reggio Calabria, Località Feo di Vito, I-89060 Reggio Calabria, Italy
2
INFN, Dipartimento di Energetica, Università di Firenze, Via S. Marta 3,
I-50139 Firenze, Italy
faggio@ing.unirc.it
Abstract. Nucleation is a critical step of diamond growth by chemical vapour

deposition (CVD), and its control is essential for optimising the properties of the
synthesised material. In this work, diamond samples have been grown by pulsed dc
glow discharge CVD onto molybdenum (Mo) substrates at different temperatures
using a CH4 –H2 gas mixture at different methane concentrations. A thorough Ra-
man and photoluminescence (PL) analysis has been carried out on a set of samples,
whose deposition process has been stopped before the grain coalescence in order to
monitor the quality of the material during the growth. Moreover, an optical charac-
terisation of continuous free-standing films has been performed, aimed at assessing
the influence of growth and nucleation parameters on the global quality of the films.
A complementary statistical study of the nucleation process has evidenced a strong
dependence of the nucleation density on CH4 concentration. A correlation between
optical characterisation results and statistical analysis has been pointed out.
1 Introduction
The success in growing synthetic diamond films using chemical vapour deposi-
tion has stimulated enormous interest in diamond, not only as subject of basic
research, but also as material for advanced technological applications. Nowa-
days, the CVD technique enables the deposition of single-crystal films on
diamond substrates (homoepitaxial growth [1]), and the growth of polycrys-
talline sample, of quality comparable with the best natural diamonds [2, 3] on
more commonly available nondiamond substrates (heteroepitaxial growth).
However, the optimisation of the nucleation process and of the early growth
stages is still a developing subject [4]. It is well known that the nucleation
process affects the quality of the deposited films in terms of roughness, struc-
tural defect density and nondiamond phase inclusion. Thus, it is essential
to deeply understand the nucleation mechanism and its consequences on the
film quality [5, 6].
Polycrystalline diamond films have been deposited on various nondiamond
substrates, including insulators, semiconductors and metals [4]. Diamond nu-
cleation on nondiamond substrates occurs mostly on an intermediate layer
330 Giuliana Faggio et al.
of diamond-like amorphous carbon, metal carbide or graphite formed at the

substrate interface. Such intermediate layers provide nucleation sites for dia-
mond growth enhancing diamond nucleation density and offer an opportunity
for controlling morphology, orientation and texture of diamond films [4, 7].
Most of heteroepitaxial diamond films are grown on single-crystal sili-
con (Si) wafers. The chief reason for the wide diffusion of the heteroepitaxial
growth of diamond on silicon is that, among carbide-forming substrates, Si
wafers are easily available and extensively used in electronics industry. How-
ever, one of the major problems associated with the growth of CVD diamond
on Si or metal substrates is the low nucleation density in the absence of
substrate pretreatments [8]. Thus, a scratching procedure of Si substrates
with diamond powder is generally carried out for enhancing the nucleation
density [9, 10].
Diamond growth on metal substrates is also of interest and exhibits
appealing features [4, 11]. Unlike Si wafers, molybdenum substrates offer
the possibility to attain diamond nucleation without any surface pretreat-
ment [12, 13], or to use CVD reactors operating at very high power, which
results in considerably higher growth rates. Diamond nucleation on Mo sub-
strates occurs mostly on an intermediate layer of metal carbide (Mo2 C) [4].
The nucleation density obtained is about one order of magnitude higher than
on all other carbide-forming substrates under the same deposition condi-
tions [4, 14].
In this work, the nucleation process of diamond samples grown onto
molybdenum substrates by pulsed glow-discharge CVD will be studied by
Raman scattering and photoluminescence spectroscopy at low temperature.
The dependence of the sample quality on the most important deposition
parameters will be discussed. The comparative analysis of the indications
emerging from the optical characterisation of the films and from a statistical
study of grain size distribution will be carried out.
2 Experimental
Polycrystalline CVD diamond films (samples A, B, C) have been deposited
onto Mo substrates by a pulsed direct current (dc) glow discharge CVD,
thoroughly described in a previous work [12], using a CH4 –H2 gas mix-
ture. No substrate pretreatment has been carried out apart from polishing
with 1000 SiC paper before deposition. Aiming at clarifying the effect of the
methane concentration on the sample quality, the content of CH4 in H2 has
been varied from 2% up to 3%. In order to investigate the nucleation process,
the growth process of the three samples has been stopped before grain coa-
lescence. Total gas pressure p, substrate temperature T s , deposition time of
sample A, B and C are shown in Table 1.
The surface morphology of the samples has been investigated by scan-
ning electron microscopy (SEM). Raman measurements have been carried
Nucleation of CVD Diamond on Mo Substrates 331
Table 1. Deposition conditions and Raman parameters for samples A, B and C
Samples Deposition parameters Raman parameters

[CH4 ] p Ts Dep. time XP Γ
% mBar ◦ C h cm−1 cm− 1
A 3 315 1030 0.4 1335.1 10.8
B 3 320 1000 1 1334.5 8.3
C 2 320 1040 7 1332.8 4.2
out using the 514.5 nm line of an Ar+ ion laser. The scattered light has been
dispersed by an Instrument S.A. Ramanor U1000 double monochromator
equipped with a microscope (Olympux BX40) for micro-Raman sampling.
In such a configuration, the laser spot is focused, by a X100 objective, to a
diameter of about 1 µm. In addition, using the microscope confocal optics, a
nominal focus depth of 2 µm is obtained. Photoluminescence measurements
have been carried out with the same experimental setup at room and liquid
nitrogen temperature, using a Linkam freezing stage. Because of the stage
size, a long focal distance X50 objective has been used.

3.1 Raman and Photoluminescence Analysis
Optical techniques, such as Raman spectroscopy and photoluminescence (PL)

are commonly used for a careful and nondestructive analysis of the quality of
diamond samples. Raman spectroscopy represents the most useful diagnostic
technique for characterising diamond, because it allows a ready and accurate
identification of the different carbon allotropes present in the analysed sam-
ple. Therefore, Raman scattering studies are widely used to evaluate phase
purity and crystalline quality of synthetic diamond samples. Position XP and
full width at half maximum Γ of the diamond Raman peak give an estimation
of the crystalline quality of the synthesised material. An inhomogeneous dis-
tribution of stress, in fact, shift and broaden the diamond peak with respect
to natural diamond values (XP = 1332 cm−1 , Γ = 2 cm−1 ) [15, 16]. More-
over, amorphous carbon (a-C) phases, commonly found in polycrystalline
CVD films, can be easily identified due to their characteristic band at about
1500 cm−1 .
PL spectroscopy is very useful in the study of impurities incorporated in
synthetic diamond during the growth process. Many impurities, in fact, form
optical centres having characteristic vibronic structures in PL spectra. The
detection of a zero-phonon line (ZPL) and vibronic sidebands in PL spectra
allows an unambiguous identification of the optical centre.
Fig. 1. Micro-Raman diamond lines

measured at diamond crystallites of the
samples A, B and C. The peak position
XP and the full width at half maximum
Γ of each line have been indicated
3.1.1 Discontinuous Samples
In order to clarify the role played by the deposition parameters in determining

the quality of the samples grown on molybdenum, a careful micro-Raman
analysis of the diamond crystallites of the samples A, B and C has been
carried out. The average values XP of the position and of the linewidth Γ
of the diamond peaks are reported in Table 1. Some micro-Raman peaks of
the three samples are shown in Fig. 1.
As a general trend, an upshift of the position of the diamond Raman
lines is observed with respect to natural diamond (1332 cm−1 ), indicating
the existence of compressive stress, due to a thermal expansion coefficient
mismatch between film and substrate [17, 18]. Diamond crystallites of better
crystalline quality are obtained at longer deposition times, as witnessed by
narrower and less shifted Raman lines. In addition, sample C, grown at lower
CH4 concentration, shows lower shift than samples A and B, indicating a
lower stress level and, consequently, an improved crystalline quality. Owing
to the restricted temperature range investigated, a clear correlation between
substrate temperature and film quality cannot be assessed.
Additional information comes out from the analysis of the micro-Raman
spectra (Fig. 2) measured on a wider range (800–1800 cm−1 ). The character-
Fig. 2. Micro-Raman spectra of samples A and B grown at 3% CH4 concentration
istic diamond Raman line and two broad bands at about 1350 and 1550 cm−1 ,
associated to sp2 -bonded carbon phases [19, 20, 21], are visible in sample A,
whereas in sample B only the diamond line is observed together with a weakly
increasing photoluminescence background. Samples A and B, grown at 3%
CH4 concentration and at different deposition times, show diamond crystal-
lites of different quality. The SEM analysis of the growth surface of samples
A, B and C showed diamond crystallites of different average size (Fig. 3).
The deposition process of sample B, in fact, has been stopped after 1 h and
its diamond crystallites have grown longer than those of sample A. The total
area of grain boundaries, compared to the volume of diamond crystallites, is
correspondingly lower. As defects and nondiamond carbon phases are chiefly
accumulated at grain boundaries [3, 22, 23, 24], in sample B the signal origi-
nating from grain boundary regions becomes more hardly detectable. In the
light of the above consideration, the absence of nondiamond carbon band in
the Raman spectrum of sample B can be attributed to the growth of diamond
crystallites of larger average size and so of higher phase purity.
Aiming at studying the nature and distribution of impurity-induced opti-
cal centres, PL measurements have been carried out. The comparison of the
PL spectra registered at room temperature on sample B, grown at 3% CH4
concentration for 1 h, and on sample C, grown at 2% CH4 concentration for
7 h, is shown in Fig. 4a.
The diamond Raman peak is clearly visible in both spectra, however,
sample B exhibits a stronger photoluminescence. In order to understand the
origin of this large band, PL measurements at low temperature (80 K) on
sample B have been carried out (Fig. 4b). Thanks to both low temperature
Fig. 3. SEM micrograph of the samples A (top, left), B (top, right) and C (bottom)
and greater focus depth due to the use of a long focal distance X50 objec-
tive, a richer scenario is obtained. The spectrum shows, besides the diamond
Raman peak, the a-C Raman band originating from more defective layers at
the substrate interface, and some PL features associated with nitrogen im-
purities [10, 21, 25, 26]. The peak at 2.156 eV corresponds to the ZPL of the
nitrogen-vacancy [N-V]0 centre in the neutral state. In addition, the spectral
features at 1.945 eV and at 1.882 eV can be attributed to the ZPL and the
vibronic sideband of the [N-V]− centre. Such PL bands are commonly seen in
PL spectra of nitrogen-doped CVD diamond films [27, 28, 29]. As in [30, 31],
sample B has been grown with gases nominally without nitrogen. However,
the observation of nitrogen-related centres in the PL spectra witnesses the
presence of residual nitrogen in the growth chamber, probably due to impu-
rities in the gas mixture or to not perfect vacuum conditions of the chamber
before the deposition.
3.1.2 Continuous Samples
So far, the optical characterisation has been performed on noncontinuous

samples. In order to study the global quality of CVD diamond films grown
Fig. 4. PL spectra of samples B and C at (a) room temperature and (b) 80 K. The
energy positions of the diamond and a-C-related Raman peaks together with the
PL bands have been indicated
in the same experimental conditions of the discontinuous samples previously

analysed, the deposition process of two samples was continued until they
coalesced into continuous films. The two CVD diamond films were deposited
at Ts = 1000◦C and [CH4 ] = 3%, and at Ts = 970◦ C and [CH4 ] = 2%,
respectively. The resulting polycrystalline films have been made freestanding,
with final thicknesses of approximately 50 µm and 40 µm, respectively.
Figure 5 shows the micro-PL spectra taken at the growth surface of the
two freestanding films. For a better comparison, the diamond peak is mag-
nified in the inset. The change in CH4 concentration from 3% to 2% results
in a narrowing of the diamond line from 3.5 to 2.6 cm−1 . The PL back-
ground of the sample grown at 2% CH4 concentration is very low, without
evident nitrogen-related PL bands, witnessing an excellent quality of the sur-
face grains. At higher methane content, the PL background is higher, but
still unstructured. Bergman et al. [21, 22] pointed out a correlation between
broadband PL and phase purity. On the basis of this correlation, the broad-
band PL emission has been attributed to a continuous distribution of in-gap
states introduced by sp2 carbon phases in the film. These results indicate
Fig. 5. Micro-PL spectra measured at the growth surface of two continuous sam-
ples, grown at [CH4 ] = 2% and [CH4 ] = 3%
that, in the range of substrate temperatures investigated, better crystalline

quality and higher phase purity are obtained at lower methane concentration.
Figure 6 shows the comparison between the micro-Raman spectra reg-
istered on the growth surface and the nucleation side of the freestanding
film grown at 2% methane concentration of Fig. 5. As can be seen, both
the surfaces show a complete absence of nondiamond carbon phases. Only
a broadening of the diamond line from 2.6 to 3.2 cm−1 is observed when
the laser spot is focused from the surface to the nucleation side. This indi-
cates that the polycrystalline film grown at 2% CH4 concentration exhibits
diamond grains of high phase purity even on the nucleation surface.
3.2 Statistical Study of the Nucleation Process
A statistical study of the nucleation process of pulsed glow-discharge CVD

diamond films onto molybdenum substrates has been carried out. This study
is based on an analysis of the statistical distribution of the grain sizes per-
formed by scanning electron microscopy of the growth surface of the three
samples A, B and C, previously characterised by means of Raman spec-
troscopy and photoluminescence. A model of a two-step nucleation process
has been deduced, and a correlation between the nucleation density and the
most important thermodynamic parameters, such as substrate temperature
and CH4 concentration, has been assessed.
Figure 7 shows the frequency distribution of the grain linear size of the
sample C, whose SEM micrograph is shown in Fig. 3. The grain linear size
Fig. 6. Comparison between the micro-Raman spectra, measured at the growth

surface and the nucleation side, of a continuous sample grown at [CH4 ] = 2%
concentration and Ts = 970◦ C
Fig. 7. Grain size frequency distribution of sample C. The points represent the
experimental data. The fit function (continuous line) is the sum of two components,
related to a “fast” (dashed line) and “slow” (dotted line) nucleation process, as
described by (1)
Fig. 8. Grain size frequency distribution of samples A, B and C. The points repre-
sent the experimental data; the continuous line corresponds to the fit function (1)
has been defined as the greatest distance (expressed in µm) between couples
of points on a same grain. The frequency distribution is quite asymmetric,
composed of a sharp maximum, with a Gaussian-like tail on the right, and a
broader distribution on the other side.
As shown in Fig. 8, all samples exhibit a similar grain linear size frequency
distribution, notwithstanding the different deposition conditions (Table 1). If
we adopt the grain size as an obvious indicator of the time elapsed after the
nucleation of each grain, and consider the grain distribution smoothed by the
spread in the grain growth rate, then we can explain our experimental data
assuming a two-step nucleation process.
– First step: “fast” nucleation. At first, a sudden and simultaneous nucle-
ation of a great many grains occurs, represented by the sharp maximum
in the frequency distribution.
– Second step: “slow” and constant nucleation. Then, other nuclei form,
with constant probability per unit time, resulting in a roughly square
distribution, smoothed by the differences in grain growth rate.
Under this assumption, the nucleation rate, i.e., the number of nuclei that
appears per unit time and unit surface, can be expressed by:
f (t) = kδ(t) + νχ[0,T ] (t),
where k is the surface nucleation density of the “fast” phase, δ(t) is the Dirac
distribution, ν is the nucleation rate (nucleation density per unit time) of the
“slow” phase and T is the deposition time, χ[0,T ](t) being the characteristic
function of the time interval [0,T ]. If, with a reasonable approximation, we

assume a Gaussian distribution g(v) for the growth rate of the grains:

1 (v − v0 )2
g(v) = √ exp − ,
σv 2π 2σv2
where v0 denotes the mean value and σv stands for the standard deviation of
the distribution, then the experimental data shown in Fig. 7 are well fitted
by the function:
T x
+ dx
1 t t
F (x) = dt f (t ) · dv g(v ), (1)
0 dx x
t
where x = vt is the statistic variable, i.e., the grain linear size. In Table 2 the
parameters derived by the fitting procedure are reported.
Table 2. Fitting parameters entering (1)
Sample Ts (◦ C) [CH4 ] (%) v0 (µm/h) σv /v0 k (mm−2 ) ν (mm−2 h−1 )

A 1030 3 14 0.28 1 800 1 900
B 1000 3 13 0.32 750 450
C 1040 2 2.6 0.30 77 6.5
The evidence of a two-component grain size distribution in the three sam-

ples under consideration indicates that, in our working conditions, a quite
complex nucleation pattern occurs onto Mo substrates. Although the small
number of samples does not allow us to be conclusive about the dependence
of the nucleation density on temperature and methane concentration, some
hypotheses can be formulated.
– The dependence of the nucleation density on temperature does not seem
to be very critical, at least in the temperature range under consideration,
as it is observed by comparison of samples A and B, grown at the same
CH4 concentrations and different temperatures.
– On the other hand, by comparison of samples A and C, grown at approx-
imately the same substrate temperature, a difference from 2% to 3% in
the methane concentration produces a two-order of magnitude difference
in the “slow” nucleation probability and a one-order difference in the
“fast” one.
The different dependence of the two distributions on the methane concen-
tration indicates two different mechanisms of nucleation. The fit parameter
σv /v0 is quite the same in the three samples, in spite of the differences in the
deposition time. This result suggests that at time t = 0, when the grain size is
zero, the grain size distribution has only the Dirac delta component, i.e., the
first component is almost instantaneous at the beginning of the deposition

process. Therefore, in the first step of the nucleation process, nucleation sites
form, whose number strongly depends on methane concentration.
The slow component of the nucleation, being a process ruled by a constant
probability per unit time, can be explained by the nucleation model proposed
by Angus [32]. The model establishes the dependence of the nucleation rate
on the concentration of active species, [CHx ], forming hydrogenated graphitic
fragments, which can act as nucleation centres for diamond [33]. Given n the
minimum number of carbon atoms in the fragment that acts as a nucleation
center, Angus finds an expression for the nucleation rate in terms of the con-
centration ratio [CHx ]/[H]. This expression, in the limit of very high hydrogen
concentration (which is assumed to be satisfied for the high temperature-high
activation state of a dc pulsed-glow discharge plasma [12]) can be written:
n
[CHx ]
rn = K · v0 · , (2)
[H]
where v0 is the growth rate and K depends on temperature.

Moreover, if we suppose the concentration of active species [CHx ] propor-
tional to the methane percentage, the data reported in Table 2, relative to
the samples A and C, grown at approximately the same temperature, allow
us to establish that the minimum number of carbon atoms in the hydro-
genated graphitic fragment is approximately n ≈ 10. Interestingly enough,
10 is just the number of carbon atoms of the lightest hydrocarbon possess-
ing the diamond unit structure, hadamantene (C10 H16 ), which is believed to
be a possible precursor of diamond, produced by a homogeneous gas-phase
nucleation [4].
4 Conclusions
In this work, a comparative study on pulsed glow-discharge CVD diamond
grown onto molybdenum substrates, using a CH4 –H2 gas mixture at different
methane concentrations has been presented. The effect of the change in the
deposition parameters on the nucleation and growth process has been stud-
ied by Raman spectroscopy and photoluminescence. A correlation between
the results of the optical characterisation and of the statistical study of the
nucleation process can be evidenced.
The micro-Raman analysis carried out shows unshifted and narrow dia-
mond lines at the growth surface of samples grown at low CH4 concentration,
indicating low stress level and high crystalline quality of individual grains.
Furthermore, a high phase purity of the diamond crystallites has been evi-
denced by the complete absence of any feature related to nondiamond carbon
and by a very weak luminescence background. Micro-Raman measurements
carried out on the nucleation side of free-standing films have only evidenced
the presence of residual stress in diamond grains. The absence of nondiamond

carbon phases witnesses that the films grown at lower CH4 concentration ex-
hibit high phase purity even at the nucleation side.
Thanks to the statistical analysis, it has been assessed that the growth
parameters have a more evident dependence on the CH4 content than on
the substrate temperature, at least in the temperature range investigated.
Lower nucleation density has been found in the samples grown at lower CH4
concentration.
On the basis of these preliminary results, it is possible to conclude that
the use of a lower CH4 content in the gas mixture, inducing lower nucleation
density on molybdenum, results in the synthesis of CVD diamond samples
of higher crystalline quality and phase purity. The same correlation between
CH4 concentration and film quality has been found in free-standing samples.
These experimental evidences may be explained as follows. Once the nu-
cleation process onto Mo substrate has started, diamond crystallites grow
larger until they coalesce into a continuous film. Then the growth proceeds
only upwards, following a columnar structure. With increasing film thickness
the size of diamond crystallites increases, and the grain boundary density cor-
respondingly decreases, improving the film quality. If the number of diamond
crystallites per unit surface is low (low nucleation density), the grain coales-
cence will occur at greater crystallite size. In this way, the layer close to the in-
terface between the diamond film and the Mo substrate will be poorer of grain
boundaries, where defects and nondiamond carbon phases are preferentially
accumulated. As a consequence, the resulting film will exhibit a good quality
even on the nucleation side. However, it should be highlighted that the corre-
lation between lower nucleation density and higher crystalline quality found
in diamond films deposited on molybdenum is strictly dependent on the sub-
strate used. In fact, recent studies carried out on diamond samples grown
on Si substrates [34] have evidenced that a lower nucleation density could
have the drawback of a greater incorporation of Si atoms in the growing film,
which could strongly degrade the final quality of the sample. Further studies
are necessary to cover a large range of deposition parameters and to clarify
the correlation between molybdenum substrate temperature and film quality.
Acknowledgements
We wish to thank the Scientific Director of the Centro Microscopie Elettron-
iche, CEME-CNR, Florence, Prof. Laura Morassi Bonzi, for the availability
of the SEM facilities and helpful discussions.
References
[1] M. G. Donato, G. Faggio, G. Messina, S. Santangelo: in Chap. 17 329
[2] C. Jany, A. Tardieu, A. Gicquel, et al.: Diam. Relat. Mater. 9, 1086 (2000)
329

S. Santangelo, A. Tucciarone, G. V. Rinati: Eur. Phys. J. B 20, 133 (2001)
329, 333
[4] H. Liu, D. S. Dandy: Diam. Relat. Mater. 4, 1173 (1995) 329, 330, 340
[5] L. L. Regel, W. R. Wilcox: Acta Astr. 48, 129 (2001) 329
[6] S. Barrat, P. Pigeat, I. Dieguez, et al.: Thin Sol Films 304, 98 (1997) 329
[7] S. T. Lee, Z. Lin, X. Jiang: Mater. Sci. Eng. 25, 123 (1999) 330
[8] W. L. Wang, K. J. Liao, L. Fang, et al.: Diam. Relat. Mater. 10, 383 (2001)
330
[9] K. Mitsuda, Y. Kojima, T. Yoshiba, et al.: J. Mater. Sci. 22, 1557 (1987) 330
[10] M. G. Donato, G. Faggio, G. Messina, S. Santangelo: Optical characterisation
of diamond films grown by chemical vapour deposition, in G. Mondio, L. Silip-
igni (Eds.): Progress in Condensed Matter Physics (Società Italiana di Fisica,
Bologna 2003) pp. 389–406 330, 334
[11] H. D. Zhang, H. Q. Li, J. H. Song, et al.: Appl. Surf. Sci. 246, 90 (2005) 330
[12] S. Sciortino, S. Lagomarsino, F. Pieralli, et al.: Diam. Relat. Mater. 11, 573
(2002) 330, 340
[13] G. S. Ristic, D. Bogdanov, S. Zec, et al.: Mat. Chem. Phys. 80, 529 (2003)
330
[14] M. Ece, B. Oral, J. Patscheider: Diam. Relat. Mater. 5, 211 (1996) 330
[15] D. S. Knigth, W. B. White: J. Mater. Res. 4, 385 (1989) 331
[16] R. J. Nemanich, J. T. Glass, G. Lucovsky, et al.: J. Vac. Sci. Technol. A46,
1783 (1988) 331
S. Santangelo, A. Tucciarone, G. V. Rinati: Diam. Relat. Mater. 10, 1535
(2001) 332
[18] D. F. Bahr, D. V. Bucci, L. S. Schadler, et al.: Diam. Relat. Mater. 5, 1462
(1996) 332
[19] L. Fayette, B. Marcus, M. Mermoux, et al.: J. Mater. Res. 12, 2686 (1997)
333
[20] P. K. Bachmann, H. D. Bause, H. Lade, et al.: Diam. Relat. Mater. 3, 1308
(1994) 333
[21] L. Bergman, M. T. McCloure, J. T. Glass, et al.: J. Appl. Phys. 76, 3020
(1994) 333, 334, 335
[22] L. Bergman, B. R. Stones, K. F. Turner, et al.: J. Appl. Phys. 73, 3951 (1993)
333, 335
[23] L. C. Nistor, J. V. Landuyt, V. G. Ralchenko, et al.: J. Mater. Res. 12, 2533
(1997) 333
[24] L. C. Nistor, J. V. Landuyt, V. G. Ralchenko, et al.: Diam. Relat. Mater. 6,
159 (1997) 333
[25] M. C. Rossi, S. Salvatori, F. Galluzzi, et al.: Diam. Relat. Mater. 7, 255 (1998)
334
[26] K. Iakoubovskii, G. J. Adriaenssens: J. Appl. Phys. 76, 1349 (2000) 334
[27] A. Wotherspoon, J. W. Steeds, B. Catmull, et al.: Diam. Relat. Mater. 12, 652
(2003) 334
[28] J. J. Shermer, F. K. de Theije: Diam. Relat. Mater. 8, 2127 (1999) 334
[29] E. Rzepka, F. Silva, A. Lusson, et al.: Diam. Relat. Mater. 10, 542 (2001) 334
[30] P. Kania, P. Oelhafen: Diam. Relat. Mater. 4, 425 (1995) 334
[31] L. T. S. Lin, G. Popovici, Y. Mori, et al.: Diam. Relat. Mater. 5, 1236 (1996)
334
[32] J. C. Angus: in A. Paoletti, A. Tucciarone (Eds.): The Physics of Diamond,
International School of Physics “Enrico Fermi”, Course CXXXV, Varenna,
Italy, 1996, vol. 135 (IOS, Amsterdam 1997) 340
[33] W. R. L. Lambrecht, C. H. Lee, B. Segall, et al.: Nature 364, 607 (1993) 340
[34] M. G. Donato, G. Faggio, G. Messina, S. Santangelo, M. Marinelli, E. Milani,
G. Pucella, G. V. Rinati: Diam. Relat. Mater. 13, 923 (2004) 341
Index
chemical vapour deposition (CVD), 329 nucleation probability, 339
nucleation rate, 338
defects slow phase, 338
grain boundaries, 333, 341 statistical study, 336–339
in diamond, 329, 333, 334, 341 surface nucleation density, 338
diamond, 329
diamond nucleation, 329, 336, 338, photoluminescence (PL), 329, 331,
339 333–335
micro-PL analysis, 333–336
heteroepitaxial growth, 329
homoepitaxial growth, 329
Raman spectroscopy, 329–333, 335–337
Mo substrates, 329–341 micro-Raman analysis, 331–333, 337
nucleation, 329–341 scanning electron microscopy (SEM),

fast phase, 338, 339 330, 333, 334, 336
Optical Characterisation of High-Quality
Homoepitaxial Diamond
Maria G. Donato1 , Giuliana Faggio1 , Giacomo Messina1 , Saveria

Santangelo1 , and G. Verona Rinati2
1
Dipartimento di Meccanica e Materiali, Università “Mediterranea” di Reggio
Calabria, Località Feo di Vito, I-89060 Reggio Calabria, Italy
2
Dipartimento di Ingegneria meccanica, Università di Roma “Tor Vergata”, Italy
Abstract. Recently, great effort has been devoted to the deposition of homoepitax-
ial diamond for electronic applications. Its single-crystal nature avoids the problems
due to grain boundaries in polycrystalline diamond films, although the quality of
such materials may still be very high. However, the optimisation of the deposition
process of single-crystal diamond has not yet been achieved. In fact, the surface mor-
phology of homoepitaxial diamond seems to be very sensitive to the quality of the
diamond substrate and to even slight variations in the deposition parameters (com-
position of the gas mixture, microwave power, substrate temperature). Moreover,
the growth of flat samples needs very low growth rates, which are not practical for
technological exploitation of the material. Thus, a deposition process at relatively
high growth rates needs to be developed. In this view, characterisation studies of
the material deposited play a key role, because they may give precious hints for the
optimisation of a fast growth process of electronic-grade homoepitaxial diamond.
In this work, we report on the characterisation of single-crystal diamond grown on
diamond substrates of different origin. The samples have been deposited by means
of microwave plasma-enhanced chemical vapour deposition (MWPECVD) with a
CH4 –H2 gas mixture at different methane concentrations and at approximately
560◦ C substrate temperature. The growth rates range between approximately 1
µm/h and 4.5 µm/h. Optical and scanning electron microscopy have been used
to study the surface morphology of the samples. A Raman investigation has been
carried out to study the crystalline quality and the spatial homogeneity of the ma-
terial, by means of accurate measurements of position and width of the diamond
Raman peak at different points in the samples. Photoluminescence (PL) has been
used to investigate both the phase purity and the distribution of possible impurities
in the deposited material. The results indicate that homoepitaxial CVD diamond
samples having very high crystalline quality (full width of the diamond Raman
peak ∼ 1.75 cm−1 ) and no impurity-related defect centers can be obtained in the
adopted deposition conditions.
1 Introduction
Diamond has been looked for, down the centuries, owing to its undisputed
beauty and rarity. However, in recent years, the research about diamond has
assumed a more scientific aspect. In fact, its optical and electrical properties,
among which transparency to visible light, high breakdown field, high radia-
346 Maria G. Donato et al.
tion hardness and chemical inertness, would allow, in principle, the realisation
of highly performing electro-optical devices, such as solar-blind UV detectors
or high-energy particle detectors, which might substitute the Si-based devices
in harsh environments. However, natural diamond, mainly because of lack in
standardisation, cost and inclusion of impurities, is not a feasible proposal as
an engineering material. For these reasons, a great effort has been and is still
being done to establish a highly reproducible deposition process allowing the
large-scale synthesis of diamond crystals.
At the beginning of the research about a suitable synthesis process, dia-
mond crystals were obtained from graphite subjected to high-pressure high-
temperature conditions (HPHT process). However, the cost of the deposition
set up and, above all, the reduced size of the specimens obtained led the
research towards a more convenient deposition process.
In the same period, diamond synthesis was also obtained by a low pressure
technique, involving decomposition of a gas mixture containing a hydrocar-
bon and hydrogen. Typically, in microwave chemical vapour deposition (MW-
CVD) a microwave discharge transforms a methane-containing gas mixture
into plasma. Diamond deposition is enhanced by the presence of hydrogen,
which has an effective role in the etching of nondiamond carbon phases.
When diamond is deposited by CVD onto nondiamond substrates (het-
eroepitaxy), polycrystalline films are obtained. The most commonly used
substrate is silicon, thanks to its widespread availability. Continuous progress
in the deposition techniques produced diamond samples having most of the
exceptional properties of natural diamond [1, 2]. However, the polycrystalline
nature of such material may put some serious limitations on its technolog-
ical exploitation. Grain boundaries, in fact, act [3] as preferential sites for
the incorporation of defects and impurities, which, in turn, may cause a dra-
matic worsening in the electronic properties of the material. Anyway, poly-
crystalline diamond films show values of the mobility-lifetime product of the
carriers comparable to those found in GaAs and SiC [4] and, until a highly
reproducible and standardised growth process of homoepitaxial diamond is
achieved, polycrystalline diamond seems to be a good candidate for diamond-
based electronics.
If diamond substrates are used for the deposition (homoepitaxy), single-
crystal specimens are obtained. For this purpose, 100-oriented commercial
HPHT diamond is commonly used. Recently [5] very high values for the elec-
tron and hole mobility measured on a CVD homoepitaxial diamond sample
have opened the way to the realisation of diamond-based highly performing
electronic devices. However, a deposition process which could guarantee very
high electronic properties has not still achieved, as pointed out in [4]. More-
over, very often homoepitaxial samples exhibit structural defects as pyra-
midal hillocks or unepitaxial crystallites, which, incorporating defects and
impurities, cause the deterioration of the electronic quality of the material [6].
Optical Characterisation of High-Quality Homoepitaxial Diamond 347
The problem of depositing homoepitaxial diamond having a flat surface

by CVD has been faced by different authors [7,8,9,10,11,12,13]. Watanabe et
al. [7, 8, 14] have shown that homoepitaxial samples having an atomically flat
surface can be deposited at very low (0.05 %) methane concentration. How-
ever, such a deposition process has the drawback of being extremely slow. In
fact, growth rates of some tens of nanometers per hour have been obtained.
Thus, the deposition of samples having suitable thickness for advanced elec-
tronic applications would require unacceptably long growth times. A moder-
ate improvement in the growth rate (approximately 0.5 µm/h) has been ob-
tained by Stammler et al. [9] with hot filament CVD. The authors succeeded
in depositing samples exhibiting very low density of unepitaxial crystallites;
however, both nitrogen and silicon atoms are incorporated in the growing
material. To further enhance the growth rate, Teraji et al. [10, 11] proposed
the use of quite high microwave power. They showed that at 3.8–4.2 kW
microwave power, flat samples can be obtained at a ratio of the methane
flow to the total deposition gas flow [CH4 ] = CH4 /(CH4 + H2 ) = 4%. In
these conditions, a relatively high growth rate (approximately 2.5 µm/h) has
been achieved. Higher growth rates (up to 25 µm/h) can be obtained at even
higher methane concentrations. However, the samples deposited show the in-
clusion of nitrogen atoms, probably coming from the hydrogen used for the
deposition [10].
To suppress the formation of unepitaxial crystallites, Bauer et al. [12]
grew homoepitaxial samples onto slightly off-axis substrates. Such a proce-
dure is based on the observation that unepitaxial crystallites form at the top
of the pyramidal hillocks commonly found in homoepitaxial samples, but not
on their lateral faces.
In order to enhance the quality of homoepitaxial CVD diamond, Tal-
laire et al. [13] used a two-step deposition process: before the diamond de-
position, the substrates were subjected to an etching process in a hydrogen–
oxygen plasma. After this pretreatment, the effective diamond deposition at
3.2 kW microwave power has been carried out. In these conditions, the best
quality sample has been grown at 4% CH4 with a growth rate of approxi-
mately 6 µm/h.
In this work, we report on the characterisation of homoepitaxial diamond
grown by the research group of the University of Rome Tor Vergata [15]. As
our group has already obtained the optimisation of the deposition process
of polycrystalline CVD diamond [1], approximately the same experimental
conditions were used as a starting point for the deposition of single-crystal
CVD diamond specimens. After the growth, the morphology of the deposited
samples was studied by optical microscopy (OM) and SEM. The crystalline
quality of the samples was studied by means of Raman spectroscopy. Pho-
toluminescence(PL)at different excitation wavelengths was used to evidence
the possible presence of defects and/or impurities in the deposited material.
The results show that homoepitaxial diamond having extremely high quality
can be obtained with the deposition conditions used.
2 Experimental
Homoepitaxial diamond samples have been grown onto nonselected diamond
substrates by microwave plasma-enhanced CVD (MWPECVD) at the Uni-
versity of Rome Tor Vergata. Three samples have been deposited at 1%
CH4 /H2 at a substrate temperature of approximately 560◦ C . As a test for
the effect of a higher concentration of methane in the deposition gas mixture,
a CVD film was deposited at 4% CH4 /H2 . For all the samples, the microwave
power was 720 W. The characteristics of the samples are summarised in Ta-
ble 1.
The Raman scattering measurements were carried out at room tem-
perature with an Instrument S.A. Ramanor U1000 double monochromator,
equipped with a microscope Olympus BX40 for micro-Raman sampling. The
514.5 nm line of an Ar+ ion laser (Coherent Innova 70) was used to excite
Raman scattering. Using a X100 objective, the laser beam was focused to a di-
ameter of about 1 µm. A depth resolution of about 4 µm was obtained with a
confocal aperture of 200 µm. Micro-photoluminescence (µ-PL) measurements
were carried out at room temperature by using the same experimental setup
used for micro-Raman spectroscopy. The 514.5 nm line (2.41 eV) and the
457.9 nm (2.71 eV) of the argon-ion laser were used to excite luminescence;
the spectra were taken in the region 1.54–2.7 eV.
Table 1. Deposition parameters of the samples analysed in this work. ND and

HCVD stand for natural diamond and homoepitaxial CVD diamond, respectively
Sample Substrate [CH4 ] Ts Pressure Thickness Growth rate

% K mbar µm µm/h
SCD4 ND 1 560 149 171 1.1
SCD5 IIb 1 560 147 90 1
SCD7 HCVD 1 550 147 52 1.1
SCD15 ND 4 560 149 89 4.6
3 Results
3.1 Optical and SEM Characterisation
In their studies about the deposition of atomically flat homoepitaxial dia-

mond, Takeuchi et al. [14] proposed a map of the surface morphologies which
Fig. 1. Optical micrographs (×10 objective) of the samples studied in this work
may be obtained at different CH4 /H2 concentrations in the deposition gas

mixture and at different misorientation angles of the (001) diamond sub-
strates. They found that homoepitaxial diamond can develop three different
types of surface morphology, at CH4 /H2 concentrations ranging between 0%
and 2%, and misorientation angle ranging between 0◦ and 5◦ .
In the region of misorientation angles lower than 1.5◦ and CH4 /H2 ratios
higher than 0.05%, surface morphologies exhibiting unepitaxial crystallites
and pyramidal hillocks are obtained. If the misorientation angle is higher than
1.5◦ , macro-bunching steps are observed. At the boundary between these two
regions, surface morphologies with unepitaxial crystallites, pyramidal hillocks
and macro-bunching steps are observed. Finally, at methane concentrations
lower than 0.15%, the samples show flat surfaces, independently of the mis-
orientation angle of the diamond substrates.
The optical micrographs of the samples deposited in this work are shown
in Fig. 1. All the different surface morphologies are clearly observed. In par-
ticular, whereas in sample SCD5 nearly square-shaped hillocks are clearly
seen, in sample SCD7 unepitaxial crystallites are observed. Their distribu-
tion and size are different on the surfaces of the various samples. SCD4 and
SCD15, grown on the top and back side of the same diamond substrate, but
at different methane concentrations, show a bunching-step-like morphology.
The different surface morphologies are obviously observed also in SEM mi-
crographs (Fig. 2). This analysis confirms that the different conditions of
Fig. 2. SEM micrographs (×500 magnification) of samples studied in this work. The
micrograph referring to SCD7 has been registered at lower (×250) magnification
the initial diamond substrates have a “driving” effect on the morphology of

the overlaying homoepitaxial diamond [14]. In particular, the different mor-
phologies observed in films grown at the same methane concentrations are
attributed to different misorientation angles of the substrates. The similar
morphology observed on the samples grown on the two sides of the same
substrate, but at different methane concentrations, is associated with a sim-
ilar misorientation angle of the substrate surfaces.
3.2 Raman Characterisation
3.2.1 First-Order Raman Scattering
Raman spectroscopy is a very powerful tool to study the crystalline quality of

synthetic diamond. In fact, a perfect diamond structure has a triply degener-
ate optical phonon at the center of the Brillouin zone whose energy is approx-
imately 165 meV. The inelastic light scattering from this phonon gives rise
in the first-order Raman spectrum of diamond to a sharp peak at 1332 cm−1
whose full width at half maximum (Γ ) is only approximately 2 cm−1 . The
study of the crystalline quality of synthetic diamond is commonly performed
by comparing the values of position and width of the measured Raman peaks
with the corresponding values of natural diamond. To properly compare the
measured linewidths with the narrow standard value, slit-induced broadening
has to be considered. In the following, we will show diamond Raman peaks
registered with only 50 µm slit aperture, determining an instrumental line
broadening on the order of 0.15 cm−1 .
Fig. 3. Micro-Raman spectra registered on the surface of the samples studied in

this work. Excitation wavelength λ = 514.5 nm. Values of position (XP ) and width
(Γ ) of the peaks are also indicated
The Raman spectra shown in Fig. 3 demonstrate the extremely high crys-
talline quality of the deposited samples. In particular, the diamond peak reg-
istered on the sample grown at 4% CH4 /H2 (SCD15) is only 1.75 cm−1 wide.
Raman linewidths registered on the samples range between 1.75 cm−1 and
2.31 cm−1 . Such linewidths are narrower than the ones obtained on the sub-
strates and even narrower than the linewidths typically found on unselected
natural diamond specimens. Lower values have been recently found on a
520 µm thick homoepitaxial diamond [13] and on a 350 µm thick freestanding
sample [12]. In a recent work, Tallaire et al. [16] showed that also in homoepi-
taxial diamond the linewidth of the Raman peak becomes narrower as the
sample thickness increases. In this regard, we remark that the thickness of
our samples does not exceed 170 µm and that the thickness of the sample on
which the narrowest Raman peak has been registered is only 80 µm. On the
basis of these considerations, we can conclude that the linewidths measured
on our relatively thin samples point out a crystalline quality comparable to
that of the best quality samples reported in the literature.
3.2.2 Second-Order Raman Scattering
In the previous section, we discussed the Raman scattering from the optical
phonon at the centre (k = 0) of the Brillouin zone. This is the only mode
which is Raman-active in the first-order spectrum. When one-phonon scatter-

ing is considered, the conservation of the crystal momentum requires that the
wave vector of the scattering phonon is approximately zero [17, 18]. However,
if a two-phonon process is considered, the conservation of the crystal momen-
tum requires that the two scattering phonons have equal and opposite wave
vectors. The second-order Raman spectrum arises from critical points in the
two-phonon density of states of diamond; moreover, to be Raman-active, the
modes have to satisfy selection rules which can be deduced by group theory
analysis [18]. As a consequence, the second-order Raman spectrum of dia-
mond ranges between 1800 cm−1 and 2667 cm−1 , and is explained in terms
of combinations and overtones of phonons in diamond [18]. The intensity of
the second-order Raman spectrum is much lower than the intensity of the
first-order Raman spectrum, and thus it can be easily obscured by Raman
or PL signals originating from defects and/or impurities in the material. In
fact, Solin and Ramdas [18] measured the second order spectrum at visible
excitation (488 nm) in a IIb natural diamond, whereas Wagner et al. [19]
observed it in polycrystalline films, but with a more energetic excitation
(4.82 eV), to diminish the contribution originating from a PL background
in the second-order Raman region. Thus, the observation, consequent to a
visible excitation, of second-order bands in the Raman spectrum of diamond
is an indicator of high material quality, because it requires the complete ab-
sence of any other spectral feature arising from defects and impurities. As a
further confirmation of the great quality of our homoepitaxial diamond sam-
ples, the spectral features due to second-order Raman scattering are clearly
detected in spectra registered, at 514.5 and 457.9 nm exciting wavelengths,
in the 2000–2700 cm−1 range (see inset of Fig. 7).
3.3 Photoluminescence Characterisation
Photoluminescence consists in light emission following the excitation of

the electronic states of the material by means of a UV–visible radiation
source [20]. As the diamond band gap is 5.5 eV, PL emission in the visible
range would be impossible in a perfect diamond crystal. Instead, impurity-
related defects having energy levels within the diamond band gap may de-
termine PL emission in the visible range. Obviously, each defect centre has
its typical PL emission. Thus, the study of PL spectra in diamond allows the
identification of its possible contaminants.
A typical PL spectrum of a defect centre has its own characteristic struc-
ture, consisting of a sharp zero-phonon line (ZPL) and vibronic sidebands.
The ZPL corresponds to the transition between the lowest vibrational state
of the excited electronic state to the lowest vibrational state of the ground
electronic state. The vibronic sidebands, i.e., one-phonon, two-phonon, etc.,
bands correspond to the transition from the lowest vibrational state of the
excited electronic state to the first, second, etc., excited vibrational state in
Normalised PL intensity (a. u.) Optical Characterisation of High-Quality Homoepitaxial Diamond 353
1 0,2
x10 Raman peaks
H3 vibronic band
2.156 eV
vibronic band
1.945 eV
vibronic band
0,0
1,6 1,8 2,0 2,2 2,4 2,6
0
1,6 1,8 2,0 2,2 2,4 2,6
PL energy (eV)
Fig. 4. Room-temperature micro-PL spectra registered at the same point of the
sample SCD4 using two different excitation wavelengths: 514.5 nm (black ) and
457.9 nm (light gray)
the ground electronic state. To identify optical centres, the spectral posi-
tions of both ZPLs and vibronic sidebands must be accurately measured. As
overlapping multi phonon transitions are observed at high temperatures, the
optical centres are usually identified by means of PL measurements carried
out at low temperatures (usually 77 K or less).
Nitrogen is the main contaminant of diamond. The inclusion of nitrogen
in CVD diamond films is quite commonly found in the literature results,
probably due to small residual amounts of this element in the growth chamber
after evacuation or to the use of not ultra pure gases. The incorporation of
nitrogen during the growth gives rise to the well-known PL centre at about
2.156 eV, attributed [21] to a [N-V]0 defect centre. A different state of charge
of this defect [21] gives rise to a PL band with ZPL at 1.945 eV ([N-V]−
defect centre).
Another PL feature commonly associated to N impurities consists of a
ZPL centered at 2.463 eV and vibronic sidebands corresponding to a phonon
energy of 40 meV. This structure is attributed to the defect complex involving
two nitrogen atoms and a vacancy ([N-V-N] centre), usually called H3 defect.
The type of defect which is observed in PL measurements depends on the
excitation energy used. In fact, only defects having energy lower than the ex-
citation energy may be observed. For this reason, PL measurements are gen-
erally performed with different excitation wavelengths. In our measurements
we have used two excitation wavelengths: the 514.5 nm line (corresponding
354
Normalised PL intensity (a. u.) Maria G. Donato et al.
1 0,1
x10 Raman peaks
1.945 eV
2.156 eV
vibronic band
0,0
1,5 1,7 1,9 2,1 2,3
0
1,6 1,8 2,0 2,2 2,4 2,6
PL energy (eV)
Fig. 5. Room temperature micro-PL spectra registered at the same point of the
to an excitation energy of 2.41 eV) and the 457.9 nm (2.71 eV) line. The first
line may evidence the presence of the two nitrogen-related defects at 2.156 eV
and 1.945 eV, while the second may evidence, besides these, the H3 defect
(2.463 eV). However, resonance effects, or selective absorption [21], may hide
the presence of the PL structure related to a particular centre, so only the
comparison between the spectra obtained with both the excitation energies
may give a complete identification of the defects present in the material.
In Fig. 4 the spectra obtained on the same surface point of sample SCD4
are shown. From these data, the quality of the sample is evident: the main
features of the spectra are the two sharp peaks corresponding to the Raman
scattering from the diamond crystal. However, a weak photoluminescence
background is observed, due to the inclusion of nitrogen atoms, in both the
H3 [N-V-N] and in the [N-V] defect configuration. The presence of both types
of defect indicates a moderate inclusion of nitrogen atoms during the growth.
It is interesting to observe the resonant excitation of 1.945 eV emission with
the 2.41 eV excitation energy. Such an effect has been already observed in
polycrystalline samples [22], and it has been explained hypothesising the
presence of a higher excited electronic state of the [N-V]− centre at an energy
value approaching 2.41 eV.
The resonant PL emission of the 1.945 eV defect centre is more impres-
sively observed in Fig. 5, where the spectra obtained on the same surface point
of the sample SCD5 are shown. When PL is excited with the green laser line
(2.41 eV) the ZPL and the vibronic band of the [N-V]− defect centre are
1
Normalised PL intensity (a. u.)
Raman peaks
0
1,6 1,8 2,0 2,2 2,4 2,6
PL energy (eV)
Fig. 6. Room-temperature micro-PL spectra registered at the same point of the
seen, together with a weak contribution of the 2.156 eV centre. However, the
more energetic laser line at 457.9 nm does not excite the centres. Moreover,
the [N-V-N] defect is not seen in this sample.
In Fig. 6, the spectra registered on the sample SCD7 are shown. No ni-
trogen-related defects are seen. Finally, in Fig. 7 the PL spectra obtained
on the sample SCD15 are shown. The exceptional crystalline quality and
phase purity indicated by the Raman measurements are confirmed also by
PL characterisation. A completely flat spectrum and the second-order Ra-
man spectrum of the diamond structure is observed with both the excitation
wavelengths used.
4 Discussion and Conclusions

In the previous sections we have shown measurements carried out on ho-
moepitaxial diamond samples grown onto diamond substrates of different
origin. Optical and SEM micrographs have shown very different surface mor-
phologies, which have been associated with the different conditions of the
diamond substrates, and in particular with the misorientation angles of the
100 diamond substrate. Thus, we may conclude that the initial state of
the diamond substrate plays a crucial role on the final morphology of ho-
moepitaxial diamond. This may be a problem in view of the technological
1
1st order x10
Raman peaks
Normalised PL intensity (a. u.)
2nd order
Raman bands
500 1500 2500
0
0 2000 4000 6000 8000
-1
Wavenumber shift (cm )
Fig. 7. Room-temperature micro-PL spectra registered at the same point on sample
SCD15 using two different excitation wavelengths: 514.5 nm (black ) and 457.9 nm
(light gray). To evidence the presence of the second-order Raman of diamond peak,
data have been plotted as wavenumber shifts
exploitation of this material. In fact, an accurate and standardised prepara-

tion of diamond substrates, which are not as common as silicon wafers, may
be too expensive. Moreover, the removal of diamond substrates, which are
almost always of lower quality than the homoepitaxial layer, may not be easy
to carry out. Anyway, the crystalline quality of the obtained samples, even
if deposited onto not-selected substrates, is really exceptional: on relatively
thin samples, values of Raman linewidth as low as 1.75 cm−1 are measured.
Such values indicate a great crystalline quality of the material, which becomes
really attractive if compared with the more easily obtainable polycrystalline
diamond. However, some further improvements in the experimental setup
should be done. In fact, photoluminescence measurements indicate the pos-
sible inclusion of nitrogen atoms. The incorporation of such impurities is not
an easily resolvable issue, because, as other authors have observed [9, 11, 23],
at high growth rates nitrogen atoms may also be incorporated in films grown
without the intentional addition of N2 in the deposition gas mixture. Thus,
a further step in the optimisation of the deposition process of homoepitax-
ial diamond will be the individuation of the growth conditions which both
limit the incorporation of nitrogen atoms and guarantee to obtain the best
quality material. Finally, it seems that the sample grown at higher methane
concentration (sample SCD15, 4% CH4 ) has a higher crystalline quality and
no incorporation of defects and impurities. Other authors have already ob-
served an improvement of sample quality with increasing methane concentra-

tion [11, 13]; however, their experimental conditions are very different from
ours. In fact, while Teraji et al. [11] use a high-power deposition, Tallaire et
al. [13] perform the etching of the diamond substrates before the homoepi-
taxial diamond deposition. On the contrary, when depositing polycrystalline
samples, we have observed [1], with our experimental setup, that the best
quality material is obtained at lower methane concentrations. For this rea-
son, a further systematic study on samples grown on substrates of identical
quality, changing the CH4 concentration in the deposition gas mixture needs
to be done, in order to assess the role of this growth variable on the quality
of homoepitaxial diamond obtained.
Acknowledgments
The authors wish to thank the group of Prof. Aldo Tucciarone at the Uni-
versity of Rome Tor Vergata for kindly providing the diamond samples and
performing the SEM measurements.
References
S. Santangelo, A. Tucciarone, G. V. Rinati: Eur. Phys. J. B 20, 133 (2001)
346, 347, 357
[2] J. Alvarez, A. Godard, J. P. Kleider, et al.: Diam. Rel. Mater. 13, 881 (2004)
346
[3] L. Bergman, B. R. Stoner, K. F. Turner, et al.: J. Appl. Phys. 73, 3951 (1993)
346
[4] C. Manfredotti: Diam. Rel. Mater. 14, 531 (2005) 346
[5] J. Isberg, J. Hammersberg, E. Johansson, et al.: Science 297, 1670 (2002) 346
[6] Y. Vohra, A. Israel, S. A. Catledge: Appl. Phys. Lett. 71, 321 (1997) 346
[7] H. Watanabe, D. Takeuchi, S. Yamanaka, et al.: Diam. Rel. Mater. 8, 1272
(1999) 347
[8] H. Okushi: Diam. Rel. Mater. 10, 281 (2001) 347
[9] M. Stammler, H. Eisenbeiß, J. Ristein, et al.: Diam. Rel. Mater. 11, 504 (2002)
347, 356
[10] T. Teraji, S. Mitani, C. Wang, et al.: J. Crys. Growth 235, 287 (2002) 347
[11] T. Teraji, M. Hamada, H. Wada, et al.: Diam. Rel. Mater. 14, 255 (2005) 347,
356, 357
[12] T. Bauer, M. Schreck, H. Sternschulte, et al.: Diam. Rel. Mater. 14, 266 (2005)
347, 351
[13] A. Tallaire, J. Achard, F. Silva, et al.: Diam. Rel. Mater. 14, 249 (2005) 347,
351, 357
[14] D. Takeuchi, H. Watanabe, S. Yamanka, et al.: Diam. Rel. Mater. 9, 231 (2000)
347, 348, 350
[15] High Resolution Single Crystal CVD Diamond Particle Detectors, presented
at the 16th European Conference on diamond, diamond-like materials, carbon
nanotubes and nitrides, Toulouse (France), 11–16 September 2005 347
[16] A. Tallaire, J. Achard, F. Silva, et al.: Phys. Stat. Sol. A 201, 2419 (2004)
351
[17] N. W. Ashcroft, N. D. Mermin: Solid State Physics (Saunders College,
Philadelphia 1976) p. 481 352
[18] S. A. Solin, A. K. Ramdas: Phys. Rev. B 1, 1687 (1970) 352
[19] J. Wagner, M. Ramsteiner, C. Wild, et al.: Phys. Rev. B 40, 1817 (1989) 352
[20] J. Walker: Rep. Progr. Phys. 42, 108 (1979) 352
[21] A. T. Collins: The electronic and optical properties of diamond, in A. Paoletti,
A. Tucciarone (Eds.): The Physics of Diamond, Proceedings of the Interna-
tional School of Physics Ënrico Fermi”, Course CXXXV, Varenna 1996 (Ios,,
Amsterdam 1997) p. 273 353, 354
[22] M. C. Rossi, S. Salvatori, F. Galluzzi, et al.: Diam. Rel. Mater. 7, 255 (1998)
354
[23] A. Chayahara, Y. Mokuno, Y. Horino, et al.: Diam. Rel. Mater. 13, 1954 356
Index
CVD diamond, 345–348 homoepitaxial growth, 346
defects optical microscopy (OM), 345, 347–349,

grain boundaries, 345, 346 355
in diamond, 347, 352–356
detectors photoluminescence (PL), 345, 347,
diamond-based detectors, 346 352–355
diamond micro-PL analysis, 348, 352–356
diamond properties, 345 plasma-enhanced chemical vapour
high-quality diamond, 347, 348, deposition (PECVD), 345
351–357 polycrystalline diamond, 345–347, 352,
homoepitaxial diamond, 345–348, 354, 356, 357
352, 355–357
HPHT diamond, 346 Raman scattering, 350, 351, 354
IIb type diamond, 348, 352 first-order Raman scattering, 350
natural diamond, 348, 352 second-order Raman scattering, 351,
nitrogen contamination, 347, 353–356 352
silicon contamination, 347 Raman spectroscopy, 345, 347, 348,
single-crystal diamond, 345 350–352, 355, 356
micro-Raman analysis, 348, 351–356
heteroepitaxial growth, 346
high-pressure high-temperature scanning electron microscopy (SEM),
(HPHT), 346 345, 347–350, 355
Pulsed Laser Deposition of Carbon Films:
Tailoring Structure and Properties
Paolo M. Ossi1 and Antonio Miotello2

1
Dipartimento di Ingegneria Nucleare and Centro di Eccellenza
“Nano-Engineered MAterials and Surfaces” (NEMAS), Politecnico di Milano,
Via Ponzio 34/3, I-20133 Milano, Italy
paolo.ossi@polimi.it
2
Dipartimento di Fisica, Università di Trento, I-38050 Povo (TN), Italy
miotello@science.unitn.it
Abstract. Pulsed laser deposition (PLD) is a versatile technique to synthesize

films of a given material with strongly different properties. In this chapter we discuss
the microstructure and structure of carbon films ranging from diamond-like carbon
to tetrahedral amorphous carbon, to nanometer-sized cluster assembled films, de-
pending on PLD process parameters. All films are hydrogen-free, as shown by FTIR
spectroscopy, and result invariably structurally disordered. Carbon local hybridisa-
tion was checked by Raman spectroscopy, both visible and UV. Vacuum deposited
films undergo a transition from mainly disordered graphitic to up to 80% tetrahe-
dral amorphous carbon above a threshold laser energy density. The microstructure
and morphology of the deposited films were studied in a complementary way by
SEM and AFM. All vacuum deposited films appear featureless. Cluster assembled
glass-like films display different growth modes, mainly depending on the deposited
energy per laser pulse and nature-pressure of the ambient gas. Film morphology is
correlated with the ballistic mechanisms that occur when plasma plume constituents
interact with ambient gas atoms. Cluster formation in the plume is modelled, and
we estimate the average number of atoms per cluster. Model predictions agree with
cluster sizes deduced by Raman spectroscopy and observed by TEM imaging.
1 Introduction
Carbon displays a wide variability of structures and properties. It is a good
candidate to explore the tunability of material properties that in turn make
it suitable for emerging technologies. Carbon can crystallise in several forms,
including, besides the most widely known diamond and graphite, fullerene
molecules like C60 and carbon nanotubes, which are rolled tubules of graphite
sheets. There are also many noncrystalline carbons due to the different types
of bonding and different degrees of disorder. We are thinking of amorphous
carbon and nanostructured carbon, the latter being a mixture of amorphous
carbon, nanotubes, fullerenes and graphitic fragments. Amorphous carbon
with a large fraction of tetrahedral, diamond-like sp3 bonds is known as
diamond-like carbon (DLC). The possibility to synthesise DLC at room tem-
perature is a great advantage with respect to diamond films. The interesting
360 Paolo M. Ossi and Antonio Miotello
mechanical and optical properties of DLCs lead to several applications, in-

cluding magnetic hard disk coatings, wear-protective and antireflective coat-
ings, flat panel displays and biomedical coatings.
The versatility of carbon materials is due to the critical dependence of
their physical properties on the sp2 (graphite-like) to sp3 (diamond-like) bond
ratio [1]. We know many types of sp2 bonded carbons with different degrees
of graphitic order, that range from microcrystalline graphite to glassy carbon.
Any amorphous carbon has a mixture of sp3 , sp2 and occasionally sp1 bonds,
possibly in the presence of a fraction of hydrogen that can be as high as
60%. As a rule, hydrogenated amorphous carbons have a quite small content
of sp3 C–C; DLCs with the highest sp3 contents are known as tetrahedral
amorphous carbon (ta-C).
The electronic structure of amorphous carbons essentially consists of the
strong σ bonds of sp3 and sp2 sites that form the occupied bonding states
in the valence band and the empty σ ∗ antibonding states in the conduction
band; a wide gap separates σ from σ ∗ states. The π bonds of sp2 and sp1
states result in the occupied π and unoccupied π ∗ states that largely lie within
the σ–σ ∗ gap. Each modification of the sp2 /sp3 fraction is associated with a
change of the density of states and of the energy gap. Any technique able to
synthesize carbon films, flexible enough to allow to control the above fraction,
the degree of structural disorder and the extent of clustering of sp2 phase,
its orientation and anisotropy, is suitable to prepare DLC films with largely
different structural, mechanical, optical and electronic properties. In partic-
ular, ta-C is grown by deposition techniques involving energetic ion beams,
or a plasma; such an energetic beam deposition is necessary to stabilise the
metastable sp3 bonding. It was observed that the optimum ion energy to
maximise the fraction of tetrahedrally coordinated material, around 70–80%,
lies between 100 and 200 eV. In the following we discuss the synthesis of DLC
and ta-C films by pulsed laser deposition. The intrinsic conceptual simplicity
of the technique is associated with the possibility to change several deposition
parameters, thus tailoring film structure and properties.
Hydrogen-free ta-C is characterised by its low friction coefficient and
high hardness [2], optical transparency in a wide spectral range and ab-
sorbance in the UV. Such a set of properties is promising for scratch-re-
sistant thin coatings of lenses [3]. The effectiveness of pulsed laser deposi-
tion (PLD) techniques to obtain in vacuum hydrogen-free amorphous carbon
films was ascertained long ago [4]. PLD is characterised by low deposition
temperatures and high deposition rates. The main process parameters that
affect the degree of diamond-like character of deposited films are vacuum
quality, laser wavelength λ and power density P . To obtain DLC a vac-
uum of at least 10−4 –10−5 Pa, and a minimum P value, which increases
with λ (P ≈ 3 × 106 W mm−2 at λ = 248 nm), seem to be necessary [5]. Ul-
traviolet excimer lasers allow one to obtain good quality films [6], since par-
ticulate emission is reduced and ta-C films are obtained at comparatively
Pulsed Laser Deposition of Carbon Films 361
low laser fluences with respect to visible and infra-red lasers [7]. On the
other hand, a dramatic change from dense, uniform, compact films to pro-
gressively more open microstructures is observed when films are deposited in
a background atmosphere, either inert or chemically reactive. Our discussion
includes the features of carbon films deposited by PLD at high pressures of
inert gas, which permit us to obtain nano-glassy films.
Here we report the morphological and structural analyses of carbon films
pulsed laser deposited in vacuum, in a reactive gas (molecular nitrogen) low-
pressure atmosphere and in an inert gas (helium and argon) high-pressure
atmosphere. Ablation of carbon atoms and ions from the irradiated target,
plasma formation and interaction of the plume of ejected particles among
themselves and with background gas lead to strong modifications of mi-
crostructure and structure of the deposited films [8]. We performed:
– scanning electron microscopy (SEM) and atomic force microscopy (AFM)
to investigate film morphology
– Fourier transform infrared (FTIR) spectroscopy to assess hydrogen con-
tent
– visible and UV Raman spectroscopies to study carbon atom coordination
– electron energy-loss spectroscopy (EELS) on selected samples to quantify
the fraction of tetrahedrally coordinated carbon
– transmission electron microscopy (TEM) on selected samples deposited
in inert gas to directly check the presence and size of carbon clusters
A discussion of the mechanisms underlying the formation of carbon films with
largely different properties is offered.
The films were deposited at room temperature on cleaned Si (100) substrates,
in a vacuum chamber (base pressure, 10−4 Pa) either at an operating pres-
sure of 10−2 Pa, or in different background gases. These were a molecular
nitrogen atmosphere at 1 Pa, or a helium or argon atmosphere at pressures
pHe;Ar = 0.6, 30, 50, 60, 70, 250, 103 and 2 × 103 Pa. Highly oriented, high-
purity (99.99%) pyrolytic graphite (HOPG) was ablated with laser pulses
from a KrF excimer laser (λ ≈ 248 nm, pulse duration τ = 20 ns, repetition
frequency 10 Hz, incidence angle 45◦ ), at power densities ranging from 0.25
to 19 MW mm−2 . The average deposition rate of films prepared in vacuum
and in nitrogen was about 0.6 nm s−1 , while for films prepared in helium it
was around 0.7 nm s−1 and in argon it was around 0.8 nm s−1 . More details of
the deposition system are reported in [9]. We used a JEOL JSM 6300 SEM,
with the primary beam accelerated at 15 keV and an incidence angle of 30◦ .
After metallization, both surface and cross-sectional pictures were taken for
all examined samples. A Thermo Microscopes CP Research AFM was used in
noncontact mode; measurements were made at room temperature, in ambient

air.
FTIR spectra were taken at room temperature, in dry air, using a Win-
Bio-Rad Fast Transform spectroscope over the 400–4000 cm−1 range. The
spectra were recorded in transmission and corrected for substrate contribu-
tions.
Unpolarized visible micro- (films deposited in vacuum and in nitrogen;
spot diameter, 1 µm; incident power at the sample surface, 3 mW) and
macro- (films deposited in helium and argon) Raman spectra were recorded
in backscattering geometry for 532 nm excitation from a Nd:YAG laser us-
ing a Jobin-Yvon T64000 triple grating spectrometer with a resolution of 1
cm−1 .
UV Raman spectra were collected using an UV-enhanced CCD camera
on a Renishaw micro-Raman System 1000 spectrometer, modified for use at
244 nm (resolution, 4–6 cm−1 ), with fused-silica optics throughout. All UV
spectra were corrected for system response signal, obtained by measuring
a background spectrum with an Al mirror and normalising to atmospheric
molecular nitrogen vibrations.
EELS measurements were performed with a dedicated VG501 scanning
transmission electron microscope fitted to a spectrometer with a McMullan
parallel EELS detection system.
3 Results
Among the deposited carbon films a clear a posteriori separation into two
families with distinct and often opposite properties is possible. Thus it ap-
pears obvious to separately discuss the properties of samples synthesized in
vacuum, or at low pressure, before presenting our results on cluster assem-
bled (CA) films prepared at high background gas pressure.
3.1 Low-Pressure Deposited Films
The surface of all films belonging to this family appears smooth, uniform
and featureless. SEM pictures (not shown) indicate the absence of droplets,
as expected when carbon is ablated with pulses in the UV region, while
accidental micrometer-sized debris are sometimes observed. All films of this
family adhere well to the Si substrates.
Two representative FTIR transmission spectra of DLC films deposited in
vacuum are shown in Fig. 1. In both spectra no absorption bands in the range
2800–3100 cm−1 , due to the C–Hx stretching mode, are found. This means
that the films are hydrogen-free (more precisely, H content is below the de-
tectable limit of 0.5 at. %). The overall aspect of the spectra indicates that
such films are structurally disordered. Yet, the spectrum (curve 1) of the film
Fig. 1. FTIR transmission spectra of DLC films deposited in vacuum below

(curve 1 ) and above (curve 2 ) the threshold laser power density P t . Substrate
contributions have been subtracted
deposited at low laser intensity (P = 0.5 MW mm−2 ), below the threshold

intensity value P t = 5 MW mm−2 , shows a broad band at about 710 cm−1 ,
attributed to a vibrational mode observed also in visible Raman spectra dis-
cussed below. This feature coincides with a peak in the calculated vibrational
density of states of a graphene sheet [10]. We believe that it can be attributed
to a collective in-plane deformation of graphitic islands. Indeed, the calcu-
lated IR and Raman spectra of model polyconjugated hydrocarbons, ranging
from C24 to C114 [11], display intense bands in the region between 700 and
750 cm−1 that arise from in-plane cooperative motions involving combina-
tions of C–C stretching and C–C–C bending. An evident shoulder at about
1250 cm−1 is attributed to a stretching mode of mixed sp2 –sp3 C–C bonds,
while the high-frequency shoulder at about 1550 cm−1 is due to vibrations
of sp2 C=C bonds [12]. The above bands are absent in the featureless spec-
trum (curve 2) of the transparent film deposited at P = 10 MW mm−2 , thus
above the threshold P t . Film transparency is an indirect indication that a
structural transition from disordered graphitic carbon to ta-C occurred. A set
of visible Raman spectra of films deposited in vacuum (V) and in nitrogen
atmosphere (N2 ) at the same intensity, for various increasing intensities, is
reported in Fig. 2. The spectra of vacuum-deposited films indicate that the
content of fourfold-coordinated carbon increases with laser intensity. Indeed,
given the slight differences of film thickness, we can compare spectra of dif-
ferent films. In the films deposited above P t the increasing intensity of the
peaks due to the crystalline Si substrate (first order at ∼ 520 cm−1 and sec-
ond order at ∼ 970 cm−1 ) with increasing power density indicates increasing
film transparency, associated with increasing content of fourfold-coordinated
carbon atoms. Two features are evident in all spectra: the D peak (∼ 1350
cm−1 ), due to breathing of hexagonal rings, and the G peak (∼ 1560 cm−1 )
due to the relative motion of sp2 carbon atoms [13]. Indeed both “peaks”
are broad bands, the nomenclature coming from the literature. Inset a in
Fig. 2 shows a magnification of a broad, weak band centred around 720 cm−1
in a film deposited below P t . The nice coincidence of this feature, that is
lacking in films deposited at high power densities, with the band observed
in the above-discussed IR spectra of the same films is noteworthy. The spec-
tra were fitted to a combination of a Breit–Wigner–Fano (BWF) function
for the G peak and a Lorentzian for the D peak. The evolution of relevant
parameters (Raman shift at BWF maximum and D to G peak intensity ra-
tio, I D /I G ) with increasing power density is shown (diamonds; dashed lines)
in Fig. 3. We consider the Raman shift corresponding to the peak value of
the BWF function to allow comparison with literature data where symmet-
ric lineshape fitting is used. Both fitting parameters indicate that above the
threshold intensity P t the films consist largely of ta-C. The coincidence of an
almost zero value for the I D /I G ratio and of a G peak position at a frequency
above 1550 cm−1 is a sufficient condition to assert a high sp3 content in the
films [13].
Looking at Fig. 2, the low-frequency Raman spectra of the set of films
deposited in N2 atmosphere at a partial pressure of 1 Pa with the same power
densities used for vacuum-deposited films, can be attributed to mainly three-
fold-coordinated disordered carbon networks. Thus nitrogen induces quali-
tative changes in the dynamics of the ablation plume. Fitting the spectra
to the above-discussed lineshapes results in the parameter trends shown in
Fig. 3 (triangles, dashed-dotted lines) and confirms that no ta-C was formed
in this set of films, even at the highest power density. This conclusion is sup-
ported by two immediate observations: first (see inset b in Fig. 2), we find
the band around 720 cm−1 even in the film deposited at the maximum power
density P = 18.5 MW mm−2 . Second, all films of this group are opaque,
the substrate Si peaks being lacking. In these samples the I D /I G ratio is
about unity and remains nearly constant, while the G peak position slightly
red-shifts, in agreement with the observed trends [14].
UV Raman spectra (not reported) of films grown in vacuum show the
G peak, around 1600 cm−1 , thus blue-shifted with respect to its position
in low-frequency Raman spectra. The characteristic feature is the T peak
around 1100 cm−1 , attributed to all sp3 C–C bond vibrations [15]. At the
lower power densities a further evident shoulder around 1400–1430 cm−1 is
discernible. We believe that this feature corresponds to the D peak in visible
Raman spectra [6]. The results of the peak-fitting procedure are displayed
in Fig. 4. We observe that, with increasing power density P , the T peak is
Fig. 2. Visible micro-Raman spectra of DLC films pulsed laser deposited at various
laser power densities in vacuum (V) and in nitrogen atmosphere (N2 ). The insets
(a) and (b) show the broad band centered around 720 cm−1 in non-ta-C films
red-shifted and its intensity increases. Correspondingly, the G peak is blue-

shifted, it becomes more symmetric and its intensity in turn increases. The
ratio I T /I G increases with power density up to about 0.45, then it remains
constant, again indicating that a transition to ta-C occurred [13].
UV Raman spectra of films prepared in N2 atmosphere display the T and
G broad peaks and a small, narrow extra peak around 1550 cm−1 due to at-
mospheric oxygen. Again from Fig. 4, with increasing deposited power density
the G peak around 1570 cm−1 remains asymmetric and shows no obvious po-
sition shift [14]; the position of the T peak, around 1300 cm−1 , is unchanged.
The feature around 1400 cm−1 is absent in all spectra. The ratio I T /I G is
around 0.35 and remains nearly constant. Looking at the trends of fitting
Fig. 3. Fitting parameters of visible Raman spectra versus laser power density: left
scale, intensity ratio between G and D peaks, I D /I G ; right scale, G peak position.
Diamonds, dashed lines: films deposited in vacuum; triangles, dashed-dotted lines:
films deposited in nitrogen atmosphere. The lines are guides to the eye
Fig. 4. Fitting parameters of UV Raman spectra versus laser power density: left
scale, intensity ratio between T and D peaks, I T /I G ; right scale, G peak position.
Diamonds, dashed lines: films deposited in vacuum; triangles, dashed-dotted lines:
films deposited in nitrogen atmosphere. The lines are guides to the eye
Fig. 5. Transmission EELS spectra of a selected film deposited in vacuum at laser

power density P below P t
parameters for these films it is evident that ta-C was not synthesized with
the above process conditions.
Transmission EELS measurements allow us to quantify [16] in selected
films deposited in vacuum the fraction of sp2 -hybridised, and thus also of
sp3 -hybridised carbon from the ratio of the area of the sp2 peak at 285 eV
to that of the (sp2 + sp3 ) peak at 290 eV. Such a ratio is about 40% (Fig. 5)
in a film prepared at P = 0.25 MW mm−2 , thus below the threshold laser
intensity P t . In Fig. 6 the same ratio is about 80% in a film deposited above
P t , at P = 10 MW mm−2 . In films deposited in nitrogen atmosphere at the
same values of power density, the measured sp3 fraction slightly increases
from about 40% to about 45%. Thus EELS results agree with the indications
from vibrational spectroscopies concerning ta-C formation in this family of
films.
3.2 High-Pressure Deposited Films

Carbon films were deposited at room temperature in helium and in argon
atmospheres at pressures of 0.6, 30, 50, 60, 70, 250, 1000 and 2000 Pa, with
power intensities ranging from 8.5 to 19 MW mm−2 . Both planar and cross-
section SEM observations show that columnar, nodule-like and dendritic,
highly porous microstructures occur in the films prepared in helium. Such
microstructures are found according to the above sequence, mainly as a func-
tion of increasing partial gas pressure in the deposition chamber pHe and also,
less sensitively, of the deposited power density. Notably, for various combina-
tions of power density and helium pressure, columns coexist with nodes that
Fig. 6. Transmission EELS spectra of a selected film deposited in vacuum at laser

power density P above P t
develop within the columnar structure during the film growth. At the lowest
pHe = 0.6 Pa, film surfaces are very similar to those of DLC films deposited in
vacuum; they appear flat and laterally homogeneous. The pictures in Figs. 7,
8, 9 are representative of the evolution of film microstructure with increasing
pHe . The microstructure of the films deposited at medium-low pHe (30–70 Pa)
consists of densely packed columns with several embedded spherically capped
nodes that are partly agglomerated with each other. Such films adhere well to
the substrate. With increasing pHe , the number density (cm−3 ) of the nodes
increases while their cap radius decreases. The calculated ratio between mea-
sured node diameters, at the top and at half of the height is equal, or very
close to 21/2 ; the same value is obtained when the ratio is made between the
diameter at half of the node height and that at one quarter of node height.
The above diameter ratios have been found for all well-resolved nodes in all
cluster-assembled films deposited at pHe = 50, 60 and 70 Pa. This result
indicates that nodes have parabolic shape. This definitely differs from the
conical shape predicted by ballistic growth models, where the ratio of node
diameters scales with n1/2 , n being an integer greater than 2 [17].
The irregular surface in Fig. 8 is characteristic of all films prepared at
pHe = 250 Pa, irrespective of the laser power density. Columns coexist with
nodes, both being loosely packed in comparison to films deposited at lower
pHe . The nodes are large, with a definite tendency to spherical geometry.
The degree of intrafilm cohesion increases with power density. Yet, film–
substrate adhesion is worse than in films prepared at low helium pressures,
independently of the power density.
Fig. 7. SEM micrograph cross sections of a representative cluster-assembled carbon

film. pHe = 60 Pa, P = 16 MW mm−2

film. pHe = 250 Pa, P = 8.5 MW mm−2

film. pAr = 2 kPa, P = 16 MW mm−2
Fig. 10. SEM micrograph cross sections of a representative cluster-assembled car-

bon film. pAr = 70 Pa, P = 19 MW mm−2
The morphologies of cluster-assembled films deposited in argon have many

similarities to those of films prepared in helium. Essentially, the same se-
quence of columns, nodes, dendritic structure is observed. The dendritic mi-
crostructure in Fig. 9 is highly porous and irregular. Although film thickness
can exceed 15 µm, at some positions on the substrate the surface is barely
covered. The film–substrate adhesion is poor. The aspect of such a film is sim-
ilar to the predictions of ballistic growth models with null atomic mobility,
where strongly off-geometrical equilibrium configurations are frozen.
When a comparison with films deposited in helium under identical con-
ditions is made, films synthesised in argon have three features:
1. The nodes are smaller, with spherical geometry; the spatial distribution
is more homogeneous and the size distribution is narrower.
2. The film–substrate adhesion is lower.
3. The films are thinner and laterally inhomogeneous.
Furthermore, as shown in Fig. 10, a peculiar platelet morphology is found
at pAr = 50, 60 and 70 Pa. The size of these particles lies between 1 and
5 µm with thickness between 400 and 500 nm, independently of argon partial
pressure and laser power density. The particle spatial distribution is, however,
homogeneous. Platelets initially grow normally to the substrate surface, but
with increasing deposition time they show a trend to grow parallel to the
substrate, and in thick films they are included in the columnar structure.
Noncontact AFM on representative film areas of 15 × 15 µm2 was per-
formed for films deposited at pHe;Ar = 50, 60 and 70 Pa, whose surface was
sufficiently smooth to allow for reliable measurements. Using AFM it was
possible to count the nodes and estimate film roughness. Figures 11 and 12
show, respectively, a comparison between the depositions in He and in Ar,
at the same conditions. Figure 12 shows the smaller size and narrower size
distribution of the nodes in films deposited in Ar, thus confirming our SEM
observations. As a general trend, when pHe;Ar changes from 30 to 70 Pa the
average node size decreases, the node number density increases, the size dis-
tribution of the nodes becomes progressively broader and the film roughness
(r.m.s.) correspondingly decreases.
In Fig. 13 are reported three representative Raman spectra of cluster-as-
sembled carbon films. All spectra consist of the G peak and a less evident
D peak at lower frequency. The former is always asymmetric and broad, while
the latter is well defined for the films prepared at the highest ambient gas
pressure, namely pHe;Ar = 2kPa, while it decreases to a shoulder in all other
spectra. Thus the films consist of structurally disordered carbon networks
with trigonal coordination. The presence of a resolved D peak in the films
showing the most irregular morphology indicates that graphitic aggregates of
considerable size were assembled together. The spectra were fitted using the
same combination of a BWF function for the G peak and a Lorentzian curve
for the D peak already adopted to fit Raman spectra of films deposited in vac-
uum and at low ambient gas pressure. At each power density, with increasing
Fig. 11. AFM pictures (15 × 15) µm2 of a representative cluster assembled film
deposited at P = 16 MW mm−2 and pHe = 70 Pa. The morphology dependence on
the nature of ambient gas is evident
background gas pressure, the peak intensity ratio I D /I G decreases, remain-

ing in the range ∼ 0.93 to ∼ 0.85. Correspondingly, the G peak maximum
lies between 1580 cm−1 and 1530 cm−1 . According to the Raman spectrum
analysis [13], the range of values of both I D /I G and G maximum indicate
that the fraction of sp3 -coordinated carbon in cluster-assembled films does
not exceed 10%.
From the I D /I G ratio it is possible to estimate the coherence length La ,
namely the average size of cooperatively scattering domains in these carbon
films. La values for CA films, prepared at ambient gas pressures pHe;Ar rang-
ing from 30 to 250 Pa, evaluated by the Tuinstra–Koenig (TK) relation [18],
lie in a narrow interval between 4.7 and 5.2 nm. Indeed La , calculated by the
TK relation for carbon films with different structures, prepared by several
different techniques, is correlated to I D /I G [19]. Our data for cluster-assem-
bled films fit well into such a correlation. Figure 14 shows that in the I D /I G
interval spanned by the CA films, La linearly scales with I D /I G , both with
helium (Fig. 14a) and with argon (Fig. 14b) as the background gas. Our
films are near the lower limit of validity of the TK relation and belong to
the family of nano-glassy carbon films. In the above analysis we do not con-
sider cluster-assembled films deposited at the highest pHe;Ar values because
only one power density value was adopted for such depositions. Cluster size
Fig. 12. AFM pictures (15 × 15) µm2 of a representative cluster assembled film
deposited at P = 16 MW mm−2 and pAr = 70 Pa. The morphology dependence on
the nature of ambient gas is evident
and structure as deduced from Raman analysis is confirmed by direct TEM

structural observations [20].
4 Discussion
We separately discuss the mechanisms underlying film formation in the two

families of low- and high-pressure deposited films so that qualitative differ-
ences between the experimental conditions leading to different energetic and
chemical evolution of the plasma plume can be better highlighted.
4.1 Low-Pressure Deposited Films
The existence of a threshold intensity P t at which the structure of vacuum-

deposited films changes from mainly graphitic to ta-C can be associated with
a corresponding threshold in the values of ablated particle energy. The plasma
plume produced by laser irradiation of graphite mainly consists of C, C2 , C+
and C++ [21], where the C2 constituent most probably results from ion–atom
collisions followed by charge neutralisation in the plume [21]. The critical fac-
tor to drive ta-C formation is the energy of C+ ions (and energetic neutrals)
Fig. 13. Macro-Raman spectra of representative cluster assembled carbon films.

(a) pHe = 60 Pa, P = 16 MW mm−2 ; (b) pHe = 2 kPa, P = 16 MW mm−2 ;
(c) pAr = 2 kPa, P = 16 MW mm−2
impinging on the growing C film. The latter will be graphitic if only neutral,
low-energy carbon atoms arrive at the substrate. A C+ energy threshold be-
tween 100 and 200 eV appears to be a necessary condition to obtain a high
sp3 /sp2 bond ratio [16]. We observe ta-C formation once the laser power
density exceeds the threshold power density P t ; this is expected, as C+ ion
energy increases as a function of power density [22]. As the experimental con-
ditions are very similar to ours, we assume the same energy values of plume
particles as in [22]. High-energy C+ ions impinging on the growing carbon
film are stopped in interstitial positions at different depths, progressively in-
creasing with increasing ion energy. According to SRIM simulations [23] 50,
100 and 175 eV particles are stopped within 0.5, 1.5 and 2.2 nm below the
surface. Each 175 eV ion produces 1.5 recoils, so that the thermal spike model
to explain ta-C formation [24] is ruled out for our energy range because the
required dense collision cascades do not form. Instead, our results agree with
the subplantation model that predicts film densification and transition from
trigonal to tetrahedral bond coordination with maximum efficiency at ion
energies around 100–150 eV [25].
The presence of a reactive gas such as nitrogen during the ablation process
affects plasma characteristics and laser energy absorption. In general, gas
Fig. 14. Calculated coherence length La versus I D /I G for carbon films prepared by
several techniques, in different deposition conditions. Diamonds for data from [19];
triangles for cluster assembled films deposited in (a) He atmosphere, (b) Ar at-
mosphere. Dotted interpolation lines from [19]
atoms collisionally cool down hot plasma electrons, with two consequences.
First, both electron impact excitation and plasma recombination mechanisms
are more effective, and second, the plasma is better confined at the target
surface, leading to an increase of particle emission. However, both carbon
neutrals and ions emitted from the target react with nitrogen in the plasma
region near the target. Reaction kinetics in plasma is strongly accelerated,
and a conversion of C2 to CN radical readily occurs. Indeed, the optical
emission intensity of the C2 Swan band becomes weaker, while the intensity
of the CN violet band increases with increasing nitrogen pressure [26]. CN is
highly effective to scatter energetic C+ ions in the plasma, thus decreasing
their energy. Such a ballistic mechanism reduces C+ ion ability to trigger the
transition from sp2 to sp3 bond coordination in the growing film. Besides
this, CN is transported at the growing film surface where it is incorporated.
The presence of nitrogen in the growing DLC film results in an increase of
the fraction of sp2 bonded carbon with increasing nitrogen content [14]. Thus
both ballistic and chemical mechanisms inhibit ta-C formation when pulsed
laser deposition is performed in the presence of nitrogen.
4.2 High-Pressure Deposited Films
With respect to film synthesis by pulsed laser deposition in vacuum, or at

low ambient gas pressure, PLD at high gas pressure is a recent process. In
the case of carbon films only a few studies were devoted to the formation
of cluster-assembled films [8, 27], whose molecular structure depends on the
conditions both of cluster formation during plume flight and of cluster landing
on the substrate. The presence of a nonreactive gas (helium, or argon, in our
experiments) in the deposition chamber, at high pressure pHe;Ar from 30 Pa
to 2 kPa leads to carbon cluster condensation in the plume. The net result
is the above-discussed sequence of morphologies in the growing film. The
scarce amount of tetrahedral sp3 bond coordination in all films, irrespective
of laser power density, is due to relevant scattering of plume constituents
from ambient gas and the associated cooling down of the plume to a plume–
ambient gas equilibrium temperature. In agreement with previous results [28],
with respect to plume propagation in vacuum, we visually observed:
(a) enhanced visible fluorescence, as induced by collisions, of the expanding
plume front
(b) spatial confinement of the plume, associated with a drastic change of its
shape
Within the range of the adopted laser parameters, the growth rate of a
typical carbon cluster is considerably higher than the ablation rate of carbon
atoms from the target [8]. Thus the number density of carbon atoms in the
plume is progressively reduced, while the average cluster size increases. Al-
though the processes associated with the initial stage of plume expansion are
highly nonlinear over the isothermal expansion regime that corresponds to
the maximum plume density, it is possible to estimate the number of atoms
per cluster N using average parameter values, taken over a longer time scale.
In a simple ideal gas approach, N depends on:
(a) The nature of the scattering gas, through the geometrical collision cross
section σ. A common value σ = 10−15 cm2 is assumed for σ C−C , for
σ C−He and for σ C−Ar .
(b) The number density of the inert gas n.

(c) The number density of the ablated species, na . Both plume confinement
and the collision rate of individual C atoms with a growing cluster in-
crease with increasing n. Sticking coefficient of 1 and identical behaviour
of energetic neutrals and C+ ions are assumed. Energetic neutrals most
likely arise from charge transfer reactions, each involving a fast carbon
ion and a slow gas atom.
(d) The average escape velocity v e of ablated species. When v e increases,
the time interval between two successive attachment events of individual
C atoms to a growing cluster decreases.
(e) The average cluster formation time, tf . This is shorter than the target–
substrate time-of-flight because below a minimum value of the number
density of ablated particles na the plume becomes noncollisional.
On the basis of the above considerations, the number of carbon atoms in
a cluster may be estimated as:
N = nna σ 2 (v e tf )2 ,
where the product of the cross section times the number density of gas atoms,
n or that of ablated atoms, na , indicates the probability of a carbon atom to
be scattered by, or attached to the cluster, respectively.
Ablation experiments performed in geometric and energetic conditions
similar to the present ones [29] give na ∼ 1017 cm−3 . In the absence of
details on plume energetics we assume tf to be half of the target–substrate
time-of-flight for fast species in the plume. With a target–substrate distance
of 6 cm, and v e ∼ 5 × 105 cm s−1 [8], we obtain tf ∼ 5 × 10−6 s. At the
lower limit of ambient gas pressure pHe;Ar = 30 Pa, n is 7.4 × 1015 cm−3 ,
while at the opposite limit of highest pressure pHe;Ar = 2 kPa, n is 5 × 1017
cm−3 . The average number of carbon atoms in a cluster ranges from N =
5 × 103 (pHe;Ar = 30 Pa), to N = 5 × 105 (pHe;Ar = 2 kPa). We take a value of
0.14 nm for the atomic radius of carbon, to which corresponds an estimated
film density of 1.8 gcm−3 , a typical value for porous nano-glassy carbon
films. If we assume that clusters are spherical, the data give cluster diameters
of 4.7 nm and 22 nm, respectively. Although rough, such an estimate of
cluster size is in agreement with the previously discussed coherence length
La values deduced from the fitting of Raman spectra for cluster-assembled
films deposited at pHe;Ar up to 250 Pa and with TEM observations of cluster
size.
5 Conclusions
In conclusion, we prepared carbon films by pulsed laser deposition in vac-
uum, in a reactive, low-pressure molecular nitrogen atmosphere and in high-
pressure atmospheres of inert gases with different mass, such as helium and
argon. We investigated by vibrational spectroscopies the structure of the

deposited films. In vacuum, sp3 bond coordinated ta-C is obtained above
a threshold power density P t of 5 MW mm−2 . Both ballistic and chemical
mechanisms inhibit the transition from mainly graphitic to ta-C in films
deposited in the presence of nitrogen. At high inert gas pressure, cluster-
assembled, nano-glassy, sp2 bond coordinated films, with strongly different
morphologies, are obtained as functions of the gas species and pressure as
well as of the laser power density. Cluster sizes estimated from experiments
are in agreement with those predicted by a simple model of cluster synthesis
in the ablation plume.
Acknowledgements
The authors are grateful to M. Bonelli and P. Mosaner, Department of

Physics, University of Trento; and to A.P. Fioravanti and D. Bolgiaghi, De-
partment of Nuclear Engineering, Politecnico di Milano, for film deposition;
to V. Russo, Department of Nuclear Engineering, Politecnico di Milano, for
assistance with Raman spectroscopy; and to C. Castiglioni, Department of
Chemistry, Applied Chemistry and Materials, Politecnico di Milano, for use-
ful discussions on the interpretation of vibrational spectra of carbon-based
materials. A.C. Ferrari, Department of Engineering, University of Cambridge,
performed UV Raman and EELS measurements.
References
[1] J. Robertson: Mater. Sci. Eng. R 27, 1 (2002) 360
[2] A. C. Ferrari: Surf. Coat. Technol. 190, 180–181 (2004) 360
[3] M. Bonelli, A. Miotello, P. Mosaner, C. Casiraghi, P. M. Ossi: J. Appl. Phys.
93, 859 (2003) 360
[4] Marquardt, R. T. Williams, D. J. Nagel: in R. Chang, B. Abeles (Eds.): Plasma
Synthesis and Etching of Electronic Materials, vol. 38 (Materials Research
Society Symposium Proceeding, Boston 1985) p. 325 360
[5] A. A. Voevodin, M. S. Donley: Surf. Coat. Technol. 82, 199 (1996) 360
[6] M. Bonelli, A. C. Ferrari, A. P. Fioravanti, A. Miotello, P. M. Ossi: in J. Sulli-
van, J. Robertson, O. Zhou, T. Allen, B. Coll (Eds.): Amorphous and Nanos-
tructured Carbon, vol. 593 (Materials Research Society Symposium Proceeding,
Boston 2000) p. 359 360, 364
[7] R. F. Haglund: Laser Absorption and Desorption (Academic, London 1998)
p. 15 361
[8] A. V. Rode, E. G. Gamaly, B. Luther-Davies: Appl. Phys. A 70, 135 (2000)
361, 376, 377
[9] M. Bonelli, C. Cestari, A. Miotello: Meas. Sci. Technol. 10, N27 (1999) 361
363
[11] M. Rigolio, C. Castiglioni, G. Zerbi, F. Negri: J. Molec. Struct. 79, 563–564

(2001) 363
[12] B. Dishler: in H. Hintermann, J. Spitz (Eds.): Metallurgical Coatings and Ma-
terials Surface Modifications, vol. 17 (European Materials Research Society
Symposium Proceeding, Strasbourg, FR 1987) p. 189 363
[14] J. R. Shi, X. Shi, Z. Sun, E. Liu, B. K. Tai, S. P. Lau: Thin Solid Films 366,
169 (2000) 364, 365, 376
[15] K. W. R. Gilkes, H. S. Sands, D. N. Batchelder, J. Robertson, W. I. Milne:
Appl. Phys. Lett. 70, 1980 (1997) 364
[16] P. J. Fallon, V. S. Veerasamy, C. A. Davis, J. Robertson, G. A. J. Amaratunga,
W. I. Milne: Phys. Rev. B 48, 4777 (1993) 367, 374
[17] Z. Czigany, G. Radnoczi: Thin Solid Films 5, 343–344 (1999) 368
[18] F. Tuinstra, J. L. Koenig: J. Chem. Phys. 53, 1126 (1970) 372
[19] D. S. Knight, W. B. White: J. Mater. Res. 4, 385 (1989) 372, 375
[20] D. Bolgiaghi, A. Miotello, P. Mosaner, P. M. Ossi, G. Radnoczi: Carbon 43,
2122 (2005) 373
[21] Y. Yamagata, A. Sharma, J. Narayan, R. M. Mayo, J. W. Newmann, K. Ebi-
hara: J. Appl. Phys. 86, 4154 (1999) 373
[22] B. Angleraud, J. Aubreton, A. Catherinot: Eur. Phys. J. A 5, 303 (1999) 374
[23] J. F. Ziegler, J. P. Biersack: The Stopping and Ranges of Ions in Matter SRIM
–2000.10 374
[24] H. Hofsäss, H. Feldermann, R. Merk, M. Sebastian, C. Ronning: Appl. Phys.
A 66, 153 (1998) 374
[25] Y. Lifshitz, S. R. Kasi, J. W. Rabelais: Phys. Rev. Lett. 62, 1290 (1989) 374
[26] S. Aoqui, K. Ebihara, T. Ikegami: Composites B 30, 691 (1999) 376
[27] E. Cappelli, S. Orlando, G. Mattei, C. Scilletta, F. Corticelli, P. Ascarelli:
Appl. Phys. A 72, 2063 (2004) 376
[28] D. Geohegan, in D.B. Chrisey, G. Hubler (Eds.): Pulsed Laser Deposition of
Thin Films (Wiley, New York 1994) p. 115 376
[29] M. Bonelli, A. Miotello, P. M. Ossi, A. Pessi, S. Gialanella: Phys. Rev. B 59,
13513 (1999) 377
Index
sp2 -bonded clusters C60 , 359
cluster size, 359, 372, 376, 377 cluster assembled (CA) films, 362,
sp3 /sp2 bonding ratio, 360, 374 368–375, 377
ta-C, 360, 362–365, 367, 373, 374, 376, cluster-assembled (CA) films, 372, 376,
378 378
clustering, 360
atomic force microscopy (AFM), 361,
coatings, 360
371–373
biomedical coatings, 360
background gas, 371, 372, 376, 377 coherence length, 372, 377
columnar film, 367, 368, 371
C24 , 363 node-like film, 367, 368, 371
D band, 364 nanotubes, 359

D/G band intensity ratio, 366, 375 nitrogen, 374, 376, 377
D/G band intensity ratio, 364, 372
dendritic film, 367, 371 optical properties, 360
diamond, 359
diamond-like carbon (DLC), 359, 360, plasma
362, 363, 365, 368, 376 plasma plume, 361, 373
pulsed laser deposition (PLD), 360,
electron energy loss spectroscopy 361, 376
(EELS), 361, 362, 367, 368
Raman spectroscopy, 362–365, 371, 373
friction coefficient, 360
fullerene, 359 scanning electron microscopy (SEM),
361, 362, 367, 369–371
G band, 364
subplantation model, 374
graphite, 359
hardness, 360 T band, 364, 365

T/G band intensity ratio, 366
infrared (IR) spectroscopy, 361–364 T/G band intensity ratio, 365
transmission electron microscopy
nano-glassy carbon, 372, 377 (TEM), 361, 373, 377
Raman Spectra and Structure of sp2
Carbon-Based Materials: Electron–Phonon
Coupling, Vibrational Dynamics and Raman
Activity
Chiara Castiglioni1 , Fabrizia Negri2 , Matteo Tommasini1 ,

Eugenio Di Donato1 , and Giuseppe Zerbi1
1
INSTM and Dipartimento di Chimica, Materiali e Ingegneria Chimica
“G. Natta”, Politecnico di Milano, Piazza Leonardo da Vinci 32, I-20133
Milano, Italy
chiara.castiglioni@polimi.it
2
INSTM and Dipartimento di Chimica “G. Ciamician”, Università di Bologna,
Via Selmi 2, I-40126 Bologna, Italy
fabrizia.negri@unibo.it
Abstract. We focus on several common features that dominate the Raman spec-
tra of carbon materials whenever delocalised π electrons are present. A molecular
approach, based on high-level quantum chemical calculations and experiments on
molecular models, allows us to predict the evolution of these features with the
relevant structural parameters, namely the size and topology of the conjugated do-
mains. These results allow us to obtain insight on the effect of the confinement
(in both one and in two dimensions) of conjugated electrons in terms of electronic
structure as well as of nuclear geometries. On the other hand, the relevant phys-
ical mechanisms that govern the spectroscopic response of theses systems can be
successfully predicted in a very simple and general way in the frame of the Hückel
theory. The generalisation of this theory to the determination of a vibrational po-
tential for a two-dimensional (2D) crystal (graphene) and for carbon nanotubes of
any diameter and chirality is presented and discussed in this work.
1 Introduction
Raman spectroscopy is a widely used technique for nondestructive analy-
sis of solid-state samples of newly developed carbon materials. These ma-
terials range from nanocristalline diamond, films of diamond-like carbon,
a variety of amorphous carbons, carbon fibres and whiskers, carbon nano-
tubes, nanocrystalline or microcrystalline graphites, and structurally con-
trolled graphenes. There are two aims of this research activity:
(a) To obtain a characterization of the materials through the detection of
spectroscopic markers, the determination of their characteristic parame-
ters (frequency, band width and shape, line intensity) and of their changes
with respect to the material synthesis and/or treatment (thermal history,
382 Chiara Castiglioni et al.
annealing, mechanical treatment, laser damaging, etc.). This activity re-

quired the collection of a large amount of experimental data and the
development of useful empirical correlations.
(b) To obtain an understanding of the physical origin of the spectroscopic
response. This kind of investigation requires the help of theory, mod-
els and simulations. As a consequence of this approach, several struc-
ture/properties relationship have been obtained that are useful as guide-
lines for the tailoring of the material properties through the modulation
of the structural parameters at the nanoscale.
An “output” of point (b) is a wide collection of computer simulations of
the spectroscopic behaviour of the materials modelled, which indeed gives
strong support also to the activity described in point (a). The contribution
of the authors of this work mainly concerns point (b). During about five years
of activity on graphitic materials, mainly based on Raman experiments on
polycyclic aromatic hydrocarbons (PAHs) [1] and quantum chemical compu-
tations on molecular models [2, 3, 4, 5, 6], we have collected a lot of evidence of
the intimate relationship between the molecular structure (size of the mole-
cule, topology of the edges, bond lengths, etc.) and the electronic structure
of these systems. This evidence formed the basis for the interpretation of the
spectroscopic behaviour of much more complex materials whose structure is
usually not known or only approximately known [6]. On the other hand, ex-
trapolation to carbon materials requires that a bridge is built, linking the
properties of molecules (oligomers) to those of their infinitely extended par-
ents, describing ideal crystals. This can be done if common theoretical models
are applied for the prediction of properties both of molecules and of crystals.
One of the most intriguing problem arising in the study of the vibrational
spectra of polyconjugated systems is the need of knowing vibrational poten-
tials, which are expected to be strongly sensitive to properties such as the
delocalisation/confinement of π electrons. Moreover, they are also influenced
by bond topologies, defects, etc. In this paper we will discuss the results
obtained by the generalisation of the Kakitani equations [7], developed for
constructing valence force fields for a variety of conjugated molecules. The
treatment is based on the simple Hückel theory, first developed for the mod-
elling of the electronic structure of conjugated organic molecules, and can be
easily extended to infinite two-dimensional (graphene) and one-dimensional
(polyacetylene and carbon nanotubes) crystals. On this basis we will show
that the wide family of materials, characterised by the presence of carbon
atoms in sp2 hybridization state, is affected by common physical mechanisms
which can be described in terms of simple physical models of general validity.
In particular, at the basis of their spectroscopic behaviour, is the existence
of an efficient and selective electron–phonon coupling, which is responsible
for relevant changes of the electronic structure whenever collective nuclear
displacements along preferential paths take place. The introduction of the
concept of electron–phonon coupling allows us to explain several experimen-
Raman Spectra and Structure of sp2 Carbon-Based Materials 383
tal results from Raman spectroscopy (activation, resonance enhancement,

frequency dispersion of specific Raman transitions). Moreover, it provides a
key for the interpretation of other independent experimental findings, such
as, for instance, STM images recently obtained across step edges in highly
oriented pyrolytic crystalline graphite (HOPG) [8]. The natural follow-up of
the concepts developed is the analysis and the interpretation of the Raman
spectra of carbon nanotubes, which still presents several unsolved questions.
It is surprising that the same physical concepts proposed about 20 years ago
for the interpretation of the vibrational spectra of polyconjugated polymers
work very well also in the case of more complex systems characterised by
the presence of domains formed by condensed aromatic rings (graphitic do-
mains). These kinds of structures can be ideally thought as pieces of matter
obtained from an infinite graphite sheet (2D crystal), and they can be de-
scribed as structurally modified graphenes. Structural modifications can be
described as:
(i) Confinement: this happens in the case of molecules (which have indeed
a finite size) and in microcrystals and/or nanocrystals by effect of edges
of different topologies.
(ii) Curvature: this is, for instance, the case of fullerenes and carbon nano-
tubes. Notice that in these cases the curvature is accompanied by con-
finement.
(iii) Defects: a variety of defects can be found ranging from those created
by laser damage, ion implantation or chemical doping to those already
present in the material as obtained, such as, for instance defects due to
the copresence of sp3 carbon phase in amorphous carbon films.
2 Raman Spectra of Polyconjugated Materials

As illustrated in the Chapter by Zerbi et al. (pp. 23–53), the presence of delo-
calised π electrons in polyconjugated polymers makes their Raman response
very peculiar. Raman spectra of polyenes [9] are characterised by few, strong
lines (in simple all-trans polyenes and polyacetylene only two lines dominate
the Raman spectrum), very sensitive to resonance enhancement when the
laser energy approaches their optical energy gap. By comparison of the Ra-
man spectra of simple polyenes of increasing chain length, one can readily
verify that the Raman frequencies of the main two lines show a systematic,
pronounced softening of the characteristic frequencies. In a parallel way any
polyacetylene sample shows frequency shifts of the two strong Raman lines
(peak maxima at 1472 cm−1 and 1060 cm−1 with near-infrared excitation)
while changing the exciting laser energy [10]. Even at the beginning of the
Raman investigations on polyacetylene this behaviour was ascribed to the
presence of different “conjugation lengths” in polyacetylene, each of them
characterised by a given energy gap [10]. This fact is easily justified, both on
the basis of the presence of chemical defects and as due to a wide distribution
of molecular weights, which usually happens in polymer synthesis. Few chains
in the distribution are selectively probed in the Raman spectrum when the
laser energy fulfills their suitable resonance condition: this is the origin of the
observed frequencies softening while decreasing the laser energy.
At a first sight, the Raman spectrum of graphites shows few points in com-
mon with that of conjugated chains: namely, also in this case, the first-order
features are few, relatively strong and sensitive to the laser energy. Highly
ordered and crystalline graphite has a very simple first-order Raman spec-
trum, showing a strong G line at 1580 cm−1 , in agreement with the Raman
selection rules for the 2D crystal (a perfect graphene sheet) characterised by
just one optical phonon (of E 2g symmetry) at the Γ point in the first Bril-
louin zone. A G band (showing sometimes modest frequency shift and line
broadening) is always observed in any sample which contains some amount
of carbon in the sp2 -hybridization state. However, the relevant feature in the
case of graphitic materials containing some kind of structural defects (ranging
from sp2 -rich amorphous carbons to microcrystalline graphites) is the appear-
ance of a strong Raman band, the so-called D band, between 1250 and 1350
cm−1 . This feature shows a systematic frequency softening while decreasing
the energy of the exciting laser and undergoing resonance enhancement [11].
The origin of the D band has been discussed for a long time in the liter-
ature. Thomsen et al. [12] developed a solid state theory approach based
on a double resonant process, activated by the disorder. The authors of the
present paper proposed a different approach, mainly based on molecular spec-
troscopy [2, 3, 4, 5, 6]. As discussed also in the Chapter by Zerbi et al. (pp. 23–
53), the conclusions reached through the “molecular approach” brought to
the idea that the D line (and its frequency dispersion) is the evidence of a
distribution of confined domains in disordered graphites. As in the case of
polyacetylene, according to their characteristic energy gap (modulated by the
confinement) different graphitic domains (showing different D line frequency)
are selectively probed by resonance while varying the energy of the exciting
laser in the Raman experiment [13]. What is really new, with respect to the
case of polyacetylene, is the fact that the D line is perfectly silent in the Ra-
man when the material approaches the limit of a perfect 2D crystal. The aim
of this paper is to show that it is possible to discuss analogies and differences
shown by Raman spectra of conjugated chains and graphenes on the basis of
the concept of electron–phonon coupling. Its main effects (in terms of elec-
tronic structure, nuclear arrangement, vibrational dynamics and potential)
will be described in a simple but conceptually powerful way in the frame of
the Hückel theory.
(a)
(b)
ν = 1476 cm−1 I = 1.0 106 A4 /amu
ν = 1122 cm−1 I = 0.82 106 A4 /amu
Fig. 1. (a) Sketch of the nuclear displacements associated with R− coordinate of

polyene chains and polyacetylene (b) Sketch of the nuclear displacements associated
with Raman-active normal modes (R− modes, see text) of a trans polyene chains
with 22 carbon atoms. Eigenvectors, frequencies and nonresonant Raman intensities
have been obtained from DFT BLYP 6.311G** calculations [14]
3 Electron–Phonon Coupling and Raman Features

3.1 Electron–Phonon Coupling in Polyacetylene
Since from the first experimental determination on polyacetylene and poly-

enes, the evidence of conjugation length-dependent Raman spectra suggested
that the very origin of such a behaviour was the existence of a long-range,
conjugation length-dependent, intramolecular potential [15]. Many valence
force fields were proposed in the past to include this effect. One of the most
widely debated points was to establish the range of extension of interactions
described by nondiagonal harmonic force constants involving CC stretching
coordinates of the polyene chain. Values and extension of these interactions
are indeed expected to be modulated by the degree of delocalisation of π
electrons. The wide literature about polyacetylene vibrational force fields is
referred to and commented on in the review by Gussoni et al. [16], to which
the interested reader is addressed. In this paper we would like to recall only a
few of the relevant points of this discussion. It is remarkable that, in the effort
to reproduce the experimental findings, several authors developed different
empirical force fields for “long chain polyacetylene” and for “short chain
polyacetylene” [17, 18]. Still on the side of the empirical valence force fields
we would like to mention the results of the “effective conjugation coordinate
theory”(ECCT) [19]. According to the ECCT, the effect of the π electron de-
localisation on the intramolecular potential is accounted for by the diagonal
force constant (FR− ) associated with the collective valence coordinate (R−
coordinate) which describes the oscillation of the bond alternation (simulta-
neous stretching of CC double bonds and shrinking of CC single bonds in the
whole chain). As far as the conjugation length increases, the value of FR−
decreases with a consequent softening of the Raman frequencies associated
with vibrational modes involving oscillation of bond alternation. The ECCT

for polyacetylene brought the following important results:
– It showed that a special nuclear trajectory (along the R− direction) does
exist, which is strongly sensitive to the π electron structure (Fig. 1a).
– It showed that the two strong Raman lines of polyacetylene and poly-
enes are due to R− modes, that is, to modes with high “content” of the
dimerisation oscillation (Fig. 1b). Moreover their Raman intensity and
resonance enhancement is due to the large polarization change associ-
ated with this peculiar nuclear displacement.
It is remarkable that until the introduction of high-level ab-initio Moller–
Plesset (MP2) quantum chemical (QC) predictions [20, 21] or density func-
tional theory (DFT) methods [14,22,23], first-principle calculations of the Ra-
man spectra of simple polyenes (based on Hartree–Fock approximation) [24]
were unable to give a reasonable prediction of the observed Raman features.
Quite localised CC stretching interactions were obtained by these methods
and (as a direct consequence) a very modest frequency dispersion with poly-
ene length was predicted for Raman-active bands. In contrast, semiempirical
force fields built on the basis of a quantum mechanical description restricted
to π electrons allowed one to give a correct prediction of the Raman frequen-
cies dispersion. Among others, the best results were obtained by the QCFF-π
methods [25, 26, 27]. An even simpler model that allows one to treat long-
range CC stretching interactions is based on the Hückel theory. In this frame,
according to the derivation proposed by Coulson [28, 29, 30], Kakitani [7] de-
veloped expressions for valence force constants involving CC stretching at
any distance along a polyconjugated chain (1):
∂2E ∂2E σ ∂2E π

fij ≡ = +
∂Ri ∂Rj ∂Ri ∂Rj ∂Ri ∂Rj
2
2
∂ β ∂β
= kσ + 2 2 pi δij + 2 Πij (1)
∂R ∂R
(Cλo∗ ∗
Cµe + Cλe Cµo ∗
) (Cνo Cσe ∗
+ Cνe Cσo )
Πλµ,νσ = + c.c. (2)
o,e
εo − εe
The “ingredients” that appear in the Kakitani expressions are:

(a) Bond–bond polarizabilities (Πij ), which can be simply calculated from
molecular orbital coefficients for π states as obtained according to the
diagonalisation of the Hückel Hamiltonian for the chain considered (2).
In (2) the symbols λ, µ and ν, σ refer to the pair of carbon atoms which
define, respectively, the bond i and the bond j. Πij terms are obtained
in units of β (the Hückel hopping integral between adjacent atomic sites,
which at the 0th-order approximation is assumed to be a constant of
identical value (β0 ) for each pair of adjacent carbon atoms).
(b) Kσ : a force constant relative to the (localised) contribution of σ electrons,

which can be treated as an adjustable empirical parameter (σ contribu-
tions are described as classical springs which affect only the diagonal CC
stretching force constant).
(c) The electron–phonon coupling term, namely the parameter (∂β/∂R)
which describes the modulation of the electronic states (through the hop-
ping integral) by the stretching of CC bonds.
(d) A term consisting of the product of the second derivative of the hopping
integral and the bond order pi . This parameter affects only the diagonal
force constants and can be combined with Kσ in order to define an “ef-
fective” adjustable diagonal parameter Fi (first term in parethesis, (1)).
According to (1) and (2), Hückel theory allows us to determine CC stretch-
ing interactions between the bonds i and j at any given distance, provided
that the molecular orbital coefficients for the π states are calculated. The
introduction of a few empirical parameters, namely β0 , ∂β/∂R and Fi , is
required in order to obtain the harmonic intermolecular potential involving
the CC stretching coordinates. The remaining force constants (involving CH
stretching, bending and bending/stretching interactions) can be assumed in-
dependent from conjugation length and described by short-range interactions
in terms of a set of a few empirical parameters.
Notice, moreover, that in its simpler form the theory makes use (for the
definition of the Hamiltonian, which in turn determines the Πij parameters)
of just one average hopping integral between adjacent sites, namely β0 . It
is, however, possible to choose different values for βij in order to explicitly
consider the existence of a bond alternation in the equilibrium structure. In
the case of equilibrium structures showing alternation of double and single
bonds, as in the case of polyacetylene, this choice can be expressed through
the definition of β1 and β2 parameters for the single and the double bonds,
respectively, which in turn can be written (within a linear approximation of
β(R)) as: β1 = β0 + 1/2(∂β/∂R)∆R and β2 = β0 − 1/2(∂β/∂R)∆R. Notice
that in this case another empirical parameter is required, namely the value
of equilibrium bond alternation ∆R (otherwise referred to as “dimerisation
parameter”, u).
Kakitani formulas have been extended to the case of an infinite chain
(i.e., polyacetylene, regarded as an infinite one-dimensional crystal) by Pis-
eri et al. [31, 32], obtaining a valence force field with seven empirical para-
meters. This has been done considering that polyacetylene orbitals now have
the form of Bloch functions. Accordingly, the 2 × 2 Hückel Hamiltonian of
the perfect crystal (which indeed corresponds to a simple one-electron tight
binding Hamiltonian for π electrons) gives wave vector-dependent coefficients
of the π states (θ and θ in (3) represent the phase associated with k in the
one-dimensional (1D) crystal according to the relationship θ = k · d, where
d is the unit traslation parameter and k varies in the first Brillouin zone (BZ)
of the reciprocal lattice).
Equation (3) gives the expressions so obtained in [32] for the mutual
polarizabilities, where θ and θ label, respectively, Bloch states in the valence
(occupied π orbitals, o) and in the conduction (unoccupied π orbitals, e)
electronic bands:
π π
1 Aλµ (θ, θ )Bνσ (θ, θ )
Πλµ,νσ = dθ dθ + c.c.
(2π)2 −π −π εo (θ) − εe (θ )
Aλµ (θ, θ ) = [Coλ
∗
(θ)Ceµ (θ ) + Coµ
∗
(θ)Ceλ (θ )]
Bνσ (θ, θ ) = [Coν (θ)Ceσ
∗
(θ ) + Coσ (θ)Ceν
∗
(θ )] . (3)
In [32] the above formalism has been used in order to obtain semiempirical
phonon dispersion relations for polyacetylene. Since long-range interactions
are modulated by the parameter (∂β/∂R), they obtained different dispersion
curves while varying the strength of the electron–phonon interaction. Notice
that in [32] for any choice of (∂β/∂R), four adjustable parameters of the
field were determined in such a way that the frequency values observed for
“long-chain polyacetylene” (lower limit of the observed Raman lines) were
predicted. In this way the change of (∂β/∂R) results in a deeper and deeper
dispersion of the two phonon branches associated with the strongest Raman
active transitions (R− phonons). No softening is allowed for k = 0 modes,
due to the procedure followed for the refinement of the adjustable parameters.
On the other hand, if the empirical parameters different from (∂β/∂R) are
kept fixed, frequency calculations also give a remarkable softening of k = 0
modes of R− branches as the electron–phonon coupling parameter increases
in value. This softening is still associated with an increase of the positive
slope of such phonon curves. This last result can be put in relation with the
ECCT and rationalised with the introduction of the following expression for
the diagonal collective force constant FR− for polyacetylene:
1
FR− = (K C=C + K C−C ) − [−f n C=C − f n C−C + 2f n C=C,C−C ] , (4)
2 n
where f n C=C (f n C−C ) are nondiagonal CC stretching force constants be-

tween double C=C (single C–C) bonds at distance n along the chain; these
constants are always negative. f n C=C,C−C are nondiagonal CC stretching
force constants describing interactions between single and double bonds at
distance n and result in positive values. According to (4), it can be realized
that an increase in the values of the constants in the squared brackets and/or
an increase of the range of interactions (number of the terms of the sum) give
rise to a higher positive sum. Notice that as far as the sum increases in value,
we obtain a decrease of FR− since the sum enters with a minus sign in (4).
If now we reconsider interaction force constants (1) as obtained according
to the Kakitani relationships, we can immediately correlate the softening of
FR− with conjugation length (as prescribed by ECCT) to two possible ef-
fects: (i) increase in value of the electron–phonon coupling parameter (which
450
400
350
300
250
200
Raman CC stretching parameter
150
100
50
0
-50
-100
-150
-200
-250
-300
-350
-400
-450
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21
CC bond position (odd positions are C=C)
Fig. 2. Local Raman CC stretching parameters (α = 1/3 Tr(∂α/∂RCC )) as pre-

dicted according to DFT 6.311G** calculations for a polyene chain with 22 carbon
atoms [14]. Units are bohr2 . Notice the change of the sign while passing from double
to single CC bonds
gives rise to an increase of the values of all nondiagonal interactions) and

(ii) widening of the range of interaction (increasing the number of terms in
the sum). Notice that in the case of polyenes (i.e., for molecules with a finite
size), this range is necessarily determined by the number of bonds forming
the chain.
In conclusion, with the help of the simple Hückel theory applied to poly-
acetylene we can predict:
– the existence of a long-range, conjugation length-dependent, intermolec-
ular potential, involving interaction of CC stretching coordinates
– the consequent softening of two peculiar Raman-active modes which are
characterised by a large content of the bond alternation oscillation (R−
modes)
– the origin of the effective conjugation force constant FR− in terms of
electron–phonon coupling
It is important at this point to stress that one of the key points in the physics
of polyacetylene is the existence of an energy gap due to Peierls distortion of
the 1D crystal structure [33]. In terms of Hückel theory this gap opening is
just expressed as:
∂β
E gap = 2 · u = 2∆β, (5)
∂r
where u represents the equilibrium bond length alternation, i.e., the “dis-
tortion” (along R− ) from a perfectly “equalised” metallic structure to a
“dimerised” equilibrium structure. In other words, (5) states that a pecu-
liar trajectory in the vibrational space (namely the trajectory described by
R− ) does exist, which is coupled to π electrons to such extent that it deter-
mines a transition between a metal (ideal polyacetylene, with u = 0) and a
semiconductor. According to this statement, it is obvious why ECCT works
so well. Indeed, in ECCT the explicit introduction of the R− coordinate in
the description of the molecular dynamics allows one to “summarize” in just
one parameter (FR− ) the effect of the electron–phonon interaction and its
dependence on the conjugation length of the chains.
The evidence of bond alternation is also preliminary to any discussion
regarding Raman activities of R− modes. According to DFT calculations
performed on many simple polyenes of different length [14], we demonstrated
that local Raman parameters (polarizability changes associated with the
stretching of individual CC bonds) are characterised by opposite sign ac-
cording to the character of the CC bond considered (single or double), as
shown in Fig. 2. This sign rule has the effect of determining a huge polariza-
tion change with the bond alternation oscillation, namely during R− modes.
On the other hand, if we ideally remove bond alternation, we expect to obtain
a vanishing polarization change during R− according to the new vibrational
selection rules (the crystal symmetry at k = 0 passes from C 2h to D2h ,
and R− modes become infrared-active and silent in the Raman spectrum).
This fact can also be explained in terms of local stretching Raman parame-
ters which, because of the symmetry, are forced to be equal (same sign and
value!) for all the CC stretching of the “metallic” chain.
3.2 Electron–Phonon Coupling in Graphenes
The idea is now to extend the simple treatment illustrated in Sect. 3.1 to the
case of graphenes. In [2] the Kakitani approach has been applied to a perfect
graphite sheet (2D crystal) and to several PAH molecules, seen as oligomers of
2D graphite. The force field used is that derived by Ohno [34] for small PAHs,
following the same derivation proposed by Kakitani. In the Ohno treatment
long-range interaction CC stretching force constants are expressed through
three adjustable parameters according to the following equations:
fii = f1 + f2 (pi − p0 ) + f3 (Πii − Π0 ) ,

fij = f3 Πij i = j, (6)
where p0 and Π0 are the “reference” bond order and self-polarizability of the
benzene molecule. Equation (6) can be put in correspondence with Kakitani
formulas to obtain (7).
2
∂2β ∂β
f1 = kσ + 2 2 p0 + 2 Π0 .
∂R ∂R
∂2β
f2 = 2 2 .
∂R
2
∂β
f3 = 2 . (7)
∂R
The important point is to notice that the constant f3 is determined by

the value of electron–phonon coupling parameter ∂β/∂R. Also in the case
of Ohno field, the range of interaction (and its decreasing law with bond
distance) is then determined by the mutual polarizabilities, as obtained in
the frame of the Hückel theory. In the case of graphite these terms have to
be calculated taking advantage of the translational symmetry, as illustrated
in [2] and starting from (8):
π π π π
1 fλµ,νσ (θ1 , θ2 , θ1 , θ2 )
Πλµ,νσ = dθ 1 dθ 2 dθ , θ ) dθ2 + c.c.
4
(2π) −π −π −π
1
−π ε o (θ 1 , θ 2 ) − ε e (θ 1 2
fλµ,νσ (θ1 , θ2 , θ1 , θ2 ) = gλµ (θ1 , θ2 , θ1 , θ2 )hνσ (θ1 , θ2 , θ1 , θ2 )
gλµ (θ1 , θ2 , θ1 , θ2 ) = [Coλ
∗
(θ1 , θ2 )Ceµ (θ1 , θ2 ) + Coµ ∗
(θ1 , θ2 )Ceλ (θ1 , θ2 )]
hνσ (θ1 , θ2 , θ1 , θ2 ) = [Coν (θ1 , θ2 )Ceσ
∗
(θ1 , θ2 ) + Coσ (θ1 , θ2 )Ceν
∗
(θ1 , θ2 )] . (8)
Here, θ1 and θ2 are the phases associated with displacements along a1 and
a2 , respectively, in the 2D graphite crystal (θ1,(2) = 2πk1,(2) ; k = k1 b1 +k2 b2 ,
where b1 and b2 are primitive vectors of the reciprocal 2D lattice of graphite).
The use of the complete set of adjustable Ohno parameters gave the pho-
non dispersion curves published in [2]. In Fig. 3 we illustrate the effect of the
change of the electron–phonon coupling constant (through the f3 parameter)
on the dispersion relation of graphite phonons. The f3 value as refined by
Ohno [34] and used in [2] corresponds to the red plots.
It is apparent that the more relevant effect obtained by changing the value
of f3 is on the higher frequency branch near K point, to which belong the so-
called A phonon of A1 symmetry at K (Fig. 4). According to the “molecular
approach” described in [2, 3, 4, 5, 6], this vibration has been associated with
the strong D line, shown by the Raman spectra of defected graphites and
disordered carbon phases.
Moreover, the phonon curve showing strong electron–phonon coupling
dependence in Fig. 3 has been indicated as that involved in the double-
resonant process proposed by Thomsen [35]. This same phonon branch has
been recently investigated by high-level DFT calculations, which revealed a
Fig. 3. In plane phonon dispersion curves obtained for a 2D graphene crystal

according to Ohno force field [2]. Different colors correspond to different choices
of the f3 parameter (different electron–phonon coupling constant). Blue: f3 =
2.000 mdyne/A; green: f3 = 3.000 mdyne/A; red: f3 = 3.646 mdyne/A
strong Kohn anomaly, driven by electron–phonon coupling [36]. According to

these observations we can state that Hückel theory applied to the construction
of long-range interaction valence potential of graphite immediately indicates
that the totally symmetric phonon at K is heavily affected by interactions
with π electrons. This fact suggests that the selective and strong activation
of A modes (D line) in the Raman spectra of graphite samples containing
some disorder (e.g., microcystalline graphites, see [11], or confinement, as in
PAH molecules) has its origin in a mechanism similar to that found in the
case of R− modes of polyacetylene and polyenes.
In a recent paper [37] we have shown an alternative, independent demon-
stration of the efficient electron–phonon coupling of the A phonon with π
electrons at K. With a treatment still based on the simple Hückel theory, we
have described the effect on the electronic dispersion curves of a graphene
crystal, of a nuclear relaxation along the A phonon. As illustrated in Fig. 5,
this relaxation (to which a new crystal structure corresponds, described by
a threefold unit cell (six carbon atoms) with CC bond characterized by two
different bond length), is responsible for a gap opening at K point, that is, of
a transition from a metallic to a semiconducting phase. The difference with
respect to the case of polyacetylene is the fact that for graphene, the stable
phase is the metallic one.
C114
ν = 1291 cm−1 I = 51128 A4 /amu
(a) (b)
Fig. 4. (a) Sketch of the nuclear displacements associated with the A phonon
(K wave vector, see text) of graphite. (b) Sketch of the nuclear displacements
associated with the Raman active transition giving rise to the D band of a PAH
molecule with 114 carbon atoms. Eigenvectors, frequency and nonresonant Raman
intensities have been obtained from DFT BLYP 6.31G calculations [4]
In presence of defects (or simply in presence of confinement) the graphene

structure locally relaxes (preferentially in the direction of the maximum elec-
tron–phonon coupling), giving rise to a new crystal structure characterised
by a threefold cell. This observation is corroborated by equilibrium structure
predictions on model molecules (PAHs) [6], but also by experimental evidence
from STM analysis of graphite near to step edges, which indeed shown a pe-
riodic superstructure which can be correlated to the structural relaxations
described above [8].
Another very important result which can be obtained considering the
“dimerised” graphite structure described by the threefold lattice is the con-
sequence of the lattice relaxation on the Raman selection rules. It can be
easily seen according to the standard zone folding procedure that the K point
of the usual (nonrelaxed) graphene is “folded” on Γ if the lattice relaxes ac-
cording to the threefold cell. This implies that the totally symmetric phonon
at K becomes symmetry-allowed for Raman transition, thus giving an ele-
gant explanation for the activation of the D line as a consequence of disorder
(confinement) induced relaxation of the lattice [5].
Another very impressive analogy with the case of polyacetylene can be
found considering DFT studies carried out on molecular species (PAHs) [4]
aimed at the understanding of their Raman activity in the D band region
(A modes). Also in this case the Raman stretching parameters obey a sign
2(x-1) x x+1 2(x-1) x x+1
3 3 3 3
2 2 2 2
1 1 1 1
[ε - α] / β0
[ε - α] / β0
0 0 0 0
-1 -1 -1 -1
0.6
0.4
-2 -2 -2 -2 0.2
0.0
-0.2
-3 -3 -3 -3
K Γ M K K Γ M K -0.4
Γ1 M1 Γ1 K1 M1 Γ1 Γ1 M1 Γ1 K1 M1 Γ1 -0.6
(a) -2 -1.5 -1 -0.5 0 0.5 1 1.5 2 0 DOS 0 DOS
(b) -2 -1.5 -1 -0.5 0 0.5 1 1.5 2 0
6 5 6 5
1 4 1 4
2 3 a1’ 2 3
Γ
6 5 6 5 6 5
1 4 1 4 1 4 b1
M
2 3 2 3 2 3 K
Γ b1’ Γ
6 5 a2’ 6 5 M
1 4 1 4 b2’
2 3 2 3 K
M
b2
(c) (d)
Fig. 5. (a) Blue curves: electronic structure (π bands) of “regular” graphite (iden-
tical hopping integral β0 for any CC bonded pair) as obtained according to Hückel
theory and considering the conventional graphite cell (2 atoms). Red curves: Same
calculation (identical hopping integral β0 for any CC bonded pair) referred to a
threefold cell (see text). (b) Electronic structure (π bands) of dimerised graphite
(two different hopping integrals β1 and β2 for CC bonded pair belonging to the
two different classes originating by the deformation along the A phonon). The cell
considered is the minimal one for the distorted structure with shorter (heavy lines)
and longer (light lines) CC bonds (threefold cell, 6 carbon atoms, see sketch (c)).
The first Brillouin zone of the “regular” graphite and of “dimerised” graphite are
represented in the sketch (d)
rule determined by the bond order of the CC considered. CC bonds in the

molecules are indeed arranged according to two different classes. In the case
of the PAHs characterised by armchair topology of the edges, this “bond
dimerisation” gives rise to an equilibrium structure which obeys the so-called
Clar’s rule [38] (see Fig. 5c): shorter bonds form six-membered rings and show
positive stretching polarizablity parameters; the longer ones, connecting the
benzenoid rings, show negative stretching polarizability parameters. During
A vibration (which can be described as a collective breathing of the benzenoid
rings, Fig. 4b) bonds belonging to a given set (i.e., the shorter ones, forming
the rings) oscillate out of phase with respect to (longer) bonds of the second
set, with a consequent cooperation in building up a very large change of the
total molecular polarizability.
Notice that this sign rule cannot be “transferred” to the case of the perfect
crystal of graphite because of symmetry: equilibrium structure of graphite
is indeed characterised by identical CC bond length (metallic structure) for
which the A oscillation is at K point (and then it is forbidden in the Raman).
On the other hand, structure relaxation (as, for instance, induced by defects
and/or confinement) can make active this peculiar phonon by virtue of the
consequent symmetry decrease. The experimental fact that the activation of
the A phonon is so selective (no other new and strong features are observed in
graphites with defects) strongly supports the existence of a selective electron–
phonon coupling and of a preferential path of relaxation of the atoms toward
a structure which can be schematically described in terms of the threefold
lattice illustrated above.
Another look at Fig. 3 indicates that also the phonon responsible for the
G band is coupled with π electrons: a softening of the Raman frequency
with f3 is indeed observed also for the relative phonon branches, near to the
Γ point. In the next section we will report on some preliminary results on
carbon nanotubes where this coupling is predicted to determine very peculiar
features.
3.3 Electron–Phonon Coupling in Carbon Nanotubes
The many works dealing with Raman spectra of single-wall carbon nanotubes
(SWNTs) always focus attention on the two major features of the first-order
spectrum:
– the so-called radial breathing modes (RBM) bands, which belong to the
lower frequency region of the spectrum (usually observed between 400 and
100 cm−1 ) and show clear, widely discussed and well rationalised fre-
quency dependence on the nanotube diameter and species [39, 40]
– the region near 1600 cm−1 , commonly ascribed to phonons related to the
G band (Γ = 0, E 2g symmetry) of a perfect graphite crystal
Experiments on individual SWNT showed that the frequency degeneracy of
the G line (dictated by symmetry in the case of graphite) is removed (as
expected by the symmetry lowering) in nanotubes, giving rise to two compo-

nents, usually referred as “G+ ” and “G− ” bands. While the peak position of
the higher frequency component (G+ ) seems practically insensitive to tube
diameter and chirality, the G− peak shows a frequency decrease while de-
creasing the tube diameter. Moreover, according to [41] a markedly different
trend for G− is observed in the case of semiconducting carbon nanotubes
with respect to the case of the metallic ones, which show a more pronounced
frequency softening while decreasing the tube diameter. It is difficult to imag-
ine a vibrational force field which is able to account for these experimental
features, using the same force constant values for different nanotubes.
In order to be quantitative about this point, we have computed the phonon
frequency of many different SWNTs [42] by transferring the valence force field
previously derived from graphite [2]. Calculations were performed on the basis
of internal vibrational coordinates within the GF Wilson’s formalism [43],
taking into account the real “curved” geometry and the full symmetry of any
tube. Symmetry considerations allow us to treat the dynamical problem in
terms of phonon coordinates built on the basis of the few degree of freedom
of the minimal structural unit, which coincides with the graphene cell (only
two carbon atoms). On one hand, the advantage of the treatment lies in the
fact that phonon frequencies and eigenvectors can be obtained through the
diagonalization of a small dynamical matrix (i.e., a 9×9 problem, based on the
9 valence coordinates of the structural unit). On the other hand, and even
more important, we obtain a formally identical problem (same dimension,
identical mathematical structure) for any nanotube (characterised by the
indexes n, m) we would like to consider.
According to our treatment, the phase difference between structural units
along the tube are described in terms of a pair of two phase parameters (θ1 , θ2 )
which have exactly the same meaning as for graphite (see Sect. 3.2). However,
θ1 and θ2 now obey well-defined quantization relationships, dictated by the
structure of the nanotube considered, according to (9):
k · C h = 2πµ
k · T = ξ, (9)
where k is the wave vector in the reciprocal space, C h the chiral vector
(C h = na1 + ma2 , for a n, m nanotube), T is the proper translation vector
along the tube axis. θ1 and θ2 can be written in the form described by (10):
2π(2n + m)µ + mdR ξ
θ1 = .
2(n2 + nm + m2 )
2π(2m + n)µ − ndR ξ
θ2 = . (10)
2(n2 + nm + m2 )
Notice that the index µ can assume only integer values (µ = 0, 1, . . . , 2(n2 +
nm + m2 )/gcd((2n + m), (2m + n)), while value ξ assumes rational values be-
tween −π and π.
Explicit calculations of the (k = 0) phonon frequencies in the 1500–

1600 cm−1 range done using a force field with common parameters for many
nanotubes of different chiralities clearly show the degeneracy removal of the
G line frequency [42]. This is the direct consequence of the introduction of
the appropriate geometries of the nanotubes considered. However, the results
obtained with a common field do not allow us to account for the experimental
frequencies trends observed for G+ and G− lines by Jorio et al. [41]: the
predicted range of frequency variation results indeed are very narrow (about
10 cm−1 ) in comparison with the experimental findings (showing changes of
G− frequencies in a range about 80 cm−1 wide).
On the other hand, if correctly applied, the simple Kakitani definition of
CC stretching force constants (1) and (9) gives the possibility of adapting the
field to any desired carbon structure and, in particular, to any SWNT. What
we have to do is to obtain the coefficient of the π orbitals (in the form of
Bloch functions) for any carbon nanotube with tube indexes (n, m). As in the
case of vibrational coordinates, the tube symmetry can be fully introduced
for the determination of the electronic structure, thus reducing the Hückel
Hamiltonian to the same dimensions (2 × 2) as for graphite. As described
above, the important point is to remember that θ1 and θ2 assume now dif-
ferent values for any nanotube considered, as prescribed by (10). In other
words, this fact (i.e., “selection” of special θ1 and θ2 values) can be described
following the more usual solid state physics language: we have performed a
“sampling” (due to the introduction of the new boundary conditions of nano-
tubes, i.e., due to confinement in one dimension) of the electronic π states of
graphite.
In this light, it is not surprising to find that mutual polarizabilities be-
tween bonds at the same distance have different values while changing the
tube structure (i.e., by changing the n and m parameters appearing in (10)).
An explicit expression for Πij can be worked out for any carbon nanotube
according to the following equation:
N π
1 1 π
Πi(0,0) ,j(n1 ,n2 ) = dξ dξ πij
n1 ,n2
. (11)
(2π)2 N 2 −π −π
µ,µ
The function Πij n1 ,n2

is a known function of µ, µ and ξ, ξ , through θ1 ,

θ2 , θ1 , θ2 . If we compare (11) with the corresponding one in the case of
graphite (8), we immediately realize that two of the four integrals are now
replaced by simple sums over the indexes µ, µ . This indeed corresponds to
“moving” along discrete lines (one for each µ value allowed) in the first BZ of
graphite, the path of integration being parametrically described by the vari-
able ξ(ξ ). Values of Πij obtained by numerical integration according to (11)
allow us to extend in a fully consistent way the Ohno force field already
applied to graphite (see Sect. 3.2) to any SWNT.
However, despite the relative simplicity of the theory, problems arise from
the fact that we have to establish a threshold distance of interaction d: this
Fig. 6. Raman frequencies of G+ and G− frequencies of nanotubes with decreasing

diameters, as predicted according to the valence force fields described in the text
(generalized Ohno fields). Open circles: armchair (metallic) nanotubes; filled circles:
metallic zig-zag nanotubes; filled triangles: metallic chiral nanotubes; open squares:
semiconducting zig-zag nanotubes; filled squares: semiconducting chiral nanotubes
implies to arbitrarily put to 0 bond–bond interactions involving CC pairs

at distance farther than d. The choice of d is indeed strictly related to the
degree of numerical accuracy which we are able to reach in the calculation
of integrals. Notice, moreover, that the functions which we have to integrate
are oscillating functions, which undergo faster and faster oscillations as the
distance of the bonds increases.
In light of the above observations, we present the data collected in Fig. 6
as a very preliminary results. On the other hand, the trends of G+ and
G− frequencies obtained with our generalised force fields seem to be a very
promising beginning of a more systematic study. A wide dispersion in fre-
quency while decreasing the tube diameters has been obtained; moreover, a
different behaviour for metallic and semiconducting nanotubes is predicted.
An important piece of information which we can extract from these re-
sults is the fact that the same electron–phonon coupling parameter (for all the
nanotubes considered, f3 has been fixed to the empirical value proposed by
Ohno) works in different ways in the presence of different electronic structures
(described by bond–bond polarizabilities calculated on the basis of the suit-
able π electron states) characteristic of the nanotube considered. The same
conclusion about the peculiar role of the specific electronic structure of any
carbon nanotube in determining the vibrational potential was first stressed
in [44] and illustrated by first-principle calculations on armchair carbon nano-
tubes. In a very simple way, easily generalised to any nanotube structure, our
treatment clearly states the inadequacy of the usual zone folding procedure
often followed in the past to obtain the description of phonons in SWNTs
directly from the dispersion curves of graphite.
4 Conclusions
A common simple treatment of the harmonic vibrational potential for π con-

jugated systems consisting of sequences of CC bonds arranged in different
topologies can be obtained in the frame of Hückel theory for pz electrons. Fol-
lowing this idea, we have extended the original Kakitani [7] equations (first
derived for molecular systems and then generalized by Piseri et al. to the
case of polyacetylene [31]) to the cases of graphite and of carbon nanotubes.
Notice that the extension of Hückel theory to infinite systems corresponds
to the one electron tight-binding effective Hamiltonian for π electrons, in the
language of solid state physics.
This formalism allows us to explicitly consider the role of the electron–
phonon coupling in determining the Raman response of a wide family of
carbon materials. In the theory, the electron–phonon coupling is described
by just one parameter, namely ∂β/∂R, and its effect on vibrational potential
is modulated by the characteristic electronic properties of the material con-
sidered, namely through bond–bond mutual polarizabilities. In this way the
model provides a common ground, useful for the comparison of the experi-
mental Raman features observed for many different systems.
We briefly summarize here the main results obtained for the materials
discussed in this work:
1. Electron–phonon coupling in polyenes and polyacetylene gives rise to:
– long-range CC stretching interactions and softening of R− modes
with conjugation length
– bond alternation in the equilibrium structure of PA
– selective enhancement of the Raman intensity of the R− modes
2. Electron–phonon coupling in graphite, polycyclic aromatic hydrocarbons
and related carbon materials is responsible for:
– long-range CC stretching interactions and softening of A modes with
π electron delocalisation
– bond “alternation” (e.g., appearance of benzeniod-like relaxed struc-
tures) in the presence of confinement
– selective enhancement of the Raman intensity of the D line in the
presence of confinement (A modes)
3. The electron–phonon coupling in single-wall carbon nanotubes gives rise
to long-range CC stretching interactions modulated by the different elec-
tronic structures. As a consequence, different dispersion laws for G- bands
can be predicted on this basis, in agreement with the experimental ob-
servations.
Preliminary calculations on SWNTs indicate that bond orders are also

affected by confinement and symmetry decreases, thus removing the perfectly
“equalized” bond structure of graphite. Further investigation on this point,
supported by high-level first-principle calculations, would give new insight
about the mechanism which rules Raman activity and its dependence upon
the symmetry of the nanotube and its characteristic electronic structure.
Work is in progress along this line.
Acknowledgements
The authors are indebted to F. Mauri, A. Ferrari and S. Piscanec for the
very useful discussions about electronic structure and vibrational potential
of carbon nanotubes, and to C. Vergara for his assistance in the setup of
the numerical methods for integral calculations. This work was supported
by a grant from MURST (Italy) (FIRB project “Carbon-based micro- and
nanostructures”, RBNE019NKS).
References
[1] M. D. Eatson, A. Fechtenkötter, K. Müllen: Chem. Rev. 101, 1267 (2001) 382
382
[3] C. Castiglioni, C. Mapelli, F. Negri, G. Zerbi: J. Chem. Phys. 114, 963 (2001)
382
[4] F. Negri, C. Castiglioni, M. Tommasini, G. Zerbi: J. Phys. Chem. A 106, 3306
(2002) 382
[5] C. Castiglioni, F. Negri, M. Rigolio, G. Zerbi: J. Chem. Phys. 115, 3769 (2001)
382
[6] E. D. Donato, M. Tommasini, G. Fustella, G. Brambilla, C. Castiglioni,
G. Zerbi, C. D. Simpson, K. Müllen, F. Negri: Chem. Phys. 301, 81 (2004)
382
[7] T. Kakitani: Progr. Theor. Phys. 51, 656 (1974) 382
[8] M. Tommasini, E. D. Donato, C. Castiglioni, G. Zerbi, N. Severin, T. Böhme,
J. Rabe: Proceedings of the XVIII International WintherSchool “Electronic
Properties of Novel Materials” IWEPNM2004 (American Institute of Physics,
New York 2004) pp. 334–338 383
[9] H. E. Shaffer, R. R. Chance, R. J. Silbey, K. Knoll, R. R. Shrock: J. Chem.
Phys. 94, 4161 (1991)
[10] H. Harada, Y. Furukawa, M. Tasumi, H. Shirakawa, S. Ikeda: J. Chem. Phys.
73, 4746 (1980)
[11] I. Pöcsik, M. Hunfhausen, M. Koós, L. Ley: J. Non-Cryst. Solids 1083, 227–230
(1998)
[12] C. Thomsen, S. Reich: Phys. Rev. Lett. 85, 5214 (2000)
[13] F. Negri, E. D. Donato, M. Tommasini, C. Castiglioni, G. Zerbi, K. Müllen:
J. Chem. Phys. 120, 11889 (2004)
[14] C. Castiglioni, M. Tommasini, G. Zerbi: Theme Issue of the Philosophical

Transactions of the Royal Society A on “Raman Spectroscopy in Carbons from
Nanotubes to Diamonds”, vol. 362 (2004) pp. 2425–2459
[15] E. J. Mele, M. J. Rice: Solid State Commun. 34, 339 (1980)
[16] M. Gussoni, C. Castiglioni, G. Zerbi: Spectroscopy of Advanced Materials (Wi-
ley, New York 1991) pp. 251–353
[17] F. Inagaki, M. Tasumi, T. Miyazawa: J. Raman Spectrosc. 3, 335 (1975)
[18] G. Zerbi, G. Zannoni: J. Phys. C 44, 3–273 (1983)
[19] C. Castiglioni, M. Gussoni, J. T. Lopez-Navarrete, G. Zerbi: Solid State Com-
mun. 65, 625 (1988)
[20] S. Hirata, H. Yoshida, H. Torii, M. Tasumi: J. Chem. Phys. 103, 8955 (1995)
[21] S. Hirata, H. Torii, M. Tasumi: J. Chem. Phys. 103, 8964 (1995)
[22] V. Schettino, F. L. Gervasio, G. Cardini, P. R. Salvi: J. Chem. Phys. 110,
3241 (1999)
[23] A. Bianco, M. D. Zoppo, G. Zerbi: J. Chem. Phys. 120, 1450 (2004)
[24] H. O. Villar, J. D. Dupuis, G. J. B. Hurst, E. Clementi: J. Chem. Phys. 88,
1003 (1987)
[25] F. Zerbetto, M. Z. Zgieski, F. Negri, G. Orlandi: J. Chem. Phys. 89, 3681
(1988)
[26] F. Negri, G. Orlandi, A. M. Brouwer, F. W. Langkilde, R. Wilbrandt: J. Chem.
Phys. 90, 5944 (1989)
[27] F. Negri, G. Orlandi, F. Zerbetto, M. Z. Zgieski: J. Chem. Phys. 91, 6215
(1989)
[28] C. A. Coulson, H. C. Longuet-Higgins: Proc. Roy. Soc. A 191, 39 (1947)
[31] L. Piseri, R. Tubino, G. Dellepiane: Solid State Commun. 44, 1589 (1982)
[32] L. Piseri, R. Tubino, L. Paltrinieri, G. Dellepiane: Solid State Commun. 46,
183 (1983)
[33] R. E. Peierls: Quantum Theory of Solids (Clarendon, Oxford 1955)
[34] K. Ohno: J. Chem. Phys. 95, 5524 (1995)
[35] S. Reich, C. Thompsen: Theme Issue of the Philosophical Transactions of the
Royal Society A on “Raman Spectroscopy in Carbons from Nanotubes to Dia-
monds”, vol. 362 (2004) pp. 2271–2288
[36] S. Piscanec, M. Lazzeri, F. Mauri, A. C. Ferrari, J. Robertson: Phys. Rev.
Lett. 93, 185503 (2004)
[37] M. Tommasini, E. D. Donato, C. Castiglioni, G. Zerbi: Chem. Phys. Lett. 414,
166 (2005)
[38] E. Clar: The Aromatic Sextet (Wiley, London 1972)
[39] A. Jorio, R. Saito, G. Dresselhauss, M. S. Dresselhaus: Theme Issue of the
Philosophical Transactions of the Royal Society A on “Raman Spectroscopy in
Carbons from Nanotubes to Diamonds”, vol. 362 (2004) pp. 2311–2336
[40] C. Thomsen, S. Reich, J. Maultzsch: Theme Issue of the Philosophical Trans-
actions of the Royal Society A on “Raman Spectroscopy in Carbons from Nano-
tubes to Diamonds”, vol. 362 (2004) pp. 2337–2359
[41] A. Jorio, A. G. S. Filho, G. Dresselhaus, M. S. Dresselhaus, A. K. Swan, Ünlü,
B. B. Goldberg, M. A. Pimenta, J. H. Hafner, C. M. Lieber, R. Saito: Phys.
Rev. B 65, 155412 (2002)
[42] M. Tommasini, E. D. Donato, C. Corni, L. Motta, C. Castiglioni, G. Zerbi: in

preparation
[43] E. B. Wilson, J. C. Decius, P. C. Cross: Molecular Vibrations: The Theory of
Infrared and Raman Vibrational Spectra (McGraw-Hill, New York 1955)
[44] S. Piscanec, A. C. Ferrari, F. Mauri, M. Lazzeri, J. Robertson: Proceedings of
the XIX International WintherSchool “Electronic Properties of Novel Materi-
als” IWEPNM2005 (American Institute of Physics, New York)
Index
annealing, 382 nanotubes, 381–383, 395–399
armchair nanotubes, 398
confinement, 382–384, 392–395, 397, chirality, 397
399, 400 chiral vector, 396
conjugation length, 383 single-walled nanotubes (SWNTs),
395–397, 399, 400
D band, 384, 391 zig-zag nanotubes, 398
defects, 383, 395
density functional theory (DFT), 386, oligomers, 382, 390
390–393
dimerisation, 386, 390, 393, 394 Peierls distortion, 389
bonds dimerisation, 395 polarizability, 388, 390, 391, 395, 397
dimerisation parameter, 387 bond–bond polarizability, 398
dispersion, 383, 384, 386, 398, 399 bond-bond polarizability, 386, 388
phonon dispersion, 388, 391 polyacetylene, 382–384, 386–390, 392,
393
effective conjugation coordinate theory long chains, 385, 388
(ECCT), 385, 388, 390 short chains, 385
electron–phonon coupling, 382, 385, polycyclic aromatic hydrocarbons
387–393, 395, 398, 399 (PAHs), 382, 390, 393
electron-phonon coupling, 384 polyenes, 383, 385, 386, 389, 390, 392
polymers, 383, 384
first principle calculations, 386, 398,
polyconjugated polymers, 383
400
fullerene, 383
quantum chemistry, 382
G band, 395–398
Raman scattering, 399
graphene, 381–384, 390, 392, 393, 396
Raman activity, 385, 386, 388–390,
Hückel theory, 382, 384, 386, 387, 389, 393, 400
391, 392, 394, 399 Raman intensity, 386
highly oriented pyrolytic graphite resonant Raman scattering, 383, 386
(HOPG), 383 selection rules, 384, 393
Raman spectroscopy, 381, 383–386,
Kohn anomaly, 392 390–392, 395
molecular approach, 384, 391 valence force field, 382, 385–387, 390,
molecular dynamics, 390 392, 396–398
Raman Spectroscopy and Optical Properties
of Amorphous Diamond-Like Carbon Films
Leonid Khriachtchev
Laboratory of Physical Chemistry, University of Helsinki, P.O. Box 55,

FIN-00014 Helsinki, Finland
leonid.khriachtchev@helsinki.fi
Abstract. The Raman spectroscopy and optical characterization of amorphous

hydrogen-free diamond-like carbon films is described. The samples were prepared
by using mass-separated carbon ion beam and pulsed arc discharge deposition meth-
ods. For the samples prepared with both deposition methods, the correlation be-
tween the sp3 fraction and the shape of the Raman spectra was obtained. The
refractive index and absorption of these films were also studied and compared with
the Raman spectra, and the optical properties were found to correlate with the
Raman spectra for the samples prepared with ion beam technique. The effect of
interference of light on Raman spectra was found to be important for the carbon
films with low absorption. The developed characterization methods were applied to
study modifications of carbon film structure upon irradiation with energetic ions
and thermal annealing.
1 Introduction
Studies of diamond-like carbon (DLC) films are popular because this mater-
ial is hard, chemically inert and optically transparent [1]. Hydrogen-free DLC
films have been taken into the focus of research from the early 1990s, and this
material can be prepared by using mass-separated ion beams (MSIB) [2, 3, 4],
pulsed or continuous arc discharge [4, 5], laser ablation [6], and other meth-
ods. The term tetrahedral amorphous carbon (ta-C) is often used to denote
this hydrogen-free material with a high sp3 fraction (η). No microcrystalline
inclusions or layered structure have been found for ta-C. The growth of the
ta-C network is probably a process of subplantation of the energetic par-
ticles into the bulk [7], although the surface growth mechanism was also
proposed [8]. Theoretical simulations of ta-C were performed for materi-
als with η ∼ 80-90% [9, 10]. The theoretically obtained vibrational prop-
erties are mainly characterized by unlocalized low-frequency modes in the
200 to 1300 cm−1 region, and the additional modes with frequencies above
1300 cm−1 are associated with well-localized geometrical anomalies in the
network. A simple structural model contemplates small sp2 -bonded carbon
clusters, embedded in a more transparent sp3 -bonded carbon host.
Raman spectroscopy has been successfully used to characterize DLC
materials. From the beginning, Raman spectra of hydrogenated DLC were
mainly studied [11, 12]. Raman spectroscopy of hydrogen-free DLC has also
404 Leonid Khriachtchev
gained research activity, and Raman spectra of hydrogen-free DLC films pre-
pared by laser deposition [13, 14], MSIB [15, 16], and pulsed cathodic arc
discharge (PCAD) [17] were reported. It was attractive to search for corre-
lation between the Raman spectra and the characteristic film parameters,
and this was found for various hydrogen-free DLC samples with excitation
at 514.5 nm [15, 16, 17, 18, 19]. The studies with UV excitation further de-
veloped the Raman characterization of hydrogen-free DLC films, and the
band around 1100 cm−1 originating from the sp3 -bonded carbon atoms was
demonstrated [20].
Optical properties of hydrogen-free DLC in the visible spectral region have
been studied for DLC films prepared by laser deposition [21], direct current
arc discharge [22, 23], and MSIB methods [24, 25]. For ta-C, the refractive
index is typically 2.4–2.5, approaching the value of crystalline diamond, and
the absorption coefficient is below 104 cm−1 . Raman characterization of thin
films is essentially influenced by their optical properties due to interference
of light. First, interference of light can change Raman signals measured from
thin solid films [26, 27]. Second, interference-induced modifications of the
spectral shape were found for MSIB DLC films [28]. These effects are quite
strong for transparent hydrogen-free DLC films, and they should be taken into
account while characterizing DLC with Raman spectroscopy. The interference
effect on the Raman spectra was used to estimate absorption of hydrogen-free
DLC films [29], and the systematic correlation between the Raman spectra
and the optical properties was found [25].
In the present contribution, we give an overview of some characteriza-
tion approaches and results obtained for hydrogen-free DLC by using Raman
spectroscopy and optical methods. The consideration is mainly limited to the
research performed in the Laboratory of Physical Chemistry at the University
of Helsinki [16, 17, 18, 19, 25, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37].
The hydrogen-free DLC films were prepared by using MSIB and PCAD meth-
ods. The best samples possess sp3 fractions as high as 80%. The films (typ-
ically from 100 to 500 nm thick) were deposited onto either crystalline sili-
con (Si) substrates or aluminum (Al) layers. In order to enhance Raman scat-
tering from surface structures (surface-enhanced Raman scattering, SERS),
some of the DLC films were covered with very thin (∼ 3 nm) silver (Ag) layers.
The deposition methods are described in more detail elsewhere [16,17,18,19].
The Raman measurements were carried out using two experimental set-
ups. The earlier studies used a 1 m Jarrell Ash spectrometer (8 cm−1 reso-
lution) equipped with a cooled low-noise photomultiplier tube and an Ar+
laser (Spectra-Physics, Model 164, ∼200 mW at 514.5 nm). The spectra were
typically recorded in 120–200 points in the region from 100 to 2100 cm−1 .
The second setup included a single-stage spectrometer (Acton SpectraPro
Raman Spectroscopy and Optical Properties of DLC Films 405
1,0
Raman intensity
S = (I1300 - I1100) / I1100 G band
0,5
D band
I500 I1550
0,0
0 500 1000 1500 2000
-1
Raman shift (cm )
Fig. 1. Raman spectrum of a MSIB film (E = 1000 eV) measured with excitation
at 514.5 nm. The Raman parameters used in the analysis are shown
500I, resolution 10 cm−1 ) equipped with a charge-coupled device camera

(Andor InstaSpec IV) and an Ar+ laser (Omnichrome 543-AP, ∼100 mW at
514.5 nm). The laser radiation (in most experiments at 514.5 nm) was focused
to the sample surface, and Raman light was detected in the transverse di-
rection without polarization analysis. Since we did not notice any difference
between the results for various pressure and temperature conditions, most
of the spectra were recorded under normal laboratory conditions. No visible
degradation of the coatings was produced by laser radiation.
For the optical studies, we measured Raman intensity and reflection from
film areas with varying thickness. The studied films were wedged, which was
due to inhomogeneous carbon-ion deposition flux, so that the thickness de-
pendence could be studied by changing the probed point. The film opti-
cal thickness nD can be found by using the fact that the thicknesses of
(2m + 1)λ/4nA and mλ/2nA correspond to minimal and maximal reflection,
respectively, where A is the geometrical parameter depending on the irradi-
ation scheme and λ is the laser wavelength. The optical thicknesses between
the mλ/4nA values are evaluated by extrapolation.
3 Raman Diagnostics
Figure 1 demonstrates the Raman spectrum for a MSIB film, deposited with
ion energy E = 1000 eV, after Si background subtraction. The spectrum
mainly consists of two broad bands around 1550 and 500 cm−1 and some ad-
ditional scattering between them. The asymmetrical peak at about 1550 cm−1
is attributed to well-localized sp2 -bonded clusters residing in the DLC net-
work. The broader band in the 400 to 800 cm−1 range originates from a
wider spatial area of the structure characterized by different sorts of mixed
bonds [9,10,13]. The approach of resonant Raman spectroscopy demonstrated
that this spectrum is the result of Raman scattering, while possible contri-
butions of photoluminescence were small.
In order to describe the Raman spectra numerically, we use the following
Raman parameters (Fig. 1):
(i) The R = I500 /I1550 intensity ratio, where I500 and I1550 are the Raman
intensities at 500 cm−1 and at the main peak maximum, respectively.
Since these Raman intensities correspond to differently bonded carbon
atoms, this parameter is expected to correlate with the sp3 fraction.
(ii) The S = (I1300 − I1100 )/I1100 parameter, which describes an average
slope in the 1100 to 1300 cm−1 interval. This parameter reflects the
shape of the main peak in the spectral region of the D band, typically
located around 1300–1400 cm−1 . Furthermore, Raman spectra with UV
excitation reveal a band around 1100 cm−1 originating from sp3 -bonded
carbon [20], and this contribution probably influences the value of S
measured with the visible excitation.
(iii) The width and the position of D and G bands and the D/G intensity
ratio. These data are conventionally used to characterize carbon ma-
terials, providing good results for hydrogenated or poor hydrogen-free
DLC [11, 12]. However, these parameters are not fully satisfactory in
diagnostics of ta-C when the D band is very weak.
Figure 2 presents these Raman parameters measured from central MSIB
film areas as a function of the deposition ion energy. The minimum of S =
0.02, the maximum of R = 0.65, and the highest frequency of the G bands
occurs for deposition energies from 60 to 300 eV. This energy interval corre-
sponds to the optimal deposition conditions when considering film parameters
such as sp3 fraction, electrical resistance, and roughness [38, 39].
A similar dependence of the Raman parameters on the discharge volt-
age was found for the PCAD films [19, 30]. The best films (voltages from
300 to 500 V) are characterized by R = 0.60 and S = 0.10, the G band is
located at 1560 cm−1 and has a width (FWHM) of ∼240 cm−1 , and the D/G
intensity ratio is ∼0.20. These values only slightly differ from the parame-
ters obtained for the optimal MSIB films. The D/G intensity ratio increases
for high arc voltages (≥ 1 kV). In addition, new spectral features at ∼715
and 450 cm−1 were found for films deposited with the 4.5 kV voltage. This
double-peak low-frequency spectrum corresponds to the vibrational density
of states of graphite, and it is a signature of flat sp2 -bonded carbon regions
and demonstrates an increase of the sp2 -cluster size for higher plasma ion
energy. The double-peak structure in the low-frequency region, and hence
(a)
Intensity ratio, R
0,6
0,4
0,2
0,8
Slope, S
0,4
0,0
(b)
G-band position (cm )
-1
1570 (c)
1550
1530
G-band width (cm )
-1
240
210
180 (d)
1 2 3 4
10 10 10 10
Ion energy (eV)
Fig. 2. Raman parameters as a function of the MSIB energy. The measurements
were performed for central film areas with excitation at 514.5 nm
flat graphite-like layers, is retained in the case of a substrate parallel to the

plasma flow [19].
Raman spectra of the MSIB films were studied with excitation by various
Ar+ -laser lines [18]. On one hand, the Raman spectra of DCL show clear
resonance enhancement. For excitation at 488 nm, the I500 /I1550 intensity
ratio exceeds 0.8 for samples prepared with the optimum deposition energy.
On the other hand, the basic behavior of spectral features demonstrated in
Fig. 2 for 514.5 nm excitation remains qualitatively the same for other laser
wavelengths. The observed resonant behavior corresponds to the 2.0–2.5 eV
optical gap known for ta-C.
Ref. 17
0,6 Ref. 18
Ref. 19
Intensity ratio, R
0,4
0,2
0,0 0,2 0,4 0,6 0,8
3
sp fraction
Fig. 3. Correlation between the sp3 fraction (η) and the I500 /I1550 intensity ra-
tio (R) for MSIB [18] and PCAD [17, 19] films
The sp3 fraction is an important parameter of carbon materials; however,

its extraction based on Raman spectroscopy is indirect. Raman spectra with
visible excitation probe mainly sp2 -bonded carbon clusters due to a much
larger scattering cross section of the graphite network, and the signal es-
sentially depends on the structure and size of the sp2 -bonded clusters. In
principle, different Raman spectra for the same sp3 fraction can be obtained
depending on the method of film preparation. Nevertheless, the experiments
show that the Raman spectra correlate with the sp3 fraction quite straight-
forwardly [17,18,19]. We used the available data on the sp3 fraction to obtain
the Raman parameters as a function of η. The result for the R parameter is
shown in Fig. 3. The presented data allow sp3 fractions to be estimated from
the Raman spectra.
The dependence of R on η can be modeled assuming three phases in the
carbon network that are characterized by bonds between sp2 sites only, mixed
bonds between sp2 and sp3 sites, and bonds between sp3 sites only. If the
I1550 signal mainly originates from the sp2 -bonded sites, we can write
R = a1 + a2 η + a3 η/(η − 1), (1)
where the coefficients a1 , a2 , and a3 characterize the scattering at 500 cm−1

from the sp2 , sp2 -sp3 mixed, and sp3 networks, respectively, and the cor-
responding coefficient responsible for Raman intensity at 1550 cm−1 (sp2
carbon atoms) is equal to 1. The fit of the experimental data by (1) yields a1
= 0.19, a2 = 0.21, a3 = 0.07 for the MSIB films, and a1 = 0.26, a2 = 0.08, a3
= 0.08 for the PCAD films. Although the employed model is very simple, the
obtained result is satisfactory. The fit gives a3 1, which means a smaller
scattering cross-section for the sp3 -bonded carbon atoms as compared with
graphite. A difference in coefficient a2 , which describes the mixed bonds, is
seen for the two deposition methods, which might originate from sufficiently
different short-range order, namely, larger sp2 -bonded carbon clusters for the
PCAD material. Indeed, it is the case of large clusters when mixed bonds do
not contribute much to the Raman scattering.
Differences between Raman parameters obtained for MSIB films with sim-
ilar sp3 fractions but with different deposition energies can be noticed. For
E = 1000 eV, we obtained R = 0.50, S = 0.15, whereas for E = 40 eV the
results are R = 0.35 and S = 0.55 although the sp3 fractions are similar (η =
0.65). This observation can be interpreted assuming that the mean size of
sp2 -bonded clusters is smaller for the higher carbon ion energy. Indeed, the
D/G intensity ratio increases with the cluster size [12], and this corresponds
to a relative increase of S. For smaller clusters, the proportion of mixed bonds
responsible for the low-frequency vibrations increases, leading to a relative
increase of R.
The spectral data shown in Fig. 2 were obtained for central areas of
the MSIB samples. It was found that the Raman parameters change with
the probed area. A decrease of R and an increase of S were systematically
observed towards the sample edges, which reflects a decrease of the sp3 frac-
tion (see Sect. 5) [18, 30]. This behavior was connected with impurities im-
planted into the network during deposition. The effect of impurities is rel-
atively stronger for lower deposition rates, i.e., for edges of the deposition
area.
4 Effects Induced by Interference of Light

The measured Raman intensity depends on the film thickness due to inter-
ference of light reflected from the air–film and film–substrate interfaces. The
local laser intensity controlling the total intensity scattered from a given point
is a function of the distance from this point to the reflecting surfaces. The
Raman light leaving the film is also modified by interference of light. In ad-
dition to the simple amplitude change, a spectral effect generally takes place
because different parts of Raman spectra have different wavelengths; hence
the phase difference between interfering light beams changes with the Raman
shift [28].
In our calculations, the Raman signal from a film with thickness D, ab-
sorption coefficient α, and refractive index n1 deposited onto a substrate with
refractive index n2 was examined. A collimated laser beam is nearly orthog-
onal to the film and backscattering light is collected in a small solid angle.
Raman scattering is assumed to be spherically symmetric and proportional to
the local laser intensity. The phase difference between the interfering beams
Ag-500V-Al
100
Raman intensity, I1550
10
100eV-Al
100eV-Si
0 1 2 3 4 5 6
Film thickness (λ/4n)
Fig. 4. Raman intensity at 1550 cm−1 as a function of the film thickness for a
MSIB film (E = 100 eV) on silicon, a MSIB film (E = 100 eV) on aluminum, and
a Ag-coated PCAD film (V = 500 V) on aluminum. The lines guide the eye
is considered as a function of Raman shift. The experimental angles between

the film surface, the laser beam, and the detection axis are computationally
taken into account. A computational approach to a more complicated opti-
cal system with many layers (superlattice) is described elsewhere [40]. The
present calculations give the Raman intensity as a function of the film opti-
cal properties. The interference effect appears quite strong for thick weakly
absorbing films with a high refractive index deposited onto highly reflecting
substrates. Because of the localization of the surface scattering, its intensity
undergoes much stronger interference-induced modifications as compared to
the signal from the film bulk. For the film thickness D = mλ/2n, the Raman
signal from the surface structure is strongly suppressed, and the bulk signal
is mainly measured. The reasons for this selective discrimination of the sur-
face scattering at this thickness are small laser intensity on the film surface
and out-of-phase return of the Raman light from the substrate. In particu-
lar, it is the case for the half-wave thickness when the middle part of the
film is essentially probed and the Raman signal from the optical interfaces is
suppressed.
Figure 4 shows the experimental Raman intensity I1550 as a function of
film thickness provided by various film areas for three different samples. The
observed oscillating behavior is due to interference of light as described in this
section. The interference effect is stronger for the aluminum-substrate film
than for the sample on silicon. The effect is further enhanced for the sample
covered by a thin silver layer when Raman scattering from the surface is
strong [41]. For the SERS sample, the 1550 cm−1 intensity decreases by a
large factor (∼85) when the film thickness increases from the quarter-wave
to half-wave value. At the half-wave thickness, the contribution of surface-
enhanced Raman scattering is about 30%, while the rest of the signal at
1550 cm−1 originates from the bulk. The ratio of the surface signals at the
half-wave and quarter-wave thickness (Q12 ) is ∼1/200 so that the surface
signal can be, indeed, efficiently suppressed using interference of light.
The film absorption can be extracted by comparing experimental and
computational data on the interference effect if the film refractive index is
known. The refractive index can be found by measuring the corresponding
Brewster angle. Using this method for the central area of the MSIB film
(E = 100 eV) on a Si substrate, we obtain an amplitude absorption coeffi-
cient of (0.60 ± 0.20) · 104 cm−1 at 514.5 nm, which agrees with the data by
other authors [24]. For the silver-covered PCAD film (V = 500 V, η = 77%) on
aluminum, the measured suppression factor Q12 = 1/200 yields α = 3.6 · 104
cm−1 , i.e., the PCAD material is much more absorbing than the MSIB mate-
rial with a similar sp3 fraction. For another SERS sample on silicon (MSIB,
E = 100 eV), the suppression factor Q12 is 0.13, yielding α = 1.2 · 104 cm−1 ,
which corresponds to an averaged absorption coefficient of the two probed
points with D = λ/4n and λ/2n. By employing the intensity ratio between
two other points Q23 = ID=λ/2n /ID=3λ/4n , which is 0.091, we obtain for
the averaged absorption coefficient of this film area α = 0.5 · 104 cm−1 . This
latter result shows an increase of the absorption towards the MSIB film edges.
In addition to the simple amplitude changes, it was experimentally con-
firmed that light interference can modify the shape of Raman spectra [28,29].
This spectral effect was demonstrated by using an Al-substrate sample
(MSIB, E = 100 eV). Figure 5a shows a dramatic difference between two
spectra recorded from adjacent film areas with similar structural properties,
and the only available explanation of this spectral change is based on in-
terference of light. For a similar Si-substrate sample, the observed spectral
modifications are smaller by a factor of ∼5, which agrees with the calcu-
lations. Figure 5b presents the thickness dependence of the I500 /I1550 ratio
measured for the central area of the MSIB sample on Al. The data points
demonstrate very strong interference-induced changes of the Raman inten-
sity ratio. The change of the intensity ratio is found to be very sensitive
to the film absorption. Assuming nearly constant optical properties in the
central film area, we fit the calculated curve to the experimental data with
the absorption coefficient as a parameter. With realistic values n1 = 2.50,
r2 = 0.90 (amplitude reflection), and R = 0.70, the best fit is obtained for
α = (0.6 ± 0.2) · 104 cm−1 , in perfect agreement with our earlier estimates.
A possible contribution of the surface scattering to the ordinary Raman
spectra deserves brief analysis. The interference effect on the signal originat-
ing from surface structures is independent of surface enhancement of Raman
scattering, and the suppression of the surface signal should be similar for films
1,0 D = 3λ/2n
(a)
D = 7λ/4n
Raman intensity
0,5
0,0
0 500 1000 1500 2000
-1
Raman shift (cm )
(b)
Intensity ratio, R
1,2
0,8
0,4
5,5 6,0 6,5 7,0 7,5 8,0

Film thickness (λ/4n)
Fig. 5. (a) Raman spectra measured from two close film areas with different thick-
ness of a MSIB film (E = 100 eV) on aluminum. (b) Raman intensity ratio R as a
function of the film thickness for the same sample. The line is from calculations
with and without silver coatings. At the half-wave thickness, the Raman spec-
tra mainly originate from the film bulk and the surface signal is suppressed.
We analyzed the experimental Raman spectra of MSIB and PCAD films with
varying thickness and found no systematic changes in the spectral shape aris-
ing from the interference-controlled depth selection of the Raman scattering.
All the periodic changes of Raman signals could be explained on the basis
of Raman scattering in the film bulk. It follows that the reported Raman
spectra of DLC films on silicon substrates characterize the bulk structure.
The Raman signal from a substrate is also influenced by interference of
light in the covering DLC film. The laser intensity penetrating into the sub-
strate material and the scattering light leaving the film both depend on the
0,4
2,4 (a)
Refractive index
Thickness (µm)
0,3
2,1
0,2
1,8
0,1
1,5 0,0
4
514.5 nm
3
α (10 cm )
-1
2
4
1 633 nm
(b)
0
0,8
R
Raman parameters
0,6
0,4
0,2
S
(c)
0,0
-4 -2 0 2 4
Displacement (mm)
Fig. 6. Properties of MSIB films (E = 100 eV) as a function of the displacement
from the deposition center. (a) Refractive index and thickness. (b) Absorption co-
efficient at 514.5 and 633 nm. The data at 633 nm were obtained with two methods
(see text for details). (c) Raman parameters (R and S)
film thickness. The light interference effect on Raman signal from a substrate
can be also used to extract film absorption (see Sect. 5).
5 Optical Characterization
In this section, optical properties of DLC films are considered in correla-
tion with the Raman spectra. The optical thickness profiles were obtained
by measuring the positions of maxima and minima of normal reflection [25].
The data obtained for excitation at 488, 514.5, and 633 nm show that the
refractive indexes are similar at these wavelengths. To estimate the refractive
index for various film areas, the corresponding Brewster angles were mea-
sured. The refractive index (accuracy ∼ 2%) versus the displacement from
the deposition center is presented in Fig. 6a by solid circles. The highest re-
fractive index (n = 2.50) is observed for the central film area, and this value
agrees with the refractive indexes of the optimal films prepared with laser ab-
lation [21] and continuous cathodic arc deposition methods [22, 23]. The film
edges possess a lower refractive index (n = 1.80), which is more specific for
hydrogenated DLC [11]. The thickness profile can be obtained as the optical
thickness divided by the refractive index, and the result is shown in Fig. 6a.
The obtained thickness of (380 ± 10) nm at the film deposition center agrees
with the profilometric measurements.
The path length of laser radiation inside the material is known, and the
Brewster-angle reflection estimates the absorption coefficient with the accu-
racy of reflection from the DCL–substrate interface. Using an amplitude re-
flection coefficient at the DCL–Al interface r2 = 0.90, we obtain the amplitude
absorption coefficient presented in Fig. 6b by solid triangles for λ = 633 nm.
The absorption coefficient changes from ∼ 0.2 · 104 cm−1 at the deposition
center to ∼ 2.0 · 104 cm−1 at the 50 nm thick edges.
The film absorption coefficient can be extracted in a different way. For
the three-medium optical system (ambient-film-substrate) the maximum and
minimum reflection coefficients can be expressed analytically in terms of opti-
cal properties. The absorption coefficient is provided by fitting the calculated
values to the experimental data. The extracted values are presented in Fig. 6b
by open symbols for two laser wavelengths (514.5 and 633 nm). The absorp-
tion at these wavelengths differs by a factor of ∼2. The data obtained for
irradiation at the Brewster angles and at 90◦ (λ = 633 nm) are in good
agreement, which confirms the validity of the approaches employed.
In Fig. 6c, the Raman parameters R and S are shown as a func-
tion of the displacement from the deposition center, the data for seven
MSIB (E = 100 eV) samples being averaged. The decrease of R and the in-
crease of S indicate a decrease of the sp3 fraction towards the film edge. The
R parameter behaves similarly to the refractive index, and the S parameter
qualitatively follows the film absorption. The correlation between the absorp-
tion coefficient and the R parameter obtained in these experiments (method
A) is presented in Fig. 7 by open symbols.
The optical measurements were further developed [32]. As mentioned ear-
lier, the Raman signal from a Si substrate depends on the optical properties
of the deposited DLC film. For a MSIB film (E = 100 eV), Fig. 8 presents Ra-
man intensity at 520 cm−1 from the Si substrate and the reflection coefficient
versus the position on the film, the curves being normalized by the signals
outside the film. These data were measured simultaneously using nearly nor-
mal irradiation and collection of light. The extreme values of the signals
correspond to film thickness of λ/n1 , 3λ/4n1 , λ/2n1 , and λ/4n1 . The Raman
signals from the Si substrate and the reflection coefficients at these points
can be written analytically [32]. The expressions for the Raman signal and
100 eV / method A
4
100 eV / method B
30 eV / method B
3
α (10 cm )
-1
2
4
0
0,2 0,3 0,4 0,5 0,6 0,7 0,8
Raman intensity ratio, R
Fig. 7. DLC absorption coefficient at 514.5 nm as a function of the R = I500 /I1550
ratio. The data were obtained using two methods A and B (see text for details)
2,0
Raman intensity
Raman intensity / Reflection
1,5
1,0
0,5
Reflection
0,0
-6 -4 -2 0 2 4 6
Displacement (mm)
Fig. 8. Raman intensity of the silicon substrate at 520 cm−1 and normal reflection
as a function of the displacement from the deposition center. The data are obtained
for a MSIB film (E = 100 eV) on a silicon substrate with excitation at 514.5 nm
reflection coefficient apply to the same sample point, and they can be solved
numerically providing α and n1 with the accuracy of 2% and 10%, respec-
tively. For the example shown in Fig. 8, I520 = 1.65, r = 0.14 at D = 3λ/4n1 ,
and I520 = 0.70, r = 0.91 at D = λ/n1 , which yields α = 0.50 · 104 cm−1 and
n1 = 2.50. For the MSIB samples, the decrease of the deposition energy from
100 to 30 eV leads to an increase of both refractive index and absorption co-
efficient. A similar trend for refractive index to increase when the deposition
energy decreases was observed by Lossy et al. for direct current arc discharge
deposition, and the effect was attributed to variations in the arrangement of
carbon atoms [22]. The increase of absorption for the lower deposition en-
ergy is probably associated with a rise of the sp2 fraction as compared to
the optimum deposition conditions and structural modification of the sp2
carbon clusters. The extracted absorption coefficient is not proportional to
the sp2 fraction, and the change of sp2 fraction from 20% to 30% (E = 100
and 30 eV) increases absorption by a factor of ∼4. The rearrangement of the
carbon network should change the optical gap, and some decrease of the op-
tical gap for small ion deposition energies was observed experimentally for
films prepared by a filtered cathodic arc [22] and MSIB deposition [24]. The
correlation between R and α measured for MSIB films is shown in Fig. 7
(method B), and it agrees with the data provided by method A. For a PCAD
samples (η = 77%), we obtained a relatively strong absorption (α = 4 · 104
cm−1 ). We can speculate that the structure of sp2 clusters is different for
our MSIB and PCAD samples; namely, the larger cluster size takes place in
the latter case. Both absorption coefficient and refractive index of the PCAD
samples show extensive variations that are probably resulted from various
deposition details.
6 Some Applications
The irradiation of carbon materials by energetic ions is of fundamental and
practical value, and it has been repeatedly addressed [33, 34, 35, 42, 43, 44,
45, 46, 47]. We studied modifications of hydrogen-free DLC films by 50 kV
Ga+ focused ion beam [33]. The best PCAD film possessed a sp3 fraction of
80% and mass density of 3.1 g/cm3 . The Raman spectra were found to be
very sensitive to the ion iradiation. The data obtained for the Raman pa-
rameters R and S suggest radiation-induced graphitization of the material.
It was estimated from the Raman spectra that the sp3 fraction decreases to
50% for the 0.01 nC/µm2 dose and to 30% for the 0.02 nC/µm2 dose. Upon
ion irradiation, the D/G intensity ratio increases and the G band shifts to
lower energies (Fig. 9a). The increase of the D/G intensity ratio features
rising sp2 -bonded clusters, however, the material remains amorphous in the
whole range of doses used. After ion irradiation, the signal from the Si sub-
strate considerably decreases reflecting an increase of film absorption due to
2
1580
(a)
1560 1
1540
G-band position (cm )
-1
D/G intersity ratio

0
0,00 0,05 0,10
2
Dose (nC/µm )
1600 4
(b)
1580
2
1560
0
0 250 500 750 1000
o
Annealing temperature ( C)
Fig. 9. G band position and D/G intensity ratio as a function of (a) dose of 50 kV
Ga+ ions and (b) annealing temperature
rising sp2 -bonded imperfections. It was estimated that the absorption coeffi-
cient increases for a dose of 0.1 nC/µm2 by a factor of 12 as compared with
the as-grown material. The properties of strongly irradiated DLC materials
(> 0.1 nC/µm2 ) are quite independent of their original structure. In agree-
ment with the observed irradiation-induced decrease of the G-band energy,
similar mode softening was observed for MSIB films when very energetic
carbon ions (5 keV) were used for deposition, leading to destruction of the
carbon network during its formation [16]. A downshift of the G band was
observed upon Si implantation into DLC [35].
Thermal stability of DLC is a practically important property [34, 48, 49].
The Raman spectra of DLC films (initially η ∼ 45%) display progressive
graphitization when the annealing temperature increases, as suggested by an
increase of both G band frequency and D/G intensity ratio (Fig. 9b) [34].
The R parameter decreases for higher annealing temperatures indicating an
efficient reduction of the sp3 fraction. The thermal mobility of various atoms
in the carbon network (for instance, Si [35]) was studied at high anneal-
ing temperatures (above 1000◦ C). The observed annealing-induced degrada-

tion of diamond-like properties means that the obtained diffusion coefficients
practically apply to a sp2 -bonded carbon structure. The elevated deposition
temperature (∼ 400◦C) leads to an essentially graphitic structure [34].
The carbon materials doped with various atoms are more complicated to
analyze based on Raman spectroscopy, with the possible exception of hydro-
gen [11, 12]. Our studies in this area include hydrogenated DLC [37], carbon
materials doped with nitrogen [31], argon [30], silicon [34, 35], and B–C–N
coatings [36]. For example, it was found that the extensive doping of the car-
bon network during deposition with Si (33 at. %) makes the structure less
sensitive to the annealing and deposition temperature and ion-beam irra-
diation [34]. Doping with nitrogen leads to a strong photoluminescence [31],
which further complicates the Raman analysis. Some additional uncertainty is
introduced in the case of high-dose implantation, as discussed elsewhere [33].
7 Conclusions
Raman spectroscopy can be successfully used to characterize structure of
amorphous hydrogen-free DLC films. The correlation of the Raman spec-
tra with the deposition energy, and hence with the sp3 fraction, has been
found, demonstrating the quantitative Raman diagnostics of this material.
The developed characterization method essentially uses the Raman intensity
at ∼ 500 cm−1 .
Raman spectra of DLC films are modified due to interference of light.
The interference changes numerical parameters of the Raman spectra, which
influences the diagnostics of the microstructure and this effect should be
always carefully considered. The use of interference of light allows selective
enhancement of Raman scattering from the bulk structure or from the film
surface. This analysis shows that the reported Raman spectra essentially
originate from the film bulk structure.
By analyzing the thickness dependence of Raman intensity and reflection,
we estimated the absorption coefficient and refractive indexes of various DLC
films. The Raman scattering from the substrate can be also employed for this
analysis. For the optimal DLC materials, we found a refractive index of ∼ 2.5
and an absorption coefficient of ∼ 0.2 · 104 cm−1 (at 633 nm).
For MSIB films, the Raman spectra show a straightforward correlation
with the absorption coefficient, which indicates the close connection between
the optical and structural properties. In particular, the most transparent films
have the highest sp3 fractions. For PCAD films, relation between the Raman
spectra and the optical properties is less clear.
The developed methods were applied to study structural modifications of
DLC films upon thermal annealing and ion irradiation. Some data can be
obtained for doped carbon materials, but the interpretation of the results
becomes less straightforward.
Acknowledgments
The work was supported by the Academy of Finland and the Center for
International Mobility (Finland). I would like to thank all my collaborators
in these studies, especially A. Stanishevsky, R. Lappalainen, E. Vainonen-
Ahlgren, J. Likonen, T. Ahlgren, M. Räsänen, and J. Keinonen.
References
[1] Y. Lifshitz: Diamond Relat. Mater. 5, 388 (1996) 403
[2] J. Ishikawa, Y. Takeiri, K. Ogawa, et al.: J. Appl. Phys. 61, 2509 (1987) 403
[3] Y. Lifshitz, S. R. Kasi, J. W. Rabalais: Mater. Sci. Forum 237, 52–53 (1989)
403
[4] J. P. Hirvonen, J. Koskinen, R. Lappalainen, et al.: Mater. Sci. Forum 197,
52–53 (1989) 403
[5] I. I. Aksenov, V. I. Strelnitskij: Surf. Coat. Technol. 47, 98 (1991) 403
[6] J. Robertson: Surf. Coat. Technol. 50, 185 (1992) 403
[7] Y. Lifshitz, S. R. Kasi, J. W. Rabalais: Phys. Rev. Lett. 62, 1290 (1989) 403
[8] N. A. Marks, D. R. McKenzie, B. A. Pailthorpe: Phys. Rev. B 53, 4117 (1996)
403
[9] C. Z. Wang, K. M. Ho: Phys. Rev. Lett. 71, 1184 (1993) 403, 406
[10] D. A. Drabold, P. A. Fedders, P. Stumm: Phys. Rev. B 49, 16415 (1994) 403,
406
[11] M. Yoshikawa: Mater. Sci. Forum 365, 52–53 (1989) 403, 406, 414, 418
[12] M. A. Tamor, W. C. Vassell: J. Appl. Phys. 76, 3823 (1994) 403, 406, 409,
418
[13] W. S. Bacsa, J. S. Lannin, D. L. Pappas, et al.: Phys. Rev. B 47, 10931 (1993)
404, 406
[14] F. Xiong, Y. Y. Wang, W. Leppert, et al.: J. Mater. Res. 8, 2265 (1993) 404
[15] S. Prawer, K. W. Nugent, Y. Lifshitz, et al.: Diamond Relat. Mater. 5, 433
(1996) 404
[16] L. Y. Khriachtchev, M. Hakovirta, R. Lappalainen, et al.: Phys. Lett. A 217,
354 (1996) 404, 417
[17] A. V. Stanishevsky, L. Y. Khriachtchev: Diamond Relat. Mater. 5, 1355 (1996)
404, 408
[18] L. Y. Khriachtchev, M. Hakovirta, R. Lappalainen, et al.: Diamond Relat.
Mater. 6, 694 (1997) 404, 407, 408, 409
[19] A. V. Stanishevsky, L. Y. Khriachtchev, R. Lappalainen, et al.: Diamond Relat.
Mater. 6, 1026 (1997) 404, 406, 407, 408
[20] K. W. R. Gilkes, H. S. Sands, D. N. Batchelder, et al.: Appl. Phys. Lett. 70,
1980 (1997) 404, 406
[21] D. L. Pappas, K. L. Saenger, J. Bruley, et al.: J. Appl. Phys. 71, 5675 (1992)
404, 414
[22] R. Lossy, D. L. Pappas, R. A. Roy, et al.: J. Appl. Phys. 77, 4750 (1995) 404,
414, 416
[23] S. Xu, B. T. Tay, H. S. Tan, et al.: J. Appl. Phys. 79, 7234 (1996) 404, 414
[24] Y. Lifshitz, G. D. Lempert, E. Grossman, et al.: Diamond Relat. Mater. 6, 687

(1997) 404, 411, 416
[25] L. Y. Khriachtchev, M. Räsänen, R. Lappalainen: J. Appl. Phys. 82, 413 (1997)
404, 413
[26] R. J. Nemanich, C. C. Tsai, G. A. N. Connell: Phys. Rev. Lett. 44, 273 (1980)
404
[27] A. H.-L. Goff: Thin Solid Films 142, 193 (1986) 404
[28] L. Y. Khriachtchev, M. Räsänen, R. Lappalainen: J. Appl. Phys. 79, 8712
(1996) 404, 409, 411
[29] L. Y. Khriachtchev, R. Lappalainen, M. Räsänen: Diamond Relat. Mater. 7,
1451 (1998) 404, 411
[30] L. Y. Khriachtchev, R. Lappalainen, M. Hakovirta, et al.: Diamond Based
Composites and Related Materials, vol. 3, NATO ASI Series 38 (1997) pp. 309–
321 404, 406, 409, 418
[31] A. Stanishevsky, L. Khriachtchev, I. Akula: Diamond Relat. Mater. 7, 1190
(1998) 404, 418
[32] L. Y. Khriachtchev, R. Lappalainen, M. Räsänen: Thin Solid Films 325, 192
(1998) 404, 414
[33] A. Stanishevsky, L. Khriachtchev: J. Appl. Phys. 86, 7052 (1999) 404, 416,
418
[34] L. Khriachtchev, E. Vainonen-Ahlgren, T. Sajavaara, et al.: J. Appl. Phys. 88,
2118 (2000) 404, 416, 417, 418
[35] E. Vainonen-Ahlgren, T. Ahlgren, L. Khriachtchev, et al.: J. Nuclear Mater.
290, 216 (2001) 404, 416, 417, 418
[36] A. Stanishevsky, H. Li, A. Badzian, et al.: Thin Solid Films 398, 270 (2001)
404, 418
[37] K. Heinola, T. Ahlgren, W. Rydman, et al.: Physica Scripta T 108, 63 (2004)
404, 418
[38] Y. Lifshitz, G. D. Lempert, E. Gossman, et al.: Diamond Rel. Mater. 4, 318
(1995) 406
[39] M. Hakovirta, J. Salo, R. Lappalainen, et al.: Phys. Lett. A 205, 287 (1995)
406
[40] L. Khriachtchev, S. Novikov, O. Kilpelä: J. Appl. Phys. 87, 7805 (2000) 410
[41] T. López-Rı́oz: Diamond Relat. Mater. 5, 608 (1996) 411
[42] D. G. McCulloch, S. Prawer, A. Hoffman: Phys. Rev. B 50, 5905 (1994) 416
[43] D. G. McCulloch, E. G. Gerstner, D. R. McKenzie, et al.: Phys. Rev. B 52,
850 (1995) 416
[44] V. D. Vankar, N. Dilawar: Vacuum 47, 1275 (1996) 416
[45] K. C. Walter, H. H. Kung, C. J. Maggiore: Appl. Phys. Lett. 71, 1320 (1997)
416
[46] J. R. Shi, Z. Sun, X. Shi, S. P. Lau, B. K. Tay, J. Pelzl: Thin Solid Films 377,
269 (2000) 416
[47] W. Y. Luk, O. Kutsay, I. Bello, et al.: Diamond Relat. Mater. 13, 1427 (2004)
416
[48] C. W. Ong, X. A. Zhao, J. T. Cheung, et al.: Thin Solid Films 258, 34 (1995)
417
[49] R. Kalish, Y. Lifshitz, K. Nugent, et al.: Appl. Phys. Lett. 74, 2936 (1999)
417
Index
ta-C, 403 optical characterization, 403, 413
optical properties, 404, 410, 411, 413,
absorption coefficient, 404, 409, 411, 414, 418
413–418
annealing, 403, 417, 418 pulsed cathodic arc discharge (PCAD),
404, 406, 408–412, 416, 418
D band, 406
D/G band intensity ratio, 406, 409, Raman parameters, 406–409, 414
416, 417 Raman scattering, 409, 418
diamond-like carbon (DLC), 403, 404, Raman intensity, 410, 414, 415
406, 412, 413, 415, 416, 418 Raman spectroscopy, 403–406, 408,
409, 418
G band, 406, 416, 417 refractive index, 403, 404, 409–411, 413,
414, 416, 418
interference of light, 409
ion irradiation, 416 surface enhanced Raman spectroscopy
(SERS), 404, 411
mass-separated ion beams (MSIB),
403–405, 407–414, 416–418 thermal stability, 417
Raman Spectroscopy of CVD Carbon Thin
Films Excited by Near-Infrared Light
Margit Koós, Miklós Veres, Sára Tóth, and Miklós Füle
Research Institute for Solid State Physics and Optics of the Hungarian Academy
of Sciences, Konkoly Thege M.u. 29-33, H-1121 Budapest, Hungary
koos@szfki.hu
Abstract. The interpretation of the Raman spectra of amorphous carbon thin

films is still controversial. The concept of their decomposition into D and G peaks
does not work in some cases, when presence of additional bands can be deduced
from the shape of the spectra. One to investigate these extra component bands is to
change the excitation wavelength. This results in the enhancement of the Raman
scattering cross-section of the different structural units. Our investigations were
aimed to detect the component bands become intense by using infrared excitation
(785 nm) in the Raman spectra of amorphous carbon thin films prepared from
benzene and methane in a wide range of deposition parameters. By comparing the
visible and infrared excited Raman spectra it will be proven that the well-known
G band of the a-C:H layers consists of two components, one of which exhibiting
no dispersion. It will be also shown that he infrared excitation makes distinguish-
able Raman bands assigned to delocalized (π) electronic structure due to resonant
enhancement of the scattering.
1 Introduction
1.1 The Raman Effect
When monochromatic light of frequency ω L scatters in a medium, scattered

light intensities shifted from ω L by certain Ω i values can be observed in the
spectrum of the scattered light. The value of these shifts depends on the
properties of the scattering media and it does not vary when changing ω L .
This phenomenon is the “Raman scattering”. During a Raman experiment
the scattered light is measured, and it is represented in relative wavenumbers,
taking the laser wavenumber being equal to zero.
Raman scattering is an inelastic scattering of light on elementary exci-
tations of the medium. The elementary excitations are usually rotational or
vibrational transitions of a molecule or lattice vibrations (phonons) in a solid.
The inelastic scattering is a two-phonon process, where the absorption of an
incoming photon of energy EL = ω L with energy transfer and the creation
of an ωS scattered photon take place simultaneously in the medium. The
energy difference is equal to the Ω energy of the elementary excitation.
If the energy transfer is positive (ω L ≥ ωS ) the process is called Stokes
scattering, the opposite case (ω L ≤ ωS ) is known as anti-Stokes scattering.
424 Margit Koós, Miklós Veres, Sára Tóth, and Miklós Füle
The E = E0 cos (ω L t ) electromagnetic field of frequency ω L acting on a

medium induces a dipole moment µin determined by the polarizability of the
medium:
µin = αE = αE 0 cos (ω L t ). (1)
A molecule or an elementary cell consisting of N atoms has 3N degrees of

freedom, 3 of which are transitional, 3 (2 for a linear system) are rotational
and 3N − 6 (3N − 5 for a linear system) are vibrational ones. 3N Cartesian
coordinates are needed to describe the motion of such a system. It is suitable
to use “internal coordinates”, describing the variations of bond lengths and
bond angles during vibrations. Combining these coordinates, the orthogonal
vibrational “normal modes” of the molecule (cell) can be described. The αij
components of the polarizability tensor can be expressed through the normal
coordinates of the system:
∂αij 1
2
∂ αij

αij = (αij )0 + Qk + Qk Qk + · · · . (2)
∂Qk 0 2 ∂Qk ∂Qk
k k,k
Restricting (2) to first-order terms, combining (1) with (2) and applying
trigonometric identities, we get:
µin,ij = (αij )0 E 0 cos (ω L t)

1 ∂αij
+ Qk E 0
2 ∂Qk 0
k
· {cos [(ω L + Ω)t + φk ] + cos [(ω L − Ω)t − φk ]} . (3)
From (3) it can be seen that during a scattering process a medium will
emit photons at frequencies ω L , ω L +Ω and ω L −Ω corresponding to Rayleigh,
anti-Stokes and Stokes scattering, respectively. While Raleigh scattering does
always exist, Raman scattering occurs only if the polarizability of the medium
changes during the scattering process. The first-order terms of the polariz-
ability tensor represent the components of the “Raman tensor” R.
Energy and momentum are conserved in the Raman process:
ωS = ω L ± Ω
kS = kL + kph , (4)
where kL , kS and kph are the momenta of the incoming and scattered photons
and of the phonon, respectively. According to (4), the energy loss of the
photon equals the energy of the lattice vibration. Considering that, for the
visible light the 104 cm−1 magnitude of kL and kS is much less than the
size of the first Brillouin zone of the crystals (about 1010 cm−1 ), it can be
concluded that only zone centre phonons participate in the scattering.
Raman Spectroscopy of CVD Carbon Thin Films 425
In the classical picture described above, the mechanism of the scattering

does not matter. The quantum mechanical treatment of the process considers
the electron–photon and electron–phonon interactions through their Hamilto-
nians He−R,σ and He−ph,ρ (σ and ρ being the polarizations of the incoming
and scattered photons, respectively). The core of the quantum mechanical
treatment of the Raman scattering is the transition matrix element K2f,10 :
ω ,f ,i|H
S e−R,ρ |0 ,f ,b0,f,b|He−ph |0,0,a0,0,a|He−R,σ |ω L ,0,i
K2f,10 = , (5)
(E L −Eaie −iγ )(E S −Ebie −iγ )
a,b
where |ω L , 0, i is the initial state characterized by the incoming photon of

E L = ω L energy, by the phonon in 0 state (no excited phonon) and by the
electron in ground state i. The final state ω S , f , i| is characterized by the
photon of E S =ω S energy, by the phonon in the f state and by the electron
in the ground state; a and b are the intermediate states participating in the
e e
scattering process. Eai and Ebi are the energy differences of states a and i as
well as b and i, respectively; γ is the lifetime of the excited states. According
to (5) the Raman scattering is a three-step process consisting of:
(a) absorption of the incoming photon with transition of the electron into an
excited state (creation of an electron–hole pair)
(b) inelastic scattering of the electron on a phonon
(c) recombination of the electron and the hole with emission of the scattered
photon
The three steps take place simultaneously. The electron can be excited from
the ground state either into a virtual state or into an existing state. In the
latter case, the scattering is “resonant”.
The K2f,10 matrix element is related to the Raman tensor (R) through
the polarization vectors of the incoming (eL ) and scattered (eS ) light:
K2f,10 = eL · R · eS .
The scattered intensity is proportional to the square of the transition matrix

element:
2
I ∼ |K2f,10 | .
It is difficult to calculate the intensity of the process since the transition

matrix elements are difficult to determined.
The number of lines observed in the Raman spectra is less than the num-
ber of phonons existing in the crystal. This is not only due to the degeneracy
of the phonons, but it is because of the number of phonons which can partic-
ipate in the scattering is limited by the “selection rules”. The first restriction
is that the scattering takes place only on zone-centre phonons. Additionally,
during the scattering the polarizability of the medium should change, which
requires the phonons to have certain symmetries. The simplest way to deter-
mine whether a given phonon will cause a nonzero polarizability change is to
take into account group theory considerations.
When the energy of the incoming or outgoing photon is close to an existing
energy transition of the medium, a significant increase in the Raman intensity
can be observed due to resonant enhancement of the scattering. Depending
on whether the resonance is for the incoming or scattered light, one can
distinguish between incoming and outgoing resonances.
In noncrystalline materials there is no elementary cell and the periodicity
is lost. This leads to breakdown of the selection rules. As a result, in these ma-
terials any phonon can participate in the scattering. However, the vibrations
will not be extended in the whole crystal, instead they will be localized. The
Raman spectrum is obtained by summing all of these localized vibrations [1]:

4 1
I (ω) = (ωL − ω) Cb [1 + n (ω, T )] gb (ω) , (6)
ω
b
where Cb is the coupling coefficient of vibrational transitions, gb is the den-

sity of vibrational states, n (ω, T ) is the Bose–Einstein distribution function.
From (6) it can be seen that the Raman intensity is determined by the density
of vibrational states.
1.2 Raman Spectra of Carbon Materials
The basic electronic configuration of a stand-alone carbon atom differs from

that during bonding. In the latter case, the electronic orbitals of carbon
combine, forming hybridized orbitals: sp3 , sp2 and sp hybridization states are
known for carbon atoms, which form three-dimensional, layered and chain-
like structures like diamond, graphite and carbyne. In disordered carbons,
atoms of different hybridization are mixed. Besides, hydrogen can also be
present in the structure. Because of the large variety of atomic arrangements
in carbon materials, their properties, both crystalline and amorphous, vary
in a wide range.
Among the several different models proposed for the a-C:H [2, 3], nowa-
days the cluster model [4, 5] has found wide acceptance since that best de-
scribes the electronic structure of amorphous carbons. Most of the experi-
mental data obtained for a-C:H films were successfully explained with this
model. The cluster model is based on considerations of the Hückel approx-
imation, treating the σ and π states separately [4]. It was found that the
π bonding of carbon atoms favours a clustering of sp2 sites predominanatly
into planar structures formed by sixfold rings. As a result, in the amorphous
carbon matrix there are sp2 clusters of different sizes surrounded by a matrix
of sp3 carbon atoms. It should be noted that the cluster model was reformu-
lated [6]. According to it, in clusters, the sp2 hybridized carbon atoms can be
arranged, besides into rings, into chains too [7, 8]. The energy of the π states
of the cluster strongly depends on the size and on the level of conjugation of
the cluster. The higher the cluster size, the lower the gap between the valence
π and conduction π ∗ states. The energy of the σ states is much higher than
that of π ones.
Raman spectroscopy has been proved to be a very suitable method for
structural characterization of carbon materials [9, 10, 11, 12, 13, 14, 15]. The
method is highly sensitive to changes in the bonding configuration of carbon
atoms, especially to that of sp2 hybridized ones, since the π states of the
sp2 clusters result in narrow band gaps, which cause the Raman scattering
process to be resonantly enhanced.
The Raman spectrum of diamond consists of one zone-centre mode at
1332 cm−1 having T2g symmetry. In the spectrum of single crystal graphite,
a peak can be observed at 1580 cm−1 corresponding to zone-centred E2g
mode, usually labelled as “G” (graphitic) band, assigned to C–C stretching
vibrations of the atoms in hexagonal rings of the graphene sheet. Besides,
an interplanar E2g stretching vibration mode is also present at 42 cm−1 [16].
In “nonperfect” or microcrystalline graphite, another band of A1g symme-
try, called the “D” (disordered) peak [17], corresponding to breathing vibra-
tions of the hexagonal rings at the grain boundaries, appears at 1350 cm−1 .
The D band is a result of a double resonant scattering process activated by
the defects of the crystal [18]. There are other bands of small intensity ob-
served in the Raman spectra of graphite [19], as well as intensive second- and
higher-order peaks above 2000 cm−1 [20], but these are not relevant for the
interpretation of the spectra of amorphous carbons. Due to the specific band
structure of graphite, where the conduction and valence bands cross at the
K point but a forbidden gap exists in other points of the Brillouin zone, the
Raman scattering in graphite is always resonant. On the contrary, the band
gap of diamond is around 5 eV, hence for visible excitation the intensity of
its 1332 cm−1 mode is about one fiftieth that of the G peak of graphite [21].
The Raman spectra of amorphous carbons consist of a broad band in the
1000–1700 cm−1 range. This composite band is usually decomposed into two
peaks centred around 1350 and 1580 cm−1 , which are near the position of
D and G bands of microcrystalline or nonperfect graphite. This similarity
was the origin of the labelling and first explanation of the Raman spectra
of amorphous carbons [9]. The D band of amorphous carbon materials is
assigned to breathing vibrations of rings or ring-like structures; the stretching
modes of these species give rise to the scattering in the G band region [14].
However, besides the stretching vibrations of carbon atoms arranged in rings,
those forming chains also contribute to the G band [13]. The parameters of
the two bands (position, width and their intensity ratio) are used for the
characterization of carbon-containing materials [10, 11].
The shape of the Raman spectrum is considered to depend on the sp2 /sp3
ratio, clustering of the sp2 hybridized carbon atoms and arrangement of these
atoms in the clusters [15]. The position of the G band indicates the type of
arrangement of carbon atoms in the clusters: if its value is around 1580 cm−1 ,
the clusters have graphitic character, while lower positions indicate a differ-
ent arrangement. The peak width is the measure of structural ordering: the
narrower the peak, the more ordered the structure. Another parameter used
for a-C:H thin layers characterization is the intensity ratio of D and G bands.
The I D /I G ratio was found to be related to the size of the sp2 clusters in the
amorphous carbon structure. It is assumed that for microcrystalline graphite
the I D /I G ratio is inversely proportional to the crystallite size LA [22]:
ID c
= . (7)
IG LA
Experimental measurements performed on ta-C thin films showed that for
amorphous carbons having sp2 clusters below ≈ 2 nm size the ratio is pro-
portional to the square of the cluster size LC [14, 23]:
ID
= c L C 2 . (8)
IG
The constants c and c in (7) and (8) were found to be dependent on the
excitation energy. The I D /I G ratio is the measure of ordering of the a-C:H
structure. The higher the intensity ratio, the higher the cluster size, thus the
higher the ordering of the structure.
Raman spectroscopic measurements with different excitation wavelengths
can furnish additional information on the a-C:H structure [10, 24, 25, 26, 27].
The bonding sites having sp2 carbon atoms are generally arranged into clus-
ters of different sizes, which exhibit band gap depending on cluster size. For
different laser energies, E L , the conditions of resonant Raman scattering will
be fulfilled for different clusters, whose band gaps are equal to that. How-
ever, the vibrational frequency depends on the cluster size: the higher the
size, the lower the frequency of the vibrations [28]. Additionally, when using
lasers in the UV region, the sp3 sites can be resonantly excited, thus their
bonding configuration can also be examined. A characteristic feature of the
D and G bands in the Raman spectra of amorphous carbon is their “disper-
sion”, the shift of the peak position when changing the excitation energy. The
dispersion was also observed for the D band of graphite [29]. It was found
that for different amorphous carbons (hydrogenated and nonhydrogenated)
the rate of the shift varies in a wide range [15]. For a-C:H thin films (poly-
meric and diamond-like) the position of both D and G bands shifts to higher
wavenumbers with the increase of E L [15, 24, 30].
It was found that in graphite the Raman scattering cross-section is
strongly enhanced for the phonons having wavevector kph equal to the
wavevector k of the electronic transition excited by the incident photon [29].
For diverse E L , this condition is fulfilled for phonons at different distances
from the K point of the zone boundary, hence the energy of the phonon par-
ticipating in the scattering also varies with E L [18]. In a-C:H the dispersion
of the bands is differently explained. Since there is no Brillouin zone, the se-
lection rules have no meaning. As noted earlier, when changing the excitation
energy, different clusters will participate in resonant scattering. The band gap
of the cluster is inversely proportional to its size. Thus, the higher the E L ,
the lower the mean size of the clusters involved in the resonant scattering.
The lower the cluster size, the higher the frequency of its vibrations.
Up to now, systematic Raman investigations on amorphous carbon thin
films of different types were performed mainly with excitations in the visible
and UV region [11,12,14,15]. The aim of the increase of the probe energy was
that the sp3 hybridized carbon sites, having higher band gap, could be excited
too. However, the resonant enhancement of Raman scattering from graphitic
structure or/and materials containing several sp2 bonded sites arranged in
nanoclusters at low-energy excitation can provide additional information on
the structure of the a-C:H films. In the following, results of Raman spectro-
scopic studies, performed with visible and infrared excitations, on a series
of a-C:H samples prepared by radio-frequency chemical vapour deposition at
different self-bias voltages, from benzene and methane, will be presented in
detail.
2 Infrared Excited Raman Spectroscopy

of Amorphous Carbon Thin Films
2.1 a-C:H Thin Films Prepared from Benzene
The a-C:H thin film samples were deposited onto Si substrates by radio-
frequency (2.54 MHz) chemical vapour deposition (CVD) method [31] from
benzene at different chamber pressures (8–20 Pa) and self-bias voltages (10–
700 V). The electrode with the substrate had negative self-bias potential,
however, for simplicity, in the following only the magnitude of the self-bias
voltage is given without the “−” sign. Raman spectroscopic measurements
were carried out on the samples using a Renishaw 1000 Raman spectrometer
attached to a microscope. A 488 nm (2.54 eV) line of an Ar ion laser and a
785 nm (1.58 eV) diode laser served as excitation sources. The 100× objective
focused the excitation beam to a spot having diameter of ≈ 1 µm. Baseline
correction on the measured spectra was performed by fitting their baseline to
a Gaussian function and subtracting the fitted data from the experimentally
measured ones. Raman spectra of a-C:H thin layers measured at 488 nm
of probe wavelength are shown in Fig. 1. The data are normalized on their
maxima. The change of the shape of the spectra with self-bias well reflects the
evolution of the amorphous carbon structure. In the spectrum of the sample
prepared at 10 V the G peak is located around 1600 cm−1 . With the increase
of self-bias up to 300 V, it shifts to lower wavenumbers and, above this voltage,
the peak positions move into opposite direction. The intensity of the D band
also increases with self-bias. The spectra of the samples prepared at low
Fig. 1. Raman spectra excited by 488 nm of a-C:H thin films prepared from benzene
at pressure of 18 Pa and different self-bias voltages
deposition voltages have bad signal/noise ratios due to the high intensity of
the luminescence background. The weak peak around 1000 cm−1 , observable
in some spectra, arises from the Si substrate.
In the 1000–1700 cm−1 wavenumber region the spectra were fitted to two
Gaussians. The dependence of the position and peak width of the G and D
bands on the self-bias voltage is shown in Fig. 2, where data obtained for a
series of samples prepared at 8 Pa are also provided. It can be seen that there
is a minimum in the position of the G peak around 300 V, where diamond-like
layers are formed, while the D band has the lowest location around 100–200 V.
Above 500 V, the G peak position increases and approaches the 1580 cm−1
peak of graphite. The peak width decreases, implying that some kind of
ordering takes place in the a-C:H structure. In this region, the graphitization
of the structure begins, and above 600 V of self-bias, the structure can be
considered as graphitic a-C:H.
Similar behaviour is observable for the G peak parameters at low voltages,
where polymeric a-C:H layers are formed. For the samples prepared below
30 V, it reaches the position of 1600 cm−1 , well above the G peak position
Fig. 2. Dependence of (a) the peak position and (b) the peak width on self-bias
voltage of the D and G bands of a-C:H thin films prepared from benzene at pressures
of 8 and 18 Pa
of graphite, suggesting that the G peak in the spectra of these samples does
not arise from graphitic domains. The band narrowing suggests the ordering
of the structure.
In Fig. 3, it can be seen that the I D /I G ratio has a minimal value in the
diamond-like a-C:H films, around 300 V self-bias. As we discussed before, the
D band can be attributed to scattering of breathing modes rings or ring-like
structure formed by sp2 -hybridized carbon atoms, while stretching modes of
both rings and chains formed by sp2 sites contribute to scattering in the G
Fig. 3. Dependence of intensity ratio of the D and G bands of the Raman spectra
of a-C:H film prepared from benzene at pressures of 8 and 18 Pa on self-bias voltage
band region. Hence, the I D /I G ratio indicates the ratio of the amount of (sp2 )
ring-like structural units to the total amount of (sp2 ) rings and chains. The
increase of the I D /I G ratio is due to the presence of benzene rings or of their
substituted forms in the structure, below 300 V self-bias. At the same time,
the formation of graphitic structure, at large deposition voltages, explains
the increase of the I D /I G ratio in that region. This explanation is in good
accordance with the change of the G peak position with self-bias. Above and
below that region, the a-C:H structure becomes more ordered. On strength
of the analysis of the G band behaviour, one can conclude that this ordering
in polymeric and graphitic layers has different origin. This is supported by
Raman spectroscopic investigations performed with infrared excitation.
In several works [13, 32, 33, 34], it is noted that the use of two peaks for
the decomposition of the Raman spectra of a-C:H thin films is inaccurate,
so the fitted curve differs from the experimental one. This suggests that not
only structural units, vibrating at frequencies of G and D peaks, contribute
to the spectra. However, since it is difficult to determine the parameters of
the extra peaks, the attempts made for their assignment have controversial
results [13, 15, 33]. One way to obtain additional information on these bands
is to change the excitation wavelength, since the dispersion of the peaks can
be different.
Earlier investigations showed that the Raman scattering spectra of the
amorphous carbon materials excited in the infrared light region can give
additional information on the bonding structure [35]. Raman spectra of the
samples measured by 785 nm excitation are shown in Fig. 4. The region
Fig. 4. Raman spectra excited by 785 nm of a-C:H thin films prepared from benzene
at pressure of 18 Pa and different self-bias voltage U SB
around 1000 cm−1 was cut out since, due to the resonance enhancement, the
Raman scattering intensity arising from the Si substrate is several orders of
magnitude higher than that of the a-C:H films. By comparing Figs. 1 and 4,
it can be concluded that there are significant differences in the spectra of
the same samples, especially at low self-bias voltages. While broad bands are
characteristic of the 488 nm excited spectra, sharp peaks, at 1200, 1300, 1450
and 1600 cm−1 wavenumbers, can be found in the infrared excited ones of
these films. Besides the 1000–1600 cm−1 region, a broad band can also be
seen around 1800 cm−1 . The increase of the self-bias results in broadening
of the bands and shifting of their positions. At 120 V self-bias new bands
appear in the spectra around 1200 cm−1 .
The spectra were decomposed by sets of Gaussian and Lorentzian curves.
It is convenient to analyse the narrow bands in the spectra of polymeric films
separately from broad bands.
Figure 5 compares the Raman spectrum of the a-C:H film prepared at
10 V self-bias with that of the benzene. Additionally, the spectrum of the
part of the film detached from the substrate is also provided. It can be seen
Fig. 5. Comparison of the Raman spectra excited by 785 nm of (a) benzene,

(b) detached from the substrate and (c) as-prepared a-C:H thin film deposited
from benzene at pressure of 18 Pa and 10 V of self-bias
that bands at 1190 and 1607 cm−1 in the spectra of the a-C:H film arise
from the presence of benzene rings in the structure. Another benzene peak
at 1007 cm−1 , masked by the band of Si substrate, is observable only in the
spectra of the detached film. The differences in the band positions can be
explained considering that there are not separate benzene rings, but substi-
tuted ones present in the structure. The substitution changes the symmetry
of the molecule, thus forbidden Raman bands appear in the spectra too.
Peaks at 1037, 1160, 1288 and 1448 cm−1 are related to substituted benzene,
whose presence was supported by infrared transmission measurements [36]
(not presented here).
It is accepted that the sp3 carbon atoms contribute to the Raman spectra
when it is excited in the UV region. Contrarily to this, information on the
bonding configuration of sp3 hybridized carbon atoms can also be deduced
from the infrared excited spectra of the a-C:H films prepared at low self-
bias voltages. The small peak around 1382 cm−1 is presumably related to
symmetric C–H deformation vibrations of sp3 CH3 groups. This mode usually
has low Raman intensity, but not in the case of sp3 CH3 group attached
to carbon atoms having double or triple bonds or to a benzene ring. The
asymmetric pair of the vibration is around 1450 cm−1 , overlapped by the
1448 cm−1 mode of benzene. The characteristic C–H mode of sp3 CH2 group
contributes to the spectra at 1300 cm−1 . The evolution of these bands with
self-bias voltage (Fig. 4) evidences the intact benzene rings are present in the
spectra of the a-C:H films up to 80 V.
The decomposition shows that, as a background of these narrow peaks,
there are two broad bands, observable in the spectra of the 10 V sample,
centred at 1378 and 1599 cm−1 . Their position is similar to that of the D and
G bands in the spectrum excited by 488 nm (Fig. 2). The G band position
(1599 cm−1 ) is higher than that in graphite (1580 cm−1 ). It is close to the
1607 cm−1 peak of the substituted benzene. This suggests that the sp2 carbon
atoms in the amorphous carbon matrix have an arrangement close to that
they have in a benzene molecule. Presumably, the amorphous phase of the
sample contains a high amount of distorted benzene rings.
As the self-bias increases, the shape of the spectra changes significantly. In
the spectra of the 80 and 120 V samples, in the 1000–1700 cm−1 wavenumber
range, four broad bands are observable, centred around 1140, 1260, 1430 and
1566 cm−1 . The close position of the 1566 cm−1 peak to the G peak in the
488 nm excited Raman spectrum of the sample (at 1560 cm−1 ) suggests that
this band is the G one, too. However, we must take into account that the
D and G bands of a-C:H have dispersion. By estimating from the dispersion
rates up to 785 nm of excitation, the D band of a-C:H films has to be located
around 1250 cm−1 and the G one is around 1430 cm−1 . By comparing these
values with the data obtained for the 120 V sample, it can be assumed that
the 1260 and 1430 cm−1 peaks of that correspond to the D and G peaks.
So one gets the conclusion that there are two bands in the infrared excited
spectrum that correspond to the G band: one, at lower wavenumbers, around
1430 cm−1 , showing dispersion, and the second, around 1566 cm−1 , having
no dispersion.
In Fig. 6 the dependence of the positions and widths of the broad bands,
found in the 785 nm excited Raman spectra, on the self-bias voltage are
shown. In the spectrum of the 10 V sample, the band showing no dispersion
(“G(n−d) ” in the following) is located at 1599 cm−1 . The increasing self-
bias causes the shift of the band position to lower wavenumbers. It reaches
1570 cm−1 at 120 V and does not vary significantly above that voltage value.
The band width is almost the same in the whole self-bias range. The position
of the G peak that shows dispersion (“G(d) ” in the following) fluctuates
around 1430 cm−1 and broadens with the increasing self-bias.
The position of the D band located at 1378 cm−1 in the spectrum of
the 10 V film decreases rapidly with the increase of the self-bias, down to
1230 cm−1 at 200 V, to the position expected with considering the disper-
sion. Above that voltage the band position does not change remarkably up
to 500 V and increases above that value. As it was noted earlier, the disper-
sion of the bands in amorphous carbons is due to the size distribution of the
clusters having different band gaps. If one goes into the details of the phenom-
enon, which was observed also in hydrocarbon polymers built of “conjugated
chains” (chains consisting of sp2 carbon atoms connected with alternating
single and double bonds), it turns out that it is not the size that determines
the band gap of a cluster, but the delocalization of the π electrons in it. The
cluster size is only the upper limit for the delocalization length. The level
of delocalization is determined by the lengths of conjugated regions, which
depend on the bonding configuration of the sp2 carbon atoms. Several fac-
Fig. 6. Dependence of (a) the position and (b) width on self-bias voltage of the
broad bands found in the Raman spectra excited by 785 nm of a-C:H thin films
prepared from benzene
tors can affect on the delocalization length of a conjugated chain, including

kinks, breaks in the periodicity of the alternating bonds and attachment of
side-chains. It is known that the benzene molecule has an electronic configu-
ration such that its π electrons are delocalized only within the ring, but not
out of it.
From the absence of the G(d) band and the downshift of position of the D
peak up to 80 V self-bias, it can be concluded that the increasing deposition
voltage causes the increase of the extent of π electron delocalization in sp2

clusters. The behaviour of the G(n−d) band indicates that the atomic arrange-
ment in the clusters of a-C:H matrix transforms. In the layers prepared at
low self-biases, mainly distorted benzene rings form the cluster, while, with
the increase of the voltage, the ratio of carbon atoms arranged into chains
rises, as well as the delocalization length characteristic for the clusters. The
amount of the intact benzene rings also decreases with the increasing self-bias
from 10 V, and above 120 V they are completely destroyed.
Above 120 V self-bias, the G(d) peak broadens and its intensity increases
relative to the G(n−d) band. In the spectra of the 300 V sample, the two
bands overlap, so as they can easily be treated as one. When the structure
becomes graphitic (700 V), the G(n−d) band rises since there are graphitic
rings present in the structure.
The decomposition showed the presence of another broad peak located
at 1130 cm−1 in the spectra of samples deposited at self-biases in the 80–
200 V range. The band shifts to higher wavenumbers with increasing self-bias.
A similar band in the spectra of a-C:H was reported earlier and was related
to vibrations of sp2 chains (trans-polyacethylene-like structural units) [37],
sp3 carbon phase [38] and nano-crystalline diamond [13, 33]. The peak ap-
pears together with the 1430 cm−1 one, which was assigned to C–C stretch-
ing vibrations of structural units having large delocalization lengths of their
π electrons. A typical representative of hydrocarbon chains, built of sp2
carbon atoms, is trans-polyacethylene. It has characteristic Raman peaks
around 1450 cm−1 and 1100 cm−1 , both showing dispersion. Presumably,
the 1130 cm−1 band can be assigned to vibrations of sp2 chains. Model cal-
culations showed that, for short polyacethylene chains, the position of the
latter band strongly depends on the chain length too, and shifts to lower
wavenumbers with increasing chain length, while there is no such effect ob-
servable for the other band [39]. The increase of the peak position with self-
bias shows that, as the films become more diamond-like, the length of the
conjugated chains decreases in the clusters. Above 200 V, the band cannot
be decomposed due to its overlapping with the D band.
In conclusion, it was shown that infrared-excited Raman spectroscopy is
an excellent tool for characterization of different types of a-C:H thin films. It
can provide additional information on the bonding configuration of carbon
atoms in the structure. The spectra recorded by near-infrared excitation were
decomposed into four peaks. It was found that the G peak has two compo-
nents, one of which shows dispersion, while the position of the other does not
change with excitation wavelength. In the lower wavenumber region, besides
the D band around 1130 cm−1 , another peak appears in the spectra, related
to vibrations of carbon atoms arranged into chains.
By using the infrared-excited Raman spectroscopy, the evolution of the
structure of a-C:H layers prepared from benzene was investigated. It was
shown that, at low self-bias voltages (up to 120 V), intact (substituted) ben-
zene rings embedded into the amorphous structure. In addition, the amor-
phous matrix of the film prepared at 10 V contains structural units not dif-
fering remarkably from benzene. Their content, as well as that of substituted
benzene decreases with increasing self-bias. The sp2 carbon atoms arrange
into chains. The small delocalization length also increases with deposition
voltage. From 120 V to 400 V, the chains dominate the sp2 clusters. The
increasing with self-bias graphitic character will dominate the spectra of the
samples above 400 V.
Visible and infrared excited Raman spectroscopic measurements were also
performed on a-C:H samples prepared from methane in a wide self-bias range.
In the following, the results of these investigations will be presented.
2.2 a-C:H Thin Layers Prepared From Methane
The methane films were prepared similarly to benzene ones at chamber pres-
sures of 13 Pa. The Raman spectra of the a-C:H films excited at 488 nm
are shown in Fig. 7. The shape of the spectra is similar to those presented
in Fig. 1. The only difference is that there was not so high a luminescence
background observable in the samples prepared at low voltages.
The spectra were analysed similarly to the benzene ones. Figure 8 shows
the variation of positions and widths of D and G bands with the self-bias
voltage. Contrarily to the layers prepared from benzene (Fig. 2), the positions
of both bands increase monotonically with the self-bias. The peak width of
the D band increases also monotonically, while that of the G peak, after
a small increase, decreases above 60 V. The comparison with the results
obtained for the a-C:H films deposited from different source gases shows
that in the low self-bias range the sample series prepared from benzene and
methane behave differently. In case of the benzene layers, the presence of
ring-like units (distorted benzene rings) in the amorphous structure causes
the G peak to appear around 1600 cm−1 . On the contrary, in the 30 V sample
prepared from methane, the G peak is located at 1537 cm−1 , suggesting that
chain-like structural units are mainly present in the sp2 clusters. The shift
of the G band to higher wavenumbers with increasing self-bias indicates the
increase of the amount of ring-like structures in the sp2 clusters of the film.
This is supported by the change of the I D /I G ratio of the films (Fig. 9),
which also differs from that of the benzene ones. It rises monotonically with
the increasing self-bias, showing the increase of the ratio of the rings in the
structure.
The huge difference in the bonding structure of a-C:H films prepared
from methane and benzene at low self-biases are also evidenced from their
Raman spectra excited at 785 nm. While the spectra of the benzene films are
characterized by narrow peaks (Fig. 4), those of methane layers are composed
of broad bands (Fig. 10). The spectrum of the layer prepared from methane
at 30 V self-bias is similar to the spectra of the films prepared from benzene
at 300–500 V potential and evolves with the increase of the self-bias similarly
Fig. 7. Raman spectra excited by 488 nm of a-C:H thin films prepared from
methane at different self-bias voltage U SB
to the benzene ones above 300 V, where the π sites have large delocalization
length.
The spectra were decomposed by a set of four Gaussians. The variation of
the positions and peak widths of the component bands with self-bias voltage
are shown in Fig. 11. The positions of component bands are similar to those
found in the benzene samples above 200 V self-bias. This implies that these
peaks have the same origin. Hence, the infrared-excited Raman spectra of
methane layers also have the D, G(d) and G(n−d) bands (around 1230, 1430
and 1570 cm−1 , respectively), as well as the peak assigned to vibrations of sp2
carbon atoms arranged into chains (around 1150 cm−1 ). The position of the
latter band, located in the spectra of the 30 V sample at 1105 cm−1 , shifts to
lower wavenumbers with the increase of the self-bias, showing the increase of
the sp2 chain lengths of in the clusters [39], and becomes undetectable in the
spectrum of the 600 V sample. In the films prepared from benzene, the band
is positioned at higher wavenumbers, indicating the smaller chain lengths in
the clusters of those layers.
Fig. 8. Dependence of (a) the peak position and (b) the peak width on self-bias
voltage of the D and G bands of a-C:H thin films prepared from methane
The position and width of the G(d) band do not change significantly with
the increase of the self-bias, but its intensity decreases as the graphitic char-
acter of the structure strengthens. This is accompanied by the shift of the
G(n−d) peak from 1560 cm−1 to higher wavenumbers.
The comparison of the bonding configuration of the a-C:H thin layers
prepared from methane and benzene shows that the structure of the films
prepared under similar conditions differs significantly at low self-biases. In
the benzene molecule, the carbon atoms have preferential arrangement. At
Fig. 9. Dependence of intensity ratio of the D and G bands of the Raman spectra
of a-C:H film prepared from methane on self-bias voltage
Fig. 10. 785 nm excited Raman spectra of a-C:H thin films prepared from methane
at different self-bias voltage U SB
Fig. 11. Dependence of (a) the position and (b) width on self-bias voltage of the
broad bands found in the Raman spectra excited by 785 nm of a-C:H thin films
prepared from methane
low ion energies, the amorphous carbon matrix mainly consists of distorted
benzene rings, so that besides intact rings, distorted benzene rings are also
present in the a-C:H matrix. This arrangement causes the π electrons to be
highly localized in the rings of clusters. As the ion energy increases, these
rings are destroyed in the plasma, as well as during the bombardment of
the surface of the growing film, and the atoms rearrange, forming sp2 clus-
ters characterized by large delocalization lengths. Clusters containing large
amount of sp2 chains develop in the self-bias region, where diamond-like a-

C:H forms. With further increase of the deposition voltage, the graphitic
clusters will dominate in the layer.
On the contrary, for the methane molecule consisting of only one carbon
atom, there is no preferential arrangement. Therefore, during film formation,
the topology of the carbon atoms develops without initial constraints, except-
ing for the bonding angles, set by the hybridization state of the carbon atom.
The sp2 clusters, formed at low self-biases, are built of carbon atoms arranged
in both rings and chains and have large delocalization lengths, compared to
the benzene ones prepared under similar conditions. The increase of the de-
position voltage up to the region, where diamond-like carbon deposits, causes
the lengthening of the sp2 chains. Above this self-bias, the graphitization of
the structure starts, similarly to the benzene layers.
In conclusion, it was shown that Raman scattering excited in the near-
infrared region can provide additional information about the bonding config-
uration of of a-C:H thin films in a wide range of deposition conditions. Hence,
it can be dependably used for quality control of these materials. The method
was proved to be highly sensitive to the arrangement of the carbon atoms in
the sp2 clusters. The discrepant dispersion of the component peaks made the
composite spectra to be decomposed more obvious, and it could give an ex-
perimental basis for the assignment of the bands found in the Raman spectra
of a-C:H.
References
[1] R. Shuker, R. Gamon: Phys. Rev. Letters 25, 222 (1970) 426
[2] S. Craig, G. L. Harding: Thin Solid Films 96, 345 (1982) 426
[3] D. R. McKenzie, R. C. McPhedran, N. Savvides, L. C. Botten: Philos. Mag.
B 48, 341 (1983) 426
[4] J. Robertson: Adv. Phys. 35, 317 (1986) 426
[5] J. Robertson, E. P. O’Reilly: Phys. Rev. B 35, 2946 (1987) 426
[6] J. Robertson: Mat. Sci. Eng. R 37, 129 (2002) 426
[7] T. Frauenheim, P. Blaudeck, U. Stephan, G. Jungnickel: Phys. Rev. B 48, 4823
(1993) 427
[8] G. Jungnickel, T. Frauenheim, D. Proezag, P. Blaudeck, U. Stephan: Phys.
Rev. B 50, 6709 (1994) 427
[9] R. Nemanich, J. Glass, G. Lucovsky, , R. Shroder: J. Vac. Sci. Technol. A 6,
1783 (1988) 427
[10] M. A. Tamor, J. A. Haire, C. H. Wu, K. C. Hass: Appl. Phys. Lett. 54, 123
(1989) 427, 428
[11] S. Xu, M. Hundhausen, J. Ristein, B. Yan, L. Ley: J. Non-Cryst. Solids 1127,
164–166 (1993) 427, 429
[12] M. A. Tamor, W. C. Vassel: J. Appl. Phys. 76, 3823 (1994) 427, 429
[13] J. Schwan, S. Ulrich, V. Bathori, H. Erhardt, , S. R. P. Silva: J. Appl. Phys.
80, 440 (1996) 427, 432, 437
[15] A. C. Ferrari, J. Robertson: Phys. Rev. B 64, 75414 (2001) 427, 428, 429, 432
[16] M. J. Pelletier (Ed.): Analytical Applications of Raman Spectroscopy (Black-
well Science 1999) 427
[17] R. Nemanich, S. A. Solin: Phys. Rev. B 20, 392 (1979) 427
[18] J. Maultzsch, S. Reich, C. Thomsen: Phys. Rev. B 70, 155403 (2004) 427,
428
[19] Y. Kawashima, G. Katagiri: Phys. Rev. B 52, 10053 (1995) 427
[20] P. H. Tan, C. Y. Hu: Phys. Rev. B 64, 214301 (2001) 427
[21] J. Wagner, M. Ramsteiner, C. Wild, P. Koidl: Phys. Rev. B 40, 1817 (1989)
427
[22] F. Tuinstra, J. L. Koenig: J. Chem. Phys. 53, 1126 (1970) 428
[23] M. Chhowalla, A. C. Ferrari, J. Robertson, G. A. J. Amaratunga: Appl. Phys.
Lett. 76, 1419 (2000) 428
[24] M. Ramsteiner, J. Wagner: Appl. Phys. Lett. 51, 1355 (1987) 428
[25] M. Yoshikawa, G. Katagiri, H. Ishida, A. Ishitani, T. Akamatsu: Solid State
Commun. 66, 1177 (1988) 428
[26] M. Yoshikawa, N. Nagai, M. Matsuki, H. Fukuda, G. Katagiri, H. Ishida,
A. Ishitani, I. Nagai: Phys. Rev. B 46, 7169 (1992) 428
[27] I. Pócsik, M. Koós, M. Hundhausen, L. Ley: Amorphous Carbon: State of the
Art (World Scientific, Singapore 1998) p. 224 428
428
[29] I. Pócsik, M. Hundhausen, M. Koós, L. Ley: J. Non-Cryst. Solids 1083, 227–
230 (1998) 428
[30] M. Ramsteiner, J. Wagner: Appl. Phys. Lett. 51, 1355 (1987) 428
[31] M. Veres, M. Füle, S. Tóth, M. Koós, I. Pócsik, J. Kokavecz, Z. Tóth,
G. Radnóczi: J. Non-Cryst. Solids 351, 981 (2005) 429
[32] J. Rao, K. J. Lawson, J. R. Nicholls: Surface Coatings Technol. 197, 154 (2005)
432
[33] S. R. P. Silva, G. Amaratunga, E. Salje, K. Knowles: J. Mater. Res. 29, 4962
(1994) 432, 437
[34] A. I. Kulak, A. V. Kondratyuk, T. I. Kulak, M. P. Samtsov, D. Meissner:
Chem. Phys. Lett. 378, 95 (2003) 432
[35] I. Pócsik, M. Koós, S. H. Moustafa, J. A. Andor, O. Berkesi, M. Hundhausen:
Michrochim. Acta [Suppl.] 14, 755 (1997) 432
[36] M. Veres, M. Koós, I. Pócsik: Diamond Relat. Mater. 11, 1115 (2002) 434
[38] R. E. Shroder, R. J. Nemanich, J. T. Glass: Phys. Rev. B 41, 3738 (1990) 437
[39] F. J. Owens: Physica E 25, 404 (2005) 437, 439
Index
π states, 423, 426 a-C:H, 426, 428–443
σ states, 427 sp2 -bonded clusters, 426–429, 435,
a-C, 428 437–439, 442, 443
cluster conjugation level, 427 dispersive G band, 435–437, 439, 440

cluster size, 426–429, 435 non-dispersive G band, 435, 437, 439,
sp2 -chains, 427, 431, 432, 435–439, 443 440
chain length, 439, 443 graphene, 427
sp2 -rings, 426, 427, 431, 434, 435, 437, graphite, 426–428, 430, 431, 435
438, 442, 443 graphitization, 430, 438, 443
distorted rings, 435, 437, 438, 442
six-fold rings, 426 Hückel approximation, 426
sp3 /sp2 bonding ratio, 427
ta-C, 428 multi-wavelength Raman spectroscopy
(MWRS), 428
normal modes, 424
chemical vapour deposition (CVD)
radio frequency CVD, 429 phonons, 423, 425
cluster model, 426 zone centre phonons, 424
clustering, 426–429, 435, 437, 438 polarizability, 424, 425
coherence length, 428 polarizability tensor, 424
crystallite size, 428
Raleigh scattering, 424
D band, 427–432, 434–438, 440 Raman scattering, 423
D/G band intensity ratio, 428, 431, anti-Stokes scattering, 423
432, 438, 441 infrared excitation, 423, 429,
diamond, 426, 427 432–435, 437–439
diamond-like carbon (DLC), 428, 430, Raman intensity, 426
431, 437, 443 Raman spectrum, 426
dispersion, 423, 428, 432, 435–437, 439, Raman tensor, 424, 425
440, 443 resonant Raman scattering, 425, 428
selection rules, 425
electron-phonon coupling, 425 breakdown of the selection rules,
electron-photon coupling, 425 426
Stokes scattering, 423
G band, 423, 427–432, 435, 437, 438, visible excitation, 423, 424, 427, 429,
440 438
The Role of Hydrogen in the Electronic
Structure of Amorphous Carbon: An Electron
Spectroscopy Study
Lucia Calliari, Massimiliano Filippi, Nadhira Laidani, Gloria Gottardi,

Ruben Bartali, Victor Micheli, and Mariano Anderle
ITC-irst I-38050 Povo (Trento), Italy

calliari@itc.it
Abstract. Two processes are examined by electron spectroscopy techniques, name-

ly hydrogen incorporation into amorphous carbon (a-C) and hydrogen evolution
from hydrogenated amorphous carbon (a-C:H). The energy E p of the π + σ plas-
mon loss (related to the material mass density) turns out to be the most sensitive
indicator of the degree of H incorporation. Other spectral features, on the other
hand, consistently confirm that H acts as a stabilizer of the sp3 phase in amor-
phous carbon systems. H evolution is thermally activated from a-C:H once the
material stability threshold (≈ 100◦ C for the film of the present study) is exceeded.
It first leads to massive sp3 to sp2 conversion, followed by graphitic order devel-
opment. No univocal relationship exists between the sp2 fraction and the π band
photoemission intensity.
1 Introduction
Carbon is unique in the variety of crystalline and disordered structures it

can give rise to, owing to the variety of possible hybridizations (sp3 , sp2
and sp1 ) of its electron states. Among the disordered structures, amorphous
carbons are a class of metastable materials produced under conditions far
from thermodynamic equilibrium, so that the lower energy crystalline phase
is not formed. They are described as a mixture of mainly threefold and four-
fold coordinated atoms.
Because a one-to-one correspondence exists between carbon coordination
and bonding state for crystalline materials, diamond and graphite, the same
was assumed also for amorphous materials, leading to interchangeably char-
acterize them either in terms of the relative amount of threefold and four-
fold coordinated atoms or in terms of the fraction of sp2 - and sp3 -bonded
atoms. However, this is not strictly true for the amorphous state, where
strained configurations may result in intermediate hybridizations of the elec-
tron states [1, 2, 3]. As a consequence, a one-to-one relationship between
atomic coordination and electron states hybridization no longer holds in this
case [4, 5]. Nevertheless, a great deal of experimental work has been suc-
cessfully performed under the assumption that atoms in a-C can indeed be
classified according to a well-defined hybridization state (sp2 or sp3 ) [2].
448 Lucia Calliari et al.
In spite of its importance, however, just knowing the fraction of atoms in

one hybridization state or the other does not univocally determine the elec-
tronic structure of a-C [6]. This is due to the π-bonding-driven interaction
among sp2 sites [6], which, for a given sp2 fraction, may result in a variety
of electronic structures depending on the spatial distribution of the sp2 sites.
Basically, the distribution can be spatially inhomogeneous, with nanosized
graphitic domains randomly distributed through the system, or spatially ho-
mogeneous, with the sp2 sites spread through the sp3 phase like in a “raisin
cake” [7]. Systems of the former type, where a conjugated π electron struc-
ture exists in graphitic regions, are referred to as “graphitic” [8], and they
are described by Robertson’s cluster model [9]. Systems of the latter type,
where the graphitic structure is destroyed and a conjugated π electron system
does not exist, are referred to as “nongraphitic” [8]. They are described by a
revised version of the original Robertson’s cluster model [10]. Needless to say,
each of the above two categories includes a variety of materials which differ
either in the size of the graphitic domains or in the way sp2 sites interact
(isolated sites, pairs, chains, even- and odd-membered rings). Knowledge of
the electronic structure of amorphous carbon systems (needed to predict the
material properties) thus requires specifying the clustering features of the sp2
sites in addition to the sp2 fraction [6].
Closely related to amorphous carbons are their hydrogenated analogues
(a-C:H), whose properties are even more difficult to foresee as compared to
a-C systems. In fact, besides the sp2 fraction and the clustering features
of the sp2 sites, shared with a-C materials, the H content and the way it
is incorporated are in this case further structural features relevant to the
material properties [6].
When incorporated into the structure of amorphous carbon materials,
H primarily saturates π bonds, thereby converting sp2 C=C into sp3 ≡CH,
=CH2 and −CH3 bonds [6, 11]. In this sense, it generally leads to an in-
creased sp3 fraction and a wide energy gap (≈ 3 eV). However, the material
mechanical properties get poorer and its mass density decreases, mainly as a
result of the reduced cross-linking in the carbon network [11].
A number of techniques were employed to characterize a-C and a-C:H
materials. They can be divided into techniques that probe either the C atom
environment (diffraction techniques), or the electron states about the C atom
(electron spectroscopy techniques). In the former case, the atomic coordina-
tion, together with bond lengths and bond angle distributions, is measured.
In the latter case, the electronic structure is described, from which the sp2
fraction and the possible existence of a conjugated π electron structure can
in principle be derived.
In the present work, selected examples from the class of a-C:H materials
are investigated by electron spectroscopy techniques that probe the occu-
pied electronic structure. These include core level photoemission, X-ray and
The Role of Hydrogen in Amorphous Carbon 449
UV excited valence band (VB) photoemission, C KVV Auger emission and

electron energy loss spectroscopy (EELS) in the plasmon region.
The focus is first on the problem of unraveling effects due to changes in
the sp2 fraction, on the one hand, and in sp2 clustering, on the other hand.
The two are in fact often tangled in the features of electron spectra.
Second, we address the issue about how and to which extent H is incorpo-
rated into the structure of amorphous carbon. It is well known in this regard
that H cannot be directly detected by electron spectroscopy techniques, so
that its presence can only be indirectly revealed.
The paper is organized as follows. After the introduction, experimental
methods are given in Sect. 2. Sections 3 and 4 describe two experiments se-
lected to discuss the electronic structure changes brought about by adding
hydrogen to a-C (Sect. 3) and by subtracting hydrogen from a-C:H (Sect. 4).
In particular, Sect. 3 deals with the investigation of a series of films deposited
by graphite sputtering in Ar–H2 plasma, with variable H2 concentration in
the plasma feed gas. On the other hand, Sect. 4 refers to a vacuum annealing
experiment (from room temperature (RT) up to 700◦C) performed on a hy-
drogenated amorphous carbon sample in order to cause hydrogen evolution
from the film. Concluding remarks are given in Sect. 5.
2 Experimental
2.1 Hydrogen Incorporation in the Structure of a-C
Amorphous carbon films were deposited on silicon wafers by sputtering a

graphite target in 5 Pa rf (13.56 MHz) discharges at variable power. A con-
stant dc self-bias on the cathode was maintained at −550 V, corresponding
to an effective load power of 40 W. The H2 concentration in the feed gas
ranged from 0 to 100 vol. %, but the useful range for film growth was limited
to 0–84%. The total flux of H2 was 30 sccm. The samples were mounted on a
rotating support, at a distance of 8 cm from the cathode and without applica-
tion of any external bias, so that the films grew at the floating potential given
by the plasma. Both the cathode and the sample holder were water cooled.
The thickness of the films, as measured by a mechanical stylus profilometer,
ranged from 100 to 480 nm. X-ray photoemission spectra (XPS) and X-ray
excited Auger emission spectra were acquired within a SCIENTA ESCA200
instrument equipped with a monochromatic Al Kα (hν = 1486.6 eV) X-ray
source and a hemispherical analyzer. No electrical charge compensation was
required to perform the analysis. Electrons were detected along the surface
normal. The energy resolution, as measured on the Ag Fermi edge, was 0.4 eV
for the C 1s spectrum and 0.6 eV for the C KVV and VB emission spectra.
2.2 Temperature-Induced Hydrogen Evolution

from a-C:H
The a-C:H film used for the annealing experiment was deposited on (100) Si
wafers in a capacitively coupled RF-PECVD system (13.56 MHz) using a
mixture of CH4 –CO2 (50% CO2 ) as gas precursor. The sample was mounted
on the water-cooled anode and ion assistance during film growth was obtained
by applying a pulsed bipolar (±50 V) bias voltage, with a 100 kHz frequency
and a constant duty cycle of 50% (for the positive part of the pulse) [12].
The annealing experiment took place in a separated ultra high vac-
uum (UHV) system equipped with a sample preparation chamber and an
analysis chamber. The film was step annealed from RT to 700◦ C in the for-
mer, while electron spectra were measured, after each annealing step, in the
latter. The sample, clamped on a Mo sample holder, was heated by radia-
tion and electron bombardment from a tungsten filament at the back of the
sample holder. The temperature was measured by a nickel–chromel thermo-
couple. For each annealing step, the temperature was increased at a rate of
10◦ C/min up to the maximum temperature, where the sample was kept for 5
min. UV photoemission spectra were excited by the HeII (hν = 40.8 eV) line
from a He discharge lamp. C KVV Auger spectra were induced by X-rays
from a nonmonochromatic Mg Kα (hν = 1253.6 eV) source, while EEL spec-
tra were associated with a backscattered primary electron beam of 2 keV. All
spectra were acquired with a double-pass cylindrical mirror analyzer (CMA)
at an energy resolution, as measured on a Pd Fermi edge, of 0.5 eV for VB
photoemission and 1 eV for C KVV emission. For the EEL spectrum, the
energy resolution is 1.3 eV, as given by the FWHM of the backscattered
electron peak.
3 Hydrogen Incorporation in the Structure of a-C

The effects of hydrogenation are examined on a series of amorphous carbon
films deposited by graphite sputtering in Ar–H2 plasma [13, 14]. The H2
content in the gas mixture, hereafter referred to as [H2 ], varies from 0 to
84%. Contrary to a possible expectation, the hydrogen content in the films
does not increase monotonically with [H2 ]. The integrated FTIR absorbance
of the C–Hx bonds in the films, related to the bound hydrogen content,
exhibits [13] a sharp maximum for low (≈ 10%) [H2 ] and decreases beyond
this maximum as [H2 ] increases. The behavior was explained elsewhere [13]
in terms of the different graphite sputtering regimes occurring at low and
high H2 concentrations in the feed gas.
Figure 1 shows the low-energy EEL spectrum associated with C 1s photo-
electrons for two films deposited at [H2 ] = 0% and [H2 ] = 16%, respectively.
For carbon systems, this spectrum is dominated by two plasmon loss features:
a low-energy peak (≈ 4 − 6.5 eV) assigned to excitations involving π electrons
Fig. 1. EEL spectrum for a [H2 ] = 0 % (continuous line) and a [H2 ] = 16 % (dotted
line) film. The zero-loss peaks are normalized to a common height
only and a high-energy peak (≈ 23–34 eV) assigned to excitations involving

both π and σ electrons. The energy, E p , of the π + σ plasmon is related to
the material electron (and hence atomic) density through the relation:

ne 2
E p = ω = , (1)
0 m
which is derived within the free electron approximation. Here n is the valence
electron density taking part in the plasma oscillation (each C atom is assumed
to contribute four electrons), e is the electron charge, 0 is the vacuum di-
electric function and m is the free electron mass. The figure shows that the
film deposited in a hydrogen-containing plasma is characterized by a lower
E p , and hence by a lower mass density, as compared to the film deposited in
pure Ar plasma.
To measure E p , a linear background was subtracted and the π+σ plasmon
feature was fitted to a Gaussian peak. Its evolution over the film series is given
in Fig. 2. The straight line is the best fit through the [H2 ] = 0 points. At
[H2 ] = 0 %, E p is ≈ 25 eV, which is typical for unhydrogenated amorphous
carbon films [15]. As soon as [H2 ] becomes different from zero, an abrupt
decrease down to ≈ 22 eV is observed for [H2 ] = 3%. From this point on,
however, E p increases with [H2 ].
In C–H systems, a decrease in mass density is related to an increased H
content in the material, and it is not only due to the low H mass, but also to
the reduced average coordination number. For the present films, a variation
in the H content was indeed demonstrated by FTIR measurements [13]. We
therefore conclude from Fig. 2 that the H content in the films increases, goes
Fig. 2. Energy of the π + σ plasmon as a function of the H2 concentration in the

palsma feed gas
through a maximum and then decreases as the H2 concentration in the gas

mixture increases from 0 to 84%.
By enabling us to tell apart films differing in the H content, E p turns out
to be the most sensitive indicator of the film H content. In contrast, other
spectral features only single out films deposited at [H2 ] = 0% from films de-
posited at [H2 ] = 0, while they are not sensitive enough to discriminate among
films in the [H2 ] = 0 region [14]. For this reason, the effects of hydrogenation
on the film electronic structure are considered in Fig. 3 by comparing spectra
from films deposited at [H2 ] = 0% (continuous line) and [H2 ] = 0 (dotted line),
without specifying the [H2 ] value.
Figure 3a shows the EEL spectrum associated with C 1s photoelectrons
and already considered with regard to the π + σ plasmon energy. Here we
focus on the π plasmon, detected at ≈ 4.4 eV from the zero-loss peak on the
[H2 ] = 0 spectrum, but missing on the [H2 ] = 0 spectrum. The π plasmon is
known to probe graphitic ordering, rather than just the sp2 fraction [6,16,17].
For graphite, it is intense and it occurs at 6.5 eV from the zero-loss peak. It
has a FWHM of 2.4 eV, which implies a decay time τ of plasma oscillations
on the order of 10−15 s. The peak has similar features for samples where,
in spite of a missing long-range graphitic order, the C atoms are nonethe-
less predominantly bonded in warped graphite-like planes, such as in glassy
Fig. 3. Comparison between [H2 ] = 0 and [H2 ] = 0 spectra: π plasmon loss associ-
ated with C 1s photoelectrons (a), C 1s photoemission (b), VB photoemission (c),
C KVV emission (d). Spectra in each panel are normalized to a common height of
the most intense C feature. Binding and kinetic energies are referred to the Fermi
level. On part (d) the vertical line marked EF represents the Fermi level offset by
the C 1s binding energy
carbon [18] or carbon nanotubes [19]. If, however, disorder is introduced into
such threefold-coordinated materials and the size of graphitic regions becomes
smaller, the π plasmon peak shifts to lower energy, decreases in height and
broadens [19]. For “nongraphitic” amorphous carbon, where no graphite-like
regions exist, the π plasmon peak vanishes, even though the material is highly
sp2 bonded [16].
Figure 3a shows that π bonding exists in the film deposited in pure Ar

plasma, though the low energy and low intensity of the π plasmon peak sug-
gest a very limited clustering of the sp2 sites. On the other hand, the disap-
pearance of the feature for films grown in a hydrogen-containing atmosphere
is most likely indicative of π bonding saturation by hydrogen.
Such an effect is observed also on the two VB-related spectra, namely VB
photoemission (Fig. 3c) and C KVV Auger emission (Fig. 3d). (Remember
that the latter basically represents a self-fold of the VB density of states
(DOS) [20,21]). For [H2 ] = 0, both spectra exhibit a decrease in the DOS near
the Fermi level E F . While in fact π states reach E F for the film deposited
in Ar plasma, these states are washed out for films grown in Ar–H2 plasma.
The related high binding energy shift in the VB top, E V (enlarged in the
inset of Fig. 3c is estimated around 1 eV, which represents a lower limit for
the energy gap. The total gap width cannot be measured, in fact, by probing
only the occupied electron states.
In addition to sweeping out π states from the Fermi level region, the
presence of H2 in the plasma feed gas also causes a narrowing of the σ band.
This is seen on the VB photoemission spectrum (Fig. 3c) for the 2s states
(the different widths of the C 2s peaks are not an artifact due to the presence
of the Ar 3s peak) and on the C KVV spectrum (Fig. 3d) for the 2p states.
Remember that the latter spectrum is contributed by σ2p states around the
peak maximum and by π2p states at the spectrum leading edge [21]. Both
the 2s and the 2p DOS thus decrease in width when hydrogen is added to
the amorphous carbon network.
By modeling X-Ray excited VB photoemission (s DOS sensitive) and
X-Ray emission (p DOS sensitive) spectra, it was demonstrated [22] that a
narrowing in the σ2s and in the combined σ2p and πp states distribution
occurs on moving from graphite to diamond. Experimentally, a narrowing of
spectral features on going from graphite to diamond has been observed for
X-Ray VB photoemission spectra [23] and for C KVV Auger spectra [24].
The narrowing of spectral features observed for hydrogen-containing films is
therefore understood in terms of an increased sp3 fraction.
Note that Ar is seen on the [H2 ] = 0 spectra. Due to a favorable photoion-
ization cross section [25], it is particularly evident on the VB photoemission
spectrum, though its concentration, as estimated from the Ar 2p and C 1s
peaks, is on the order of a few at. %. In contrast, Ar is not observed on the
[H2 ] = 0 spectrum, because of a decreased density of Ar+ ions in the plasma
when [H2 ] = 0 [13].
Figure 3b shows the C 1s spectrum after subtracting a Shirley back-
ground. The spectra are aligned to a common energy of 284.45 eV (char-
acteristic of [H2 ] = 0 spectra). A mild electrostatic charging occurring for
[H2 ] = 0 films under photon irradiation prevents us from discussing possible
changes in the C 1s binding energy, so that we only focus on peak shape.
The FWHM of the two spectra is around 1.2 eV. A small high binding en-
ergy tail is observed on all spectra and it is assigned to C–O bonds, given
that a certain amount of oxygen (4–10 %) is present in the film near-surface
region. However, a statistics of over 15 samples consistently shows that films
produced in a pure Ar plasma exhibit a high binding energy tail which can-
not be accounted for by C–O components unless unrealistic O concentrations
are assumed (note that the two spectra in Fig. 3 refer to samples having the
same O content).
Line asymmetry in core-level photoemission is associated with small en-
ergy electron–hole excitations near the Fermi level as a response to the cre-
ated photo-hole. As such, it is characteristic of metals and semimetals, like
graphite [26, 27]. The concept was used to describe the C 1s spectrum from
amorphous carbon systems [28, 29, 30], and asymmetry was shown to be re-
lated to the existence of a π DOS near the Fermi level. This is also the case
here: an asymmetric C 1s lineshape is observed only for films (deposited in
pure Ar plasma) whose π states reach the Fermi level.
All spectra in Fig. 3 thus consistently show that H removes π states from
the VB top, thereby opening a gap, and leads to an increased sp3 fraction.
4 Temperature-Induced Hydrogen Evolution

from a-C:H
Hydrogen evolution from a-C:H materials is easily obtained, for example, by
thermal annealing and was experimentally demonstrated by either measuring
the gaseous species evolving from the films [31, 32] or by characterizing the
film composition via H sensitive techniques, such as FTIR [31, 33, 34].
Here we present results from an experiment in which a hydrogenated
amorphous carbon film is step-annealed in vacuum from RT up to 700◦ C.
After each annealing step, the near-surface composition and structure are
characterized in terms of UV-excited VB photoemission, Auger emission and
EELS in the plasmon region.
The HeII excited VB photoemission spectrum from a sample annealed at
400◦ C (dotted line) is compared to the spectrum from a RT sample (contin-
uous line) in Fig. 4. A background, consisting of a third-degree polynomial
extrapolated from the energy regions at either side of the peak, was sub-
tracted from the spectra, which were then normalized to a common height.
The difference spectrum (400◦ C− RT) is also given. The spectra differ in two
energy regions, namely from 0 to 5 eV and from 8 to 12 eV. Thermal anneal-
ing clearly removes electron states from the latter region and generates new
states in the former. The upper VB (0–5 eV) is where π states are located,
while the region around 10 eV is where H-induced states in the σp band
occur. This was shown by calculated DOS [9, 35], and it was also demon-
strated in experiments in which a diamond surface undergoes repeated cycles
of hydrogenation and annealing [36, 37]. Such cycles lead, respectively, to a
H-terminated and clean, reconstructed surface. Correspondingly, a transfer of
Fig. 4. HeII (hν = 40.8 eV) excited VB photoemission spectra: RT (continuous

line), 400◦ C (dotted line). The difference spectrum (400◦ C − RT) is also shown
states from the π region to the C–H states region and vice versa is observed.
For diamond, π states are associated with the reconstruction of the clean sur-
face [37], while the H termination inhibits surface reconstruction and freezes
the atoms in their tetrahedral bulk configuration. Considering the analogy
with diamond on the one hand, and the fact that H is known [6] to stabi-
lize the sp3 phase in a-C:H on the other hand, C–H states between 8 and
12 eV have most likely to be associated with sp3 -hybridized electron states.
Annealing at 400◦ C therefore causes the disappearance of H-induced (sp3 -
hybridized) electron states and the corresponding formation of π states. In
other words, dangling bonds generated by temperature-induced H desorption
are saturated via π bond formation.
The evolution of H-induced states and π states intensity over the consid-
ered temperature range is depicted in Fig. 5a, b. To measure the respective
intensity, spectra were handled as in Fig. 4. For each temperature T , the
difference between the spectrum at temperature T and the RT spectrum was
calculated. On the difference spectrum, the negative area in the 8–12 eV re-
gion is assigned to C–H states, while the positive area in the 0–5 eV region
is assigned to π states.
The C–H states contribution to the spectrum (Fig. 5a) is stable up to
a temperature of 100◦ C. It decreases markedly on going from 100 to 250◦ C
Fig. 5. Temperature evolution of three spectral features: C–H states area (a), π
states area (b), ∆ (c). See text for the intensity measurements (referred to the RT
spectrum). The horizontal dashed line in part (c) gives the ∆ value we measured
for graphite
and tends to stabilize above this temperature, where only a slight decrease
is observed.
Also the π states contribution to the spectrum (Fig. 5b) remains constant,
at the RT value, up to 100◦ C. From 100 to 250◦ C, it increases sharply. Above
250◦ C, however, no stabilization occurs. On the contrary, the π states’ con-
tribution to the spectrum keeps on increasing, though to a lower rate, over
the 250–500◦C range, and it increases at a faster rate over the 500–700◦C
range. Thus only up to 250◦ C does their behavior mirror the C–H states’
evolution. It reveals that, from 100◦C to 250◦ C, H evolves from the film and
π states are formed at the expense of C–H (sp3 -hybridized) states. At higher
temperatures, however, the fate of the π states is not linked to the fate of
C–H states.
Figure 5c gives the temperature evolution of the energetic distance ∆,
measured on the derivative C KVV spectrum and defined in the inset. ∆ is
stable, around 16 eV, up to 150◦ C. It markedly increases to ≈ 21 eV on going
Fig. 6. HeII (hν = 40.8 eV) excited VB photoemission spectra: T = 400◦ C (con-
tinuous line), T = 700◦ C (dotted line). Also shown is the difference spectrum
(700◦ C−400◦ C)
from 150 to 300◦C, while it is nearly stable above 400◦ C. ∆ is known [38,39] to
be a sensitive probe of the electron states’ hybridization about the C atom.
It increases from ≈ 13 eV for fully sp3 -hybridized diamond to ≈ 23 eV for
fully sp2 -hybridized graphite, and it takes intermediate values for amorphous
carbon materials. The marked increase in ∆ over the 150–300◦C range is
therefore understood as a change in the electron states’ hybridization from
sp3 to sp2 . At higher temperatures, the transformation tends to saturate to
a value near that of graphite.
A consistent picture is thus derived from the three panels of Fig. 5 for T ≤
300◦ C. The film is stable up to 100–150◦C, at which temperature structural
changes are initiated. These consist in H evolution from the film between
100◦ C and 300◦C, resulting in sp3 to sp2 conversion. Above 300◦C, ∆ follows
the C–H states evolution, thus indicating that, once the H desorption process
is exhausted, the same also holds for sp3 to sp2 conversion.
It seems therefore that the π band intensity gain revealed by VB photoe-
mission (Fig. 5b) above 300◦ C cannot be assigned to an increase in the sp2
fraction. To clarify this point, Fig. 6 compares the 400◦C (continuous line)
to the 700◦ C (dotted line) HeII-excited VB photoemission spectrum. Both
spectra are background subtracted and normalized to a common height, as
explained before. The difference spectrum (700◦C–400◦ C) is also shown. At
variance with respect to Fig. 4, the π band intensity gain observed here is
not balanced by a corresponding intensity decrease in the σp band. Note that
the apparent widening of the σp band around 5 eV (leading to a peak in the
difference spectrum) has to be nearly entirely ascribed to the intensity gain
of the π band, which increases in height and width, on going from 400 to
700◦ C. The observed behavior suggests that the π band intensity gain in VB
photoemission (Fig. 5b) has different origins in different temperature regimes.
To gain further insight into the matter, Fig. 7 considers another π-related
spectral feature, namely the π plasmon of the EEL spectrum associated with
a backscattered primary electron beam of 2 keV. The π plasmon spectrum
is given at four selected temperatures: RT, 400, 600 and 700◦ C. At RT, a
nearly vanishing π peak is observed. As explained in the previous section,
this indicates either a very small sp2 fraction or, in any case, a homoge-
neous distribution of the sp2 sites within the sp3 phase, without formation of
graphite-like clusters. At 400◦ C, a π plasmon peak is observed, which reveals
the presence of π states at this stage. As T increases from 400 to 600 and
700◦ C, the π peak increases in height, its energy increases from 5.2 to 5.6 to
5.8 eV, while its FWHM decreases from 4 to 3.5 to 3 eV. The evolution in-
dicates that [19], once the sp2 fraction has grown up, graphite-like ordering
develops in the system through the formation of graphite-like clusters with
progressively increasing size.
This provides us with a key to understand the temperature evolution of
the π band intensity in the VB photoemission spectrum (Fig. 5b). Below
300◦ C where electron states are transferred from the σp to the π band, its
increase is due to an increase in the sp2 fraction. On the other hand, its
further intensity gain above 300◦C must be due to the rearrangement of sp2
sites into clusters of increasing size. Apparently, π states in a graphite-like
network have a greater photoemission yield than π states in a nongraphitic
environment.
In a recent paper [40], the question was raised whether the photoemission
π band intensity is affected by the sp2 fraction alone or in addition by the π
network structure. From the present annealing experiment, we conclude that
it is affected by both structural features. We have, in fact, clear indications
(∆ and C–H states evolution) that our film is highly sp2 -hybridized once
T is around 300–400◦C, so that only below these temperatures can the π
band intensity gain be assigned to an increase in the sp2 fraction. At higher
temperatures, on the other hand, the observed further increase in the π band
intensity can only be brought about by the film transformation from non-
graphitic to graphitic.
5 Concluding Remarks
The effects of amorphous carbon hydrogenation are examined by consider-
ing a series of films deposited in Ar–H2 plasma with variable H2 concen-
Fig. 7. π plasmon peak (EEL spectrum) at four selected temperatures. The zero-
loss peaks are normalized to a common height
tration. Hydrogen is incorporated into films grown in a hydrogen-containing

atmosphere. This leads to a reduced mass density, mainly as a result of re-
duced cross-linking in the hydrogenated network.
Hydrogen incorporation markedly changes the electronic structure of oth-
erwise graphite-like amorphous carbon films: π states are swept out from the
Fermi level region, and the system turns from semimetallic into semiconduct-
ing, due to the opening of a gap (greater than 1 eV). The saturation of π
bonds leads to an increased sp3 fraction, revealed by a narrowing in the 2s and
2p VB DOS, and to the disappearance of small graphitic domains revealed,
for the unhydrogenated films, by the π states reaching E F , the asymmetric
C 1s line and the measurable π plasmon intensity on the EEL spectrum.
The reverse process, dehydrogenation of a-C:H, is examined by thermally
inducing hydrogen evolution from the system. Three temperature regimes
are basically distinguished. From RT up to ≈ 100◦ C, the investigated film is
stable. Structural modifications are initiated above this temperature. From
≈ 100 to ≈ 300◦ C, hydrogen desorbs from the system, causing massive sp3
to sp2 conversion. At even higher temperatures, a reorganization of the sp2
sites into graphite-like regions of progressively increasing size takes place.
No univocal relationship exists between the sp2 fraction and the intensity
of the π band in UV excited VB photoemission.
Acknowledgements
The work was funded by Provincia Autonoma di Trento (Fondo Progetti per
la Ricerca).
References
[1] C. Gao, Y. Y. Wang, A. L. Ritter, J. R. Dennison: Phys. Rev. Lett. 62, 945
(1989) 447
[2] D. R. McKenzie: Rep. Prog. Phys. 59, 1611 (1996) 447
[3] P. Ruffieux, O. Gröning, M. Bielmann, P. Mauron, L. Schlapbach, P. Gröning:
Phys. Rev. B 66, 245416 (2002) 447
[4] R. Haerle, E. Riedo, A. Pasquarello, A. Baldereschi: Phys. Rev. B 65, 045101
(2001) 447
[5] J. T. Titantah, D. Lamoen: Phys. Rev. B 70, 075115 (2004) 447
[6] J. Robertson: Mater. Sci. and Engineering: R 37, 129 (2002) 448, 452, 456
[7] M. M. Golzan, D. R. McKenzie, D. J. Miller, S. J. Collocott, G. A. J. Ama-
ratunga: Diamond Relat. Mater. 4, 912 (1995) 448
[8] K. Takai, M. Oga, H. Sato, T. Enoki, Y. Ohki, A. Taomoto, K. Suenaga,
S. Iijima: Phys. Rev. B 67, 214202 (2003) 448
[9] J. Robertson, E. P. O’Reilly: Phys. Rev. B 35, 2946 (1987) 448, 455
[10] J. Robertson: Diamond Relat. Mater. 4, 297 (1995) 448
[11] J. K. Walters, R. J. Newport: J. Phys.: Condens. Matter 7, 1755 (1995) 448
[12] G. Gottardi, N. Laidani, R. Bartali, M. Filippi, L. Calliari, R. S. Brusa, S. Mari-
azzi, C. Macchi, M. Anderle: Diamond Relat. Mater. 14, 1031 (2005) 450
[13] N. Laidani, R. Bartali, P. Tosi, M. Anderle: J. Phys. D: Appl. Phys. 37, 2593
(2004) 450, 451, 454
[14] L. Calliari, M. Filippi, N. Laidani, M. Anderle: J. Electron Spectr. Relat.
Phenomena 150, 40 (2006) 450, 452
[15] J. Fink, T. Müller-Heinzerling, J. Pflüger, A. Bubenzer, P. Koidl, G. Crecelius:
Solid State Commun. 47, 687 (1983) 451
[16] J. Kulik, Y. Lifshitz, G. D. Lempert, J. W. Rabalais, D. Marton: J. Appl.
Phys. 76, 5063 (1994) 452, 453
[17] Y. Lifshitz: Diamond Relat. Mater. 12, 130 (2003) 452
[18] D. G. McCulloch, S. Prawer, A. Hoffman: Phys. Rev. B 50, 5905 (1994) 453
[19] M. M. Brzhezinskaya, E. M. Baitinger, V. V. Shnitov: Physica B 348, 95 (2004)
453, 459
[20] J. E. Houston, J. J. W. Rogers, R. R. Rye, F. L. Hutson, D. E. Ramaker: Phys.
Rev. B 34, 1215 (1986) 454
[21] L. Calliari, G. Speranza, J. C. Lascovich, A. Santoni: Surf. Sci. 501/3, 253
(2002) 454
[22] K. Endo, S. Koizumi, T. Otsuka, M. Suhara, T. Morohasi, E. Z. Kurmaev,
D. P. Chong: J. Computational Chemistry 22, 102 (2001) 454
[23] F. R. McFeely, S. P. Kowalczyk, L. Ley, R. G. Cavell, R. A. Pollak,

D. A. Shirley: Phys. Rev. B 9, 5268 (1974) 454
[24] A. P. Dementjev, M. N. Pethukov: Surf. Interf. Analysis 24, 517 (1996) 454
[25] J. Yeh: Atomic Calculation of Photoionization Cross-Sections and Asymmetry
Parameters (Gordon and Breach, Longhorne 1993) 454
[26] P. M. T. M. van Attekum, G. K. Wertheim: Phys. Rev. Lett. 43, 1896 (1979)
455
[27] F. Sette, G. K. Wertheim, Y. Ma, G. Meigs, S. Modesti, C. T. Chen: Phys.
Rev. B 41, 9766 (1990) 455
[28] J. Diaz, G. Paolicelli, S. Ferrer, F. Comin: Phys. Rev. B 54, 8064 (1996) 455
[29] J. Diaz, S. Anders, X. Zhou, E. J. Moler, S. A. Kellar, Z. Hussain: Phys. Rev.
B 64, 125204 (2001) 455
[30] J. Hong, S. Lee, C. Cardinaud, G. Turban: J. Non-Crystalline Solids 265, 125
(2000) 455
[31] Y. Bounouh, M. L. Thye, A. Dehbi-Alaoui, A. Matthews, J. P. Stoquert: Phys.
Rev. B 51, 9597 (1995) 455
[32] N. M. Conway, A. C. Ferrari, A. J. Flewitt, J. Robertson, W. I. Milne, A. Tagli-
aferro, W. Beyer: Diamond Relat. Mater. 9, 765 (2000) 455
[33] M. Clin, M. Benlahsen, A. Zeinert, K. Zellama, C. Naud: Thin Solid Films
372, 60 (2000) 455
[34] L. Calliari, M. Filippi, N. Laidani: Surf. Interf. Analysis, 36, 1126 (2004) 455
[35] J. Schäfer, J. Ristein, R. Graupner, L. Ley, U. Stephan, T. Frauenheim,
V. S. Veerasamy, G. A. J. Amaratunga, M. Weiler, H. Ehrhardt: Phys. Rev.
B 53, 7762 (1996) 455
[36] R. Graupner, J. Ristein, L. Ley: Surf. Sci. 320, 201 (1994) 455
[37] B. B. Pate: Surf. Sci. 165, 83 (1986) 455, 456
[38] J. C. Lascovich, R. Giorgi, S. Scaglione: Appl. Surf. Sci. 47, 17 (1991) 458
[39] D. E. Ramaker: Chemical information from Auger lineshapes, in D. Briggs,
J. T. Grant (Eds.): Surface Analysis by Auger and X-Ray Photoelectron Spec-
troscopy (IM Publications and Surface Spectra Ltd, UK 2003) pp. 465–500
458
[40] P. Reinke, M. G. Garnier, P. Oelhafen: J. Electron Spectr. Relat. Phenomena
136, 239 (2004) 459
Index
π states, 452, 456 hydrogen desorption, 455–458, 460
a-C, 447
a-C:H, 447 photoemission spectroscopy, 450,
453–461
Auger emission spectroscopy, 449–455, plasmon energy, 451–453
457
UV photoemission spectra (UPS), 450,
electron energy loss spectroscopy 461, 462
(EELS), 449–452, 458–460
electron energy loss spectroscopy x-ray photoemission spectroscopy
(EELS), 455 (XPS), 449
UV-Induced Photoconduction in Diamond
Emanuele Pace1 , Antonio De Sio1 , and Salvatore Scuderi2

1
Dipartimento di Astronomia e Scienza dello Spazio, Università di Firenze,
Largo E. Fermi 2, I-50125 Firenze, Italy
2
INAF – Osservatorio Astrofisico di Catania, Via S. Sofia 78, I-95123 Catania,
Italy
pace@arcetri.astro.it
Abstract. Owing to its physical properties, diamond is a very promising candi-

date for UV photon detection. The synthesis of very high quality polycrystalline
diamond thick films and single crystals is presently accomplished routinely using the
chemical vapour deposition (CVD) technique. Metal/diamond/metal ohmic junc-
tions are produced by depositing electrical contacts on the front surface (coplanar
contacts) or on both front and back (sandwich contacts). The investigation of dia-
mond properties in the wavelength range 100–300 nm is of considerable interest in
order to attain ultraviolet (UV) detectors having high quantum efficiency against
no sensitivity to visible photons, along with radiation hardness and chemical inert-
ness. Improvements in diamond synthesis techniques and processing technology now
make available very high quality polycrystalline diamond films as well as optical-
grade single crystals with dimensions sufficiently large for developing detectors.
Several problems limiting the performance have been solved, and single-pixel de-
tectors are close to being suitable for exploitation. This paper reviews the status of
the art of diamond-based UV detectors, their application in specific fields, such as
excimer lasers, photolithography and space experiments, and the perspectives.
1 Introduction
Despite their steady improvement over the last decades, the UV imaging de-
tectors currently available exhibit some limitations, inherent to silicon tech-
nology or to electron multiplication, that become crucial, for example, in the
context of space missions or UV laser lithography where reliability, dura-
bility, low cost, sensitivity, radiation hardness and solar blindness are re-
quired. In order to produce an innovative solid-state detector that copes with
this demand, the proper material must be selected among those having wide
band-gap energy, because of their negligible absorption of visible photons and
intrinsic very low dark currents. Currently, nitrides (GaN, AlGaN, InGaN),
SiC and synthetic diamond are widely investigated for these applications. Di-
amond has a unique combination of superior physical properties, displacing
concurrent materials, which makes it the ideal candidate for photoconductive
detectors replacing silicon in UV applications. Recent advances in the depo-
sition techniques allow the synthesis of diamond films with such exceptional
properties that overwhelm those of natural diamond in many cases. These
464 Emanuele Pace et al.
properties make synthetic diamond a “frontier” material for advanced elec-

tronic applications in harsh environments where conventional devices cannot
operate without special precautions.
Nowadays, the CDV technique makes available thick, large-area polycrys-
talline films. Many research teams have thoroughly investigated UV detectors
using interdigitated electrodes on polycrystalline CVD diamond [1, 2, 3, 4, 5,
6, 7, 8, 9, 10, 11, 12], and diamond-based UV detectors are commercially avail-
able [13]. Unfortunately, the results of these efforts invariably suggest that,
even if the performance of single-pixel detectors is dramatically improved,
polycrystalline material is unsuitable for imaging applications. The response
of polycrystalline diamond detectors is, in fact, not uniform over the area of
the detector; they show long time-constants and a response which varies with
time. The homogeneity of homoepitaxial CVD diamond single crystals and
the absence of grain boundaries, with respect to polycrystalline films, holds
out the promise to overcome all of these difficulties. This paper describes
the state of the art of diamond-based UV photon detectors. After a general
overview of the field, the main physical properties of diamond will be dis-
cussed in order to explain why diamond is considered a promising material
for UV detectors. The synthesis of diamond layers will be briefly described to
emphasize the crucial role of the material quality and how it correlates with
the detector performance. The following sections will describe the principle
of charge photogeneration and of detector types and operation along with
the expected performances and perspectives. A concise description of some
relevant applications of diamond UV detectors will conclude this review.
2 Overview
Synthetic diamonds were available until the late 1980s only as small crys-
tallites with variable properties. The emergence of new and improved tech-
niques, particularly CVD [14], to grow thin films of diamond encouraged
research into solid-state devices made from this material. Diamond has un-
common physical and mechanical properties: it is very hard, chemically inert,
an excellent thermal conductor [15] and has outstanding electrical and optical
characteristics [16]. Such appealing properties have attracted academic and
industrial interest to use them in practical applications. Synthesis techniques
have made important progress and Europe is at present the undisputed leader
in the production of CVD diamond. This material has been so far exploited
commercially for cutting tools, hard coatings, cold emitters for displays [17],
windows [18], thermal conductors, etc. For example, high-power CO2 laser
and gyrotron diamond windows have been produced in Europe [19]. In Japan,
Surface Acoustic Wave (SAW) devices [20] for GHz frequency telecommuni-
cations have been commercialised by Sumitomo. In the USA, CVD diamond
is intensively studied for various applications such as field emission displays,
SAWs, etc. For these applications, however, the quality of the material does
UV-Induced Photoconduction in Diamond 465
not have to be as high as that required by electronics, optoelectronics and de-

tectors, where top-level quality of diamond films is absolutely necessary [21].
Presently, CVD techniques produce mainly polycrystalline diamond layers
over nondiamond substrates (typically silicon) [14]. This method is advan-
tageous, because large sizes and controlled quality can be obtained. Nev-
ertheless, the polycrystalline structure can introduce a large concentration
of impurities, defects and grain boundaries that can trap or recombine the
photogenerated free charges [22, 23]. These interactions strongly influence
the device response by reducing the photoinduced current and slowing the
response time [24]. The distribution of these defects is not homogeneous in
the samples, and it can be cause of nonuniform sensitivity on the detector
surface. Moreover, they are responsible for the instability of devices due to
polarisation effects and long-term drifts caused by accumulation of charge by
particular deep-traps [25, 26]. These effects are not yet fully understood, and
their control is still one of the essential tasks to develop good functioning de-
vices. In addition, the nature of the surface [27,28] and it difficult to fabricate
reliable and reproducible electric contacts with peculiar geometry [29, 30]. To
date, the technology to produce high-quality polycrystalline diamond films
and devices is not firmly established [31, 32], even though several groups
have studied the opportunities offered by diamond as a radiation-sensitive
material [33].
So, the use of high-quality single crystals to produce devices is becom-
ing a necessity, especially for those groups that recently have been en-
gaged in producing pixel arrays. Therefore, the interest in single-crystal di-
amond growth techniques has increased. High-pressure and high-tempera-
ture (HPHT) single-crystal diamond synthesis is long well known. HPHT
diamonds are more homogeneous but are not completely defect-free, so they
still present problems in electronic applications. In fact, it is very difficult to
control HPHT parameters, and consequently they are not reproducible, with
nitrogen and metal catalyst inclusions. Elimination of these contaminations
increases the quality, although the samples are smaller and more expensive.
Recently, the synthesis techniques have been improved, and homoepitaxial
CVD diamond growth has increased. This technique permits the fabrication
of high-quality IIa-type single-crystal diamond samples [34, 35, 36]. The limit
of homoepitaxial CVD growth is still the size of the samples, but now single-
crystal diamond of 1 cm2 with exceptional electronic performances are avail-
able. For instance, charge mobility has increased up to 4500 and 3800 cm2 /Vs
for electrons and holes, respectively, increasing the collection distance up to
several millimetres in contrast to the hundred of microns of the best polycrys-
talline diamond films [37, 38, 39, 40, 41]. Even if the homoepitaxial growth is
more complex, it allows the realization of very promising devices, as reported
in [42, 43, 44, 45, 46]. UV detectors currently available for industrial, scientific
and medical applications are mainly based on well-developed, low-cost and
versatile materials. Although well adapted to the detection of low-energy radi-
ation, they are unsuitable whenever visible-blind, chemically inert and radia-
tion-hard sensors are required. In such circumstances, the detectors currently
used often comprised standard diodes or CCDs with shields, filters and cool-
ing systems. Moreover, cooling and protection from aggressive environments
requires sensors sealed within cameras. The use of diamond offers unmatched
performance in harsh environments or in the presence of high-doses of ion-
ising radiation. Other semiconductors, such as Si, Ge or new wide-band gap
materials, cannot be used in such environments without special precautions.
A number of potential applications exist where diamond is the only viable
solution; these range from environmental monitoring to oil recovery and from
the nuclear industry to optoelectronics. One of the most promising is UV pho-
ton sensing, in particular in satellite-borne instruments [47]. In space environ-
ment, current UV detectors, such as Si diodes, photomultipliers or CCDs, suf-
fer from problems affecting their performances or causing heavy instrument
payloads. Some of these problems are radiation hardness, visible-light sensi-
tivity, cooling systems, high-voltage biasing and thermal stability. Diamond
appears to be the most interesting solution, as its wide band gap (5.5 eV)
results in a very low leakage current (no cooling systems required) and in a se-
lective absorption of light with wavelengths shorter than 225 nm (visible-blind
detectors) [48]. Electronic properties [16, 49] and radiation hardness [50, 51]
are other attractive features. These characteristics have stimulated research
on diamond UV detectors. Experimental groups in the USA have tested nat-
ural diamond [2, 3]. Again, natural diamond is an expensive limited resource
and the optimal crystal must be selected among many of them. Nevertheless,
the encouraging results have pushed US and European teams to investigate
the functional properties of CVD diamond. Research projects involving inter-
national collaborations between universities and research centres have been
funded and are under way to study the feasibility of diamond UV detectors
for scientific and industrial applications [45, 52, 53, 54, 55, 56, 57, 58, 59, 60]. So
far, there was only one industrial product: a photoconductive detector com-
mercialised by Centronic, following an EU–Brite–Euram supported project.
This detector had a high sensitivity limited to a narrow UV band [4, 6], and
it relied on the surface properties of the material. The response time was im-
proved at the expense of sensitivity [54], and it was comparable to or faster
than the response time of well-developed silicon detectors. High-speed UV
detectors have been produced to follow short laser pulses [61, 62, 63, 64]. Fast
response to pulsed radiation can be properly ascribed to charge excitation
from valence to conduction band, typically a very fast process. However, many
applications require a fast response also to continuous radiation sources [65]
that induce different excitation mechanisms [66]. These problems need further
study but presently prevent end users from using diamond detectors.
A new frontier in the field of UV detectors is the attempt to achieve
bidimensional pixel arrays. Applications in UV astronomy, photolithography,
and laser and synchrotron beam characterisation and profiling should benefit
from the availability of such imagers. However, this research is presently at

the forefront of the diamond device technology and the nonuniformity of
diamond polycrystalline layers represents one of the main limitations to a
full development. Single crystals may be the solution, but their present size
prevents the development of large formats.
3 Properties of Diamond
Diamond crystal has a cubic structure formed by one carbon atom surrounded
by four tetrahedrally arranged nearest neighbours (sp3 σ-type bonds). This
crystal structure is found also in silicon and germanium. The bond distance
in diamond between two nearest neighbours is 0.154 nm, compared with
0.234 nm for silicon and 0.245 nm for germanium. As a result, the atomic
number density in diamond is 1.77 × 1023 cm−3 , the highest of any mate-
rials at normal pressure, and its mass density is 3.52 g cm−3 . The cohesive
energy of diamond, i.e., the energy required to disassemble a solid into its
constituent atoms, is almost twice as large as that of silicon and germanium.
This dense structure and strong bonding give diamond extreme hardness and
wear resistance.
The high propagation velocity of phonons is an effect of such a compact
structure. As the diamond elementary cell contains two carbon atoms, the
phonon spectrum has both acoustic and optical branches. Diamond is an
insulator, with an insignificant number of electrons in the conduction band.
This implies that heat conduction is due to acoustic phonons, providing the
extremely high thermal conductivity of diamond (25 W cm−1 s−1 , compared
to 4 W cm−1 s−1 for copper at room temperature). Diamond can efficiently
dissipate its own heat or the heat generated elsewhere and can tolerate high
temperature. It is one of the rare materials that naturally conduct heat yet
electrically insulate.
Its inertness to acids or other chemical agents, deriving from the high
binding energy of carbon atoms in the diamond lattice, is another important
property of diamond that is essential for applications in harsh environments.
The electrical and optical properties of diamond result from its wide band
gap energy, i.e., Eg = 5.470 ± 0.005 eV at 295 K. Electrically, diamond is an
insulator, but it can behave as a semiconductor by introducing donor or
acceptor atoms. Like silicon, diamond has an indirect band gap. The max-
imum of the valence band lies at the centre of the Brillouin zone in the
reciprocal space; the lowest point of the conduction band is near the zone
boundary in the K = (111) direction, where the band gap, Eg , is defined.
Diamond has a large breakdown electric field (about 107 V cm−1 ) and its sat-
uration velocity is approximately 107 cm s−1 . The electron and hole mobility
is 1800 cm2 V−1 s−1 and 1200 cm2 V−1 s−1 , respectively. Achieving current
saturation at a high electric field and high charge mobility are interesting
properties of diamond-based devices in view of electronic applications. Ta-

ble 1 lists some typical material and operating parameter values of diamond
compared with those of silicon.
Table 1. Physical properties of diamond at room temperature
Properties Units Diamond Silicon

Atomic number − 6 14
Density g cm−3 3.5 2.32
Binding energy eV/atom 7.37 4.63
◦
Evaporation temperature C 4100 1420
Breakdown voltage V 107 103 *
Band gap energy eV 5.5 1.1
Resistivity Ω cm > 1012 105
Dielectric constant − 5.6 11.7
Electron mobility cm2 V−1 s−1 1800 1500
Hole mobility cm2 V−1 s−1 1200 480
Saturation velocity µm/ns 220 100
∗ for a p–n junction
Optically, pure diamond exhibits no photon absorption or luminescence

in the visible spectral range, at λ > 225 nm. This property is very important
for visible-blind UV photodetectors. On the other hand, photons with λ <
225 nm are absorbed when they impinge on diamond. This process gives rise
to the electron excitation from the valence band to the conduction band, and
diamond exhibits photoinduced electrical conductivity. This will be discussed
in a following paragraph. Absorbance and reflectance are other important
optical properties that contribute to the performance of detectors [3, 16, 67].
4 Synthesis of Diamond
The promising characteristics of natural diamond for applications in detectors

and in harsh environments have been demonstrated in the USA. In spite of
the very limited resource of appropriate quality natural gems and their cost,
people have studied the functional properties of this material.
The advent of synthetic diamond films, in particular those produced by
CVD, has radically increased the availability of diamonds for technological
and mechanical applications. The production of CVD diamond films suitable
for detector applications is still a challenge that is being pursued worldwide,
and it is one of the current technological frontiers in the production of ad-
vanced materials.
The heteroepitaxial CVD growth allows the synthesis of large polycrys-
talline slices (up to 6 in diameter) on a variety of substrates [33, 54, 68, 69,
70, 71]. This approach is based on creating a hot plasma region (> 1500 ◦ C)
near a substrate (typically at 700–800 ◦ C) at a lower pressure than other tech-
niques. In the vapour phase, the thermal decomposition of carbon-containing
gases such as methane occurs.
The hydrocarbon gas is mixed with a large concentration of molecular
hydrogen gas. A power source excites the gas mixture so that the carbon
atoms link together onto the substrate, forming the sp3 bonds leading to
the diamond lattice and eventually the polycrystalline film. In various CVD
processes, it is common to start from a reactive gas phase, which is composed
of carbon and hydrogen. The most readily apparent role of atomic hydrogen
is to etch selectively nondiamond-bonded material, thereby reducing the in-
corporation of graphite in the completed film. Oxygen is frequently added,
either directly or as a part of the carbon carriers (CO or CO2 ). The presence
of oxygen can have a marked influence on the deposition processes, as can the
temperature of the CVD gas phase and substrate. An increasing CH4 content
induces a preferential orientation of diamond crystals [72]. For the highest
values of CH4 content, a (100) preferential orientation and then a textured
film can be obtained. The preferential orientation of diamond crystals can be
checked by the X-ray diffraction (XRD) technique [73].
Several techniques are employed to produce the plasma, such as hot-fila-
ment CVD, the DC glow discharge or microwave plasma-enhanced CVD [32].
The latter technique is presently the preferred one, because the plasma occurs
in a region far from any parts of the reactor, so that the contamination of
the film is very limited. Natural diamond and CVD diamond have the same
crystal lattice. However, the different production processes introduce differ-
ent impurity concentrations and different structural defect densities. These
differences may likely result in different material properties, e.g., mechanical
strength, thermal conductivity and electrical conductivity. Although natural
diamond is single crystal, it is far from perfect, containing high densities of
impurities and structural defects. As CVD diamond growth technology has
matured, the quality of the polycrystalline diamonds has improved: some of
the material properties of new CVD diamond films have already exceeded
those of natural diamond. Synthetic single diamonds have been available
since the 1950s after the ASEA Laboratories and the General Electric Com-
pany developed independently HPHT diamond synthesis. This technique is
rather simple in principle: a small volume filled with hydrocarbons is heated
at more than 2000◦C and the volume reduced by increasing the pressure to
Gbar values. In order to avoid the formation of graphite – which is ener-
getically preferred – some catalyst metals, such as nickel, cobalt or iron, are
included in the gas mixture. HPHT diamonds are usually yellowish because of
a high nitrogen concentration (up to 100 p.p.m.). Recently, a few companies
have been able to provide some small stones of pure diamond (IIa type). The
CVD technique is used also to synthesise homoepitaxial diamond films. The
parameters of the synthesis are nearly the same as those of heteroepitaxial
growth, but the substrate is a natural or HPHT diamond gem. This is at

the same time the advantage and the disadvantage of such a technique. Ho-
moepitaxial growth produces single crystals, so the overall crystalline quality
is better than in polycrystalline films, because of the absence of stresses inside
crystals, twinning crystals, defects and grain boundaries. Unfortunately, the
size of natural or synthetic gems is limited in size and therefore the maximum
size of homoepitaxial films is currently around 1 cm2 . Single-crystal diamond
films are grown on (100) HPHT diamond substrates. The crystalline qual-
ity of this material is comparable with that of natural diamond, but unless
films are grown under very carefully controlled deposition conditions [74, 75],
structural imperfections can form on the growth side [76, 77] acting as pref-
erential site for impurity and nondiamond phase incorporation. Recently, ho-
moepitaxial diamond has aroused a lot of interest owing to advances in the
synthesis of larger samples and improved electrical properties. It was used to
fabricate electronic devices [78, 79], and recently the growth of device-grade
homoepitaxial diamond films has been claimed [77].
5 Experimental Methods
Using UV photons to study the photoelectrical properties of diamond films
is a consolidated technique providing information on both the material prop-
erties and the device performance. Material properties can be investigated at
sub-band gap photon energies because nondiamond structures forming intra-
gap energy levels absorb such photons. At higher energies, the diamond film
absorbs photons, and therefore radiation can be used to analyze the photo-
electrical characteristics [80, 81, 82, 83]. In addition, it is generally accepted
that the photoconductive properties of diamond films change going deeper in
the material from the surface to the bulk and to the nucleation side.
Photoconductivity, time and spectral response are the main electro-
optical properties investigated to characterise the performance of diamond
UV photoconductors. Researchers routinely make use of different experimen-
tal equipments to measure the spectral response in the visible and in the UV
range. Optical tests in the deep UV range (at wavelengths below 180 nm) re-
quire vacuum systems, as well as vacuum-grade materials and sources, whilst
measurements at longer wavelengths can be performed in air. In addition, di-
amond is transparent at wavelengths above 225 nm, except for contributions
from impurities and defects in the crystals. As a result, special instrumenta-
tion, based on heterodyne detection techniques, is required to observe very
low level signals.
The apparatus used for the detector characterisation in the vacuum UV
range is based on a deuterium lamp (for λ > 120 nm) or a gas discharge source
(for λ < 200 nm) placed at the entrance slit of a vacuum monochromator,
which is evacuated down to a pressure p ∼ 10−6 Torr. The radiation sources
are sealed or must be interfaced directly to the vacuum. Photons entering
the monochromator impinge on a grating that can rotate around its axis
perpendicular to the optical path, so that only the selected wavelength can
illuminate the exit slit, where it is possible to arrange the detector. If the
spot size on the detector is smaller than the electrode overall dimensions, the
response is independent from the sensitive area. Generally, the estimated light
power is on the order of tens of nW/nm on the detector. By applying a bias
voltage, the response of such detectors to such a monochromatic illumination
is a photocurrent that can be measured using an electrometer.
The detector characterisation in the visible is performed through a dif-
ferent instrumentation. The monochromator is not evacuated and the photo-
current is measured after chopped illumination. The light source is typically
a tungsten lamp irradiating continuously from infrared to visible. Another
source is the mercury–xenon lamp, emitting spectral lines superimposed to
a continuous spectrum that covers the range from infrared to UV (about
200 nm). A chopper inserted between the source and the entrance slit of
the monochromator produces pulsed monochromatic radiation at the exit
slit. A lock-in amplifier can detect the alternating current induced by the
chopped illumination, while rejecting any DC signal, such as the dark cur-
rent. This experimental technique allows very high sensitivity, so that sig-
nals over 7–8 intensity decades and signal-to-noise ratio (SNR) < 1 can be
recorded [5, 80, 84].
Table 2. Attenuation lengths in IIa-type diamond films [85]
Wavelength Attenuation length, L0

nm µm
140 0.013
150 0.015
160 0.02
170 0.03
180 0.3
190 1.3
200 1.7
210 3
220 9
222.5 100
224 400
225 500
227 900
230 2 000
250 10 000
High-quality diamond layers present a sharp absorption edge at the band

gap energy. Correspondingly, the penetration depth, defined as the reciprocal
of the absorption coefficient, of photons shortens very rapidly from centime-
tres down to a few nanometres, as reported in the Table 2 [85]. This feature
provides a unique opportunity to use UV photons in a small spectral range
in order to probe the photoelectric and the charge transport properties of
diamond layers at different depths up to the surface proximity. In particu-
lar, transients of photocurrent induced by UV photons emitted by a lamp
and short pulses from an UV laser represent the optimal probe as rising and
falling photocurrent edges and time response provide information on trapping
or polarization effects.
Pulsed radiation was used from UV laser sources. The system is based on
excimer lasers emitting radiation at 125 nm, 157 nm and 193 nm, while one
can get pulses at 266 nm and 213 nm using the fourth and fifth harmonic of
the fundamental wavelength of a Nd:YAG laser. The pulses from these lasers
have duration around 10−9 s (4–10 ns) at a repetition rate of tens of Hz with
energy variable from 0.1 µJ to hundreds of µJ. It is possible to measure this
energy using pyroelectric probes with uncertainty within 5%. Photocurrent
pulses are measured using an oscilloscope directly connected to the device
through a bias-T.
6 Electro-Optical Properties
The research on diamond UV detectors shows an increasing trend of their
electro-optical performance over the last two decades. Scientific and tech-
nological efforts have provided a positive outcome with encouraging results,
allowing expectations for an imminent availability of a new generation of
high performance UV detectors. The technology required to produce CVD
diamond and photoconductor devices has been strongly improved. Finally,
the production of high-precision coplanar electric contacts with spacing d,
ranging between 10 and 100 microns, is now consolidated, even if their re-
producibility and reliability is still a concern. Hundreds of single-pixel pho-
toconductors have been developed and their electro-optical performance in-
vestigated thoroughly in the UV and visible range. The results assess the
excellent performance of the intrinsic diamond-based detectors.
In order to specify the performance of photodetectors, quantum efficiency
and dark current have to be determined along with time response for a certain
applied electric field E. The quantum efficiency, i.e., the average number of
carriers excited in the conduction band by an incident photon, is related
to the detector sensitivity, while the dark current is related to its thermal
noise and metal–diamond interface. Therefore, these quantities contribute
significantly to the SNR of the detector, a parameter that actually estimates
its detectivity.
The dark current of diamond detectors is generally measured at room
temperature in air or in vacuum, because, considering the wide band gap and
the very high resistivity, one may expect very low levels of leakage current
in diamond, even in the case of ohmic contacts. Low-level dark current at
room temperature is a very important topic, since one of the main targets of
researchers is to exploit this diamond electrical property in order to improve
the SNR while avoiding cooling systems. The dark current behaviour as a
function of the bias voltage also gives information on the characteristics of
the metal–diamond junction. The I–V curves of devices tested in air show
nonlinear and higher values than those of the same devices measured in an
evacuated environment: this is due to air and water vapour conductivity. This
is pointed out by comparing I–V curves of coplanar contacts and sandwich
geometry. Therefore, reliable measurements must be carried out in vacuum
or using guard rings, the latter only in case of transverse contacts.
The quantum efficiency gives information on the device spectral responsiv-
ity, highlights the presence of gain in photoconductors and allows the evalua-
tion of visible blindness. In addition, photoinduced charge yield in the visible
and infrared range is used to identify intragap energy levels due to impurity
species inside diamond films or defects [86].
The spectral response is determined by measuring the photocurrent in
a detector under illumination and by comparing it to that in a calibrated
photodiode. Therefore, the external quantum efficiency η can be estimated
as:
Id
η = ηp , (1)
Ip
where ηp is the external quantum efficiency of the calibrated photodiode at

a certain wavelength, Ip and Id are the photocurrents measured respectively
by the photodiode and the diamond detector, taking care that the spot light
is smaller than the contact size. Due to the presence of traps, the collected
charge Qc is different from the photogenerated charge Qph :

E
Qc = Qph µτ , (2)
d
and, correspondingly

Qc E
Id = = Iph µτ . (3)
t d
Thus, the external quantum efficiency can be expressed as

ηp E
η = EQE = Iph µτ = η0 G. (4)
Ip d
The external quantum efficiency (EQE) of diamond detectors is the prod-

uct of two terms: the first, η0 = ηp (Iph /Ip ), is the real quantum efficiency,
i.e., the number of photogenerated carriers per incident photon, while the
second, G = µτ E/d, is the gain factor, which will be discussed in depth at
the end of Sect. 7. Detectors based on high-quality diamond films currently
have quantum efficiency measured values greater than 1, and other specific
measurements are required to separate the contribution of the two terms.
A relevant and now well-established topic is that diamond photoconduc-
tors exhibit high UV sensitivity and photoconductive gain coupled to very
low dark current. For instance, it has been observed that good detectors yield
gain G > 200 at 200 nm and very low dark currents (< 10 pA/mm2 at 100 V)
along with very strong visible rejection (107 ) [42, 87, 88]. As a clear conse-
quence, the minimum detectable signal is very low. By using these data we
can estimate the noise-equivalent-power (NEP), i.e., a signal (measured in
watts) giving SNR = 1. NEP may have values around 10−11 W cm−2 .
Diamond detectors can be based on different substrate materials: natural
diamond, CVD polycrystalline diamond wafers, CVD or HPHT single-crystal
diamonds. Natural diamond was used in early experiments, when the synthe-
sis of diamond layers was at earlier stages. Nowadays, gems are used just
as reference because of the superior quality of synthetic layers, so detectors
based on poly- and monocrystalline layers will be discussed.
6.1 Polycrystalline Diamond Detectors
A discussion on the performance of a detector starts usually from its sensi-

tivity, even compared with the dark current levels, in order to highlight its
detection capability, especially in presence of low-level photon fluxes. The
main characteristic of diamond detectors is their high efficiency at UV wave-
lengths against visible light. Diamond technology is not yet mature for mass
production of UV detectors, but it has made large steps ahead during the
last decade. This means that their capability of detecting UV signals with a
photoconductive gain G > 1 while rejecting visible signals has been greatly
improved along with response time and removal of some detrimental effects
affecting even recent UV detectors. In addition, dark current levels can be
smaller than 1 pA/mm2 with an applied field of 1 V/µm.
Figure 1 shows a typical spectral response of current polycrystalline di-
amond UV detectors. These detectors exhibit a peak of efficiency around
200 nm, with gain ranging between 100 and 300 as reported by several au-
thors. The spectral range has been extended to the visible and infrared regions
to show the very high rejection factor, typically 107 [5, 87, 89]. The sharp sen-
sitivity cutoff at 225 nm is one of the main important features, which makes
diamond the solid-state detector with the best solar-blindness. After the re-
sults of Jackman’s team in 1996 [5] that claims improved UV selectivity and
global detector performance using a special gas treatment, it was generally
believed that surface treatments were needed in order to enhance the de-
tector response. More recently, several results show that similar or superior
performances can be obtained without any surface treatments. Other authors
2
10
External quantum efficiency (e/γ)
1
10
0 E = 28 KV/cm
10
-1
10
-2
10
-3
10
-4
10
-5
10
-6
10
-7
10
-8
10
200 400 600 800 1000
Wavelength (nm)
Fig. 1. Spectral response of a diamond photoconductive detector in a wavelength
range extended from the vacuum UV to the near infrared. It is worth noting the
exceptional visible rejection ratio that is about 107 at 400 nm and > 108 at 700 nm,
while the infrared discrimination factor is > 109
have extended the investigated spectral range to shorter wavelengths down

to 120 nm in order to analyse the vacuum UV photoresponse [53, 60, 64, 87].
In principle, a constant efficiency should be expected because photons are
completely absorbed, even in very thin diamond layers. The curve in Fig. 1
shows that the efficiency decreases, and the measurements from other au-
thors, even if stopping at 180 nm, confirm this trend [4, 5, 84]. A possible
explanation is a thin layer of contaminants or a high density of surface states
at the diamond surface that generates a “dead layer” [90]. This means that
taking care of the surface and the material quality this effect should be lim-
ited. Another hypothesis moves from the size of diamond crystals: a small
ratio between the volume of the crystal and the surface of the grain bound-
ary could affect the photoresponse. An indication comes from the response of
different substrates: polycrystalline films with small grains, as that in Fig. 1,
the same with greater crystals, and single crystals. The latter, as reported
below, shows much smaller efficiency reduction at short wavelengths.
We may expect that the material quality and the device processing tech-
nology influence the detector photoresponse, so it is important to identify the
critical parameters and their contribution. The relevant facts are:
(a) Some researchers have demonstrated that different deposition techniques
and parameters lead to different material quality and morphology and
that these correlate with the UV detector performance. The photolumi-
nescence (PL) and Raman peaks provide information on the material
quality; so their correlation with the detector efficiency can be expected
and indeed demonstrated by some groups (Fig. 2) [89, 91, 92, 93, 94]. The
30
25
ηav (%)
20
15
10
0.30 0.35 0.40 0.45 0.50
IPL/Idiam (adim)
Fig. 2. Average detection efficiency ηav as a function of the IPL /Idiam ratio [91].
The parameter IPL /Idiam is the ratio between the signal of the PL peak and of the
diamond Raman peak. The clear correlation between such indicators suggests that
the worsening in the detection performance is due to the higher incorporation of
defects and impurities occurring at higher CH4 concentrations
efficiency is also directly related to the average orientation of the crys-

tals in the polycrystalline films, as shown in Fig. 3 [95, 96]. Finally, the
spectral response curves in Fig. 4 confirm that the sensitivity and the dis-
crimination factor depend on the material quality that is determined by
different synthesis techniques [84]. Therefore, techniques producing high-
quality randomly oriented large crystals provide the best material for UV
detectors, maximising the ratio between the volume and the surface, i.e.,
the grain boundaries of the crystal.
(b) Several authors have discussed the time response of UV detectors, show-
ing that this is generally affected by a slow, sometimes very slow, rising
time and then by a persistence of the photocurrent (PPC) after hav-
ing turned off the radiation source (Fig. 5) [3, 8, 97]. In principle, di-
amond detectors should be very fast, owing to the very high mobility
of the photogenerated carriers. However, a variable but large concentra-
tion of impurities and defects, producing energy states inside the large
diamond band gap, causes a variation of the time response related to
their different trapping and release time. The gain–bandwidth product,
which is typical of electronic devices, suggests that UV diamond detec-
tors can exhibit improved time response at the expense of gain, and this
is confirmed by experimental results [54, 66]. By carefully tailoring the
material and the electrode processing, sensitive and fast detectors can
be achieved [60, 87, 98]. Long-time response effect is not present when
pulsed light is used to excite electrons. Very fast responses have been
transverse
co-planar
Photocurrent (nA)
10 E = 2.5x104 V/cm
0.6 0.7 0.8

P(ikl)
Fig. 3. Measured photocurrent as a function of the preferential orientation of
crystals in the diamond films (parameter P (ikl); see [34] for the details). This
plot shows the sharp decrease of the photoresponse when the CH4 concentration
increases. The correlation has been observed in the case of planar contacts as well
as of sandwich contacts, as expected
1,0
0,8
Normalised response
0,6
0,4
(iv)
(iii)
0,2
(i) (ii)
0,0
200 240 280 320 360 400
Wavelength (nm)
Fig. 4. Normalized photoresponse of some diamond detectors: a microwave plasma

CVD-grown sample (i), a hot-filament CVD-grown film (ii) and a microwave plasma
CVD-grown sample with high nitrogen content (iii). The curve (iv) is relative to
sample (ii) after a long UV exposure and this shows the effects of the persistence
of the photocurrent (data from [32])
120 3,5
@225 nm 3,0
100
2,5
Photocurrent (pA)
Photocurrent (pA)
80 2,0
1,5
60
1,0
40
0,5
20 0,0
-0,5
0 1000 2000 3000 4000 5000 6000 0 100 200 300 400 500 600 700 800 900 1000 1100
Time (s) Time (s)
Fig. 5. Transients of photocurrent for CVD polycrystalline diamond. The left panel
shows a transient affected by an ever-increasing photocurrent and the PPC after
UV irradiation at 225 nm. The right panel demonstrate the effects of the PPC on
subsequent pulses
obtained by several authors using diamond photoconductors to detect

UV excimer laser pulses [56, 82, 99, 100]. Detectors were also fabricated
exhibiting response times of less than 70 ps FWHM at 125 nm [64].
(c) The performance of different detectors may vary, even if the material
and the electrode processing are the same [101]. Two pairs of electrodes
on the same substrate may have different characteristics, thus providing
irreproducible responses. This is likely due to the difficulty of producing
identical metal–diamond interfaces and thus identical ohmic contacts.
The PPC is an important topic affecting performance and full exploitation
of UV detectors based on polycrystalline diamond. Therefore, it is important
to evaluate its effects on the photoresponse to nonpulsed radiation. The PPC
is an extra current observed after the radiation source has been turned off.
Many hours can be required to return to the dark current level measured be-
fore the irradiation. By switching on the radiation source after some seconds
or minutes, the photocurrent reaches quickly the values it had before turning
off the radiation and, it begins to grow. The overall effect is an ever-increasing
photocurrent, as can be seen in Fig. 5, still observable in many UV detectors.
Figure 6 shows another consequence of the PPC. It can affect the spectral re-
sponse when the wavelength range is scanned, producing an upward-scanned
spectral response very different from a downward-scanned one [102]. Irradi-
ating the sample for a long time with intense UV fluxes (UV “priming”) and
then scanning the spectrum, the highest quantum efficiency values have been
measured, reproducing the behaviour of the spectral response affected by the
PPC. Stabilizing the response by priming techniques is usually adopted for
X-ray and particle diamond detectors. Priming is not effective for UV detec-
tors, but UV priming can saturate all the traps and centres that affect UV
Fig. 6. The influence of the PPC on the spectral response can be observed using two
different scanning directions and after UV priming (solid line). The radiation source
has a very intense peak at 160 nm and is very stable (1% maximum variation). The
wavelength scanning after this peak is affected by the PPC. The photon fluxes at
the other wavelengths are so low that apparently they do not modify the spectrum
photoconduction, as indicated by these results. However, this priming is very

unstable: it disappears in a few hours and is strongly reduced after a few
minutes.
Very recently, Brescia et al. have reported on the photoresponse of de-
tectors fabricated on polycrystalline diamond that is comparable to that of
single-crystal detectors. The group at IAF, Frieburg, has synthesised 500 µm
thick diamond films. A layer 400 µm thick at the substrate side has been
removed and the growth surface has been polished, so as to obtain a fi-
nal thickness of approximately 50 µm. High-quality cylindrical large crystals,
much larger than the interelectrode distance then form the front surface. This
and the removal of the contaminated back layer are eventually the explana-
tion for such enhanced photoresponse. Sandwich structures have been also
studied in order to pave the way to imaging detectors. Indeed, this electronic
structure allows compact and easily addressable bidimensional pixel struc-
tures. However, the nonhomogeneous illumination of the substrate, due to
the short penetration depth of photons at wavelengths shorter than 225 nm,
gives rise to polarisation effects and to a nonlinear quantum efficiency that
saturates at higher photon fluxes. The quenching of the quantum efficiency is
apparently enhanced by increasing the beam light intensity. De Sio et al. have
also shown that the photoresponse quenching depends on the photon wave-
length [102]. All the photocurrent transients show a quasirectangular shape at
wavelengths longer than 220–230 nm, while this shape changes significantly
-1 CVD1
10 + 0.2 V/mm
- 0.2 V/mm
-2
10
EQE
-3
10
-4
10
CVD2
+ 2 V/mm
-1
10 - 2 V/mm
-2
EQE
10
-3
10
0
10 CVD2
+ 0.6 V/mm
-1
10
EQE
-2
10
140 160 180 200 220 240 260

Wavelength (nm)
Fig. 7. Absolute spectral response of two sandwich detectors fabricated on CVD
polycrystalline films having different thicknesses (500 µm for CVD1 and 40 µm for
CVD2). The lower graph shows the EQE curve obtained for the CVD2 sample in
coplanar geometry
at shorter wavelengths, exhibiting a photocurrent quenching after an initial

peak. This effect is not present in coplanar contacts. A possible explanation
lies on the nonuniform illumination of the material layer that produces polar-
isation. Photons with λ > 225 nm pass through diamond layers because the
absorption length is longer than the film thickness (see Table 2). In this case,
photons are absorbed in the whole layer. Conversely, photons are absorbed
in thinner and thinner material layers when λ < 225 nm. Collins et al. [98]
argued that, for energies above about 5.15 eV, most of the photons are ab-
sorbed before reaching the back contact, while all the carriers are generated
virtually within a few µm of the surface for energies above 5.6 eV. When the
thickness of the specimen is greater than the charge collection distance, the
photocurrent is made up of an induced current, associated with the move-
ment of the carriers in the electric field [103], together with the flow of current
from those carriers that actually reach the electrodes. For short wavelengths,
no carriers reach the back contact on the specimen, and the photocurrent is
dominated by the induced current.
An impact of this effect on the spectral responsivity may be expected.
Figure 7 shows the results for two detectors in sandwich configuration having
different thicknesses of the diamond layer. The spectral response is rather sim-
ilar and highlights a quenched efficiency at wavelengths shorter than 220 nm
or 230 nm. The different peak wavelengths can be explained taking into ac-
count the different thickness: the wavelengths between 220 nm and 230 nm
still pass through the thinner sample, thus reducing the collection efficiency,
but producing a uniform illumination of the layer. The spectral response data
suggest a possible dependence on the photon beam intensity. The radiation
source emits a line peak at 160 nm that is one order of magnitude more
intense than the emission at the other wavelengths. The EQE plots show
a minimum exactly at that wavelength with a shape resembling the source
emission line; this lets us suppose that higher intensity causes lower efficiency.
In order to analyze these aspects in more detail, De Sio et al. have il-
luminated the same samples with UV lasers, emitting nanosecond pulses at
193 nm, 213 nm and 266 nm. They have analyzed the peak voltage against the
beam intensity and observed that it saturates for photon fluxes above 1013
photons/s. This saturation, observed also by other authors [82, 104], is ac-
companied by an increasing width, but the total collected charge, calculated
as
T
1
Q= V (t) dt ,
Rosc 0
where Rosc is the oscilloscope input impedance, still shows saturation at

higher fluxes. The EQE reported in Fig. 8 confirms this charge saturation:
it should be constant if plotted against intensity, whereas it decreases and
saturates at higher fluxes. Lansley et al. have developed specific surface treat-
ments to provide linearity and to eliminate peak and width saturation, but
at the expense of sensitivity [104].
6.2 Single-Crystal Diamond Detectors
Brescia et al. [42] have reported on the first results of electro-optical char-
acterisation of single-pixel UV detectors based on CVD single-crystal dia-
mond. Photocurrent measurements were carried out illuminating the sample
Fig. 8. External quantum efficiency versus laser beam intensity at two different
wavelengths
at 160 nm. Devices on homoepitaxial diamond show currents more than one
order of magnitude higher than that of a standard polycrystalline diamond.
In addition, the ratio between photocurrent and dark current is about 3–4
orders of magnitude in the measured voltage range.
To analyse the sensitivity of such a device as a function of wavelength,
photocurrent measurements have been carried out under calibrated illumi-
nation transients. Figure 9 shows these transients, highlighting the good re-
sponse time of single-crystal detectors and the stability of the photocurrent.
Therefore, no PPC apparently affects the detector response. These data have
been used to estimate the EQE of the device in Fig. 10. Maximum EQE is
much greater than unity, due to the high photoconductive gain. However,
the device exhibits a nonoptimal solar blindness, due to the increasing ab-
sorption length (which is 4 µm, namely the CVD layer thickness, at around
215 nm) with decreasing photon energy. In such case, the authors have not
removed the Ib-type diamond substrate, and then its nitrogen content ab-
sorbs photons at longer wavelengths, affecting the photoconductive process.
No evidence of PPC effects has been observed by scanning these spectra from
short wavelengths to longer and vice versa.
The same authors have investigated if the nitrogen content may inhibit
the performance of UV detectors. Two planar interdigitated contacts were
deposited on one of the two surfaces of an HPHT Ib-type diamond crystal and
the spectral response of such device was measured. The very high quantum
efficiency obtained with this photoconductor is reported in Fig. 11. The same
Fig. 9. Response time of a single-crystal diamond UV detector. The device has

been irradiated at 160 nm at two different applied electric fields
Fig. 10. External quantum efficiency of a UV photoconductor based on a homoepi-

taxial diamond film biased at two different electrical fields. The efficiency peak at
210 nm is strongly enhanced, by a factor of 30, by a small increase of the electric
field (a factor of 5)
Fig. 11. External quantum efficiency for the same diamond detector, based on a
HPHT Ib-type diamond film, with a couple of electric contacts (upper curve) that
were removed before depositing another couple of contacts on the same surface
(lower curve)
figure also shows the much smaller spectral photoresponse obtained from the
same diamond film after removing the electric contacts, cleaning the surface
from any kind of residual and having deposited a new pair of contacts.
The perspectives of using single crystals for UV detectors are very promis-
ing. High-purity large crystals (side > 7 mm) have been already achieved [41,
105] with unprecedented optical and electronic performance [38, 44], so fur-
ther enhancements of UV performance can be expected. De Sio et al. tested
a photoconductive device based on a 500 µm thick freestanding homoepitax-
ial single-crystal CVD diamond. Photoconductive measurements in coplanar
and transverse configurations have been performed to characterize the device
sensitivity in the 140–250 nm spectral range. Very high sensitivity values
were achieved in both configurations. It has emerged that the sensitivity in
the transverse configuration is at least 300 times higher than in the coplanar
configuration. They observed high sensitivity in steady-state photoconductive
measurements carried out in a relatively thick diamond ultraviolet detector
in a transverse configuration [106].
7 Single-Pixel Detectors
7.1 Photodetectors
Typically, a diamond-based photon detector is in the form of a two-terminal

metal–insulator–metal (MIM) device. The insulator is undoped diamond. The
metal contacts can be ohmic or rectifying. The structure of the electric con-
tacts depends on the application and specifically on the energy of the photons
emitted by the excitation source used to create the excess carriers.
An electronic structure consisting of two metal electrodes separated by a
high-resistivity diamond layer is called a sandwich structure. It is generally
applied to high-energy X-ray or gamma-ray detection, since diamond films
are quasitransparent at those energies. An external voltage that is applied to
the metal contacts produces an electric field across the device. The direction
of the incident photons is nearly parallel to the electric field, and the electron–
hole pair generation occurs in the bulk of the material.
A coplanar structure, i.e., two electric contacts, usually interdigitated,
lying on the same surface, is used when the charge excitation follows the UV
photon or low-energy X-ray absorption. In this case, the direction of incident
photons is perpendicular to the applied electric field. The penetration depth
in diamond is small, and the electron–hole pair generation occurs near the
exposed surface (generally the growth surface).
The sandwich structure provides information on the bulk of the material,
while coplanar contacts give information on the electrical properties near
the surface. Therefore, using both measurements, the relationship between
bulk and surface properties can be investigated, along with the relationship
between the transport properties of charges drifting perpendicularly or along
the surface.
As already stated, UV photons are absorbed in a very thin layer close to
the illuminated surface, so coplanar device structures are generally preferred
to fabricate UV detectors. However, sandwich structures have been studied
recently [60], especially for single-crystal substrates in order to improve their
performance under UV illumination because they can be integrated more
easily in a pixel array detector. Large-array formats require many electri-
cal connections to bias and read-out each pixel; actually, it appears rather
impossible to accomplish them using coplanar contacts.
7.2 Photoconductor
A diamond-based photoconductor is operated by connecting it in series to

a DC-voltage power supply and to an electrometer. Photons are incident on
and absorbed in the very first film layers, thereby exciting electrons into the
conduction band. The result of such an excitation is a decrease of the ma-
terial resistance and hence an increase of the current flowing through the
electrometer. Therefore, current variations are proportional to photon flux
variations; this current excess is called photocurrent. The amount of photo-

current depends also on the applied bias voltage, on the photon wavelength
and on the distance between electrodes, although at high fields the current
saturates. The electric contact on diamond surface must be ohmic. It is diffi-
cult to fabricate good ohmic contacts on diamond due to its inert and wide
band gap nature [30]. It is necessary to somehow modify the structure of a
perfect diamond surface so that the voltage drop across the metal–diamond
interface is negligibly small as compared to that across the active portion
of the device. By selecting or intentionally creating damaged surfaces, good
contacts can be accomplished. However, producing reliable and reproducible
ohmic contacts is to date a major concern. Presently, the techniques used
to get ohmic contacts are ion implantation of boron atoms in the diamond
regions under the metal contacts or the use of carbide-forming metals, such
as titanium or chromium. Although the potential barrier height does not
change, ion implantation produces a narrower voltage-drop region, and pho-
togenerated charges can move from diamond to the contact by tunnelling
through the barrier. On the other hand, the ohmic behaviour of some specific
metals is likely due to the formation of carbide at the metal–diamond inter-
face. These metals are evaporated on diamond and then annealed at high
temperature (typically 400◦ C) to enhance carbide formation. A thicker layer
of gold coats the metallic contact in order to avoid oxidation; a buffer layer,
generally platinum, is used sometimes in between to avoid interdiffusion of
gold into the metal.
Some authors use gold contacts on diamond because it can be easily evap-
orated thermally. Sputtering can be another technique, but priming effects,
i.e., the saturation of intragap energy levels and then the enhancement of the
diamond photoconductive properties, due to electrons emitted from the elec-
tron gun have been observed. Gold forms Schottky barriers and its adhesion
on diamond surface, especially on smooth surfaces, is very poor. Annealing
at 400◦ C produces gold ohmic contacts, but their properties are not repro-
ducible.
The extremely high resistivity of intrinsic diamond eliminates the need
for reverse-biased junctions or doped material in order to suppress thermally
generated currents. In addition, the large band gap produces negligible dark
currents and then very low level photocurrents can be detected.
Charge generation and collection in photoconductors can be described by
considering a uniform device with mobile charge carriers, electrons and holes,
generated by incident radiation [33, 107]. If a flux F0 (t) of photons at a given
wavelength irradiates a photoconductive detector, the carrier generation rate
is G(t) = η0 F0 (t), with η0 the above-defined quantum efficiency. The number
of charges, N , is related to the generation rate according to the equation:
dN N
= G(t) .
dt τ
For a pulsed irradiation, the carrier density at the end of the excitation (t = 0)
is N = N0 , and its subsequent time evolution is

t
N (t) = N0 exp − ,
τ
while being
N = Gτ = η0 F0 τ ,
in the steady state, when the generation rate equals the recombination rate.
If an external electric field E is applied, the excited carriers move covering
an average drift distance,
L = vd τ = µEτ , (5)
which is determined by their drift velocity vd and lifetime τ .

Since electrons and holes have different effective masses (m∗e and m∗h ) and
exist in different energy bands, the values of electron mobility (µe ) and hole
mobility (µh ) may be different and, in addition, the electron (τe ) and hole
lifetimes (τh ) may not be equal, (5) should be written as L = (µe τe + µh τh )E.
Owing to the electric field, each of these carriers drifts with mean velocity
vd = µ(E)E that gives rise to a photocurrent in the external circuit, i.e.,
i = qvd /d (Ramo–Shockley theorem [103]). The total current is the product
of i by the number of carriers N

qF0 η0 vd τ τ L
I = iN = = qF0 η0 = qF0 η0 ,
d τd d
where τd = d/vd is the drift time for carriers crossing the distance d. The
factor (τ /τd ) or (L/d) represents the fraction of the distance that an excited
carrier drifts before recombining. If L/d > 1, the photogenerated charge will
be entirely collected, and L/d = µτ E/d can be considered as a photocurrent
gain factor G. On the other hand, if L/d < 1, only a fraction of the photogen-
erated charge will be detected. In this last case, an accurate detector design
must prevent the effects of space charge buildup: the uncollected charge can
accumulate in the device forming a net charge, which distorts the electric
field locally and leads to reduced detection efficiency.
7.3 Photodiode
An ideal diode junction allows current to flow easily (with low contact re-
sistance) if forward biased and very little current to flow if reverse biased.
P-type material is currently achieved by adding boron atoms to the diamond
lattice, while n-type diamond is still an open issue for many research groups.
As a result, p–n diamond diodes are not yet fully available and junctions are
currently limited to Schottky barriers.
In most applications, Schottky interfaces are reverse biased, with the volt-
age value fixing the barrier height and the width of the depletion region. Since
CVD diamond layers are available, many authors have published their results
on Schottky junctions using natural, polycrystalline or single-crystal boron-
doped diamond. The target was to produce Schottky diodes with small reverse
leakage currents operating also at high temperature (> 400 ◦ C) [28,49]. How-
ever, diodes manufactured on p-type CVD diamond suffered from relatively
large leakage currents. Adding a thin defect-free insulating layer between the
metal contact and diamond (MIS diode structure) reduces the leakage cur-
rent. Intrinsic diamond is an insulator, so MIS structures can easily achieved
by depositing an intrinsic layer on top of the p-type substrate. A thin SiO2
layer (∼ 2 nm) has also been used successfully to reduce the leakage cur-
rent (metal–oxide–semiconductor structure, MOS), but I–V characteristics
of MOS devices degrade with increasing temperature [108, 109]. The MIS
structure is more advantageous than the Schottky barrier because charge
carriers can be driven to an accumulation region. On the other hand, MIS
structures have a higher density of interface states, producing a larger current
noise.
The electrical properties of Schottky contacts on diamond, such as bar-
rier heights, exhibit small dependence on the properties of different metals,
such as their work functions or electronegativity. Treatments of the diamond
surface along with selecting proper metals, metal deposition parameters and
subsequent treatments, e.g., annealing, strongly determine the I–V charac-
teristics of the contacts. Typically, Au or Al contacts – but also other metals
such as W and Mo – on doped or undoped diamond are used.
It is possible to fabricate diamond detectors by using two electric con-
tacts, blocking or ohmic, separated by diamond. For example, a sandwich
configuration with a semitransparent blocking front contact and an ohmic
back contact gives Schottky-type photodiodes. A diamond layer deposited
on a p-type silicon substrate forms an isotype heterojunction with a highly
recombinant interface, which is essentially an ohmic contact [8]. Two copla-
nar blocking contacts are equivalent to two back-to-back Schottky diodes and
therefore described by a symmetrical relationship:

qV
Id = I0 tanh ,
ξKB T
where ξ is the diode quality factor and KB the Boltzmann constant. A similar
equation holds when these structures are illuminated, substituting (I0 + Iph )
for I0 , where Iph is the current photogenerated in the junction. So, the device
does not exhibit any photoelectric signal at zero bias. Such behaviour is rather
different from a single Schottky photodiode, which exhibits a small nonzero
short circuit current under illumination.
Photocurrent in diamond photodiodes depends on the applied external
field, but also on the device geometry and on the direction of the incident
light with respect to the applied field. For a diamond film having thickness T
and sandwich contacts, the light propagates along the electric field direction,
and the minority carrier photocurrent can be written [110]:

1 − eBT
I = qF0 (1 − R) α ,
B
where α is the absorption coefficient, R is the front surface reflectivity and
1
B =α+ ,
L
while the photocurrent from coplanar contacts, where light and field direc-
tions are approximately perpendicular, is:

d L
I = qF0 η 1 − exp − .
L d
It is noteworthy that no internal gain is expected for photodiodes: the quan-
tum efficiency can never exceed unity.
8 Pixel Array Detectors

One of the most important targets of developing detectors is eventually to
produce arrays of pixels for imaging applications. Some projects presented
by European research groups outline the interest in the field of diamond
pixel detectors. For instance, the RD42 Collaboration at CERN has started
in 2001 a program to develop two-dimensional particle detectors based on
CVD polycrystalline diamond [111], following research and investigation on
microstrip particle trackers (from 2000) [112, 113]. At the end of 2002, a
European Consortium, called BOLD, including Universities, Astrophysical
Observatories and Research Centers [114], has started a program supported
by the European Space Agency (ESA) to develop image sensors for the XUV
range. Other groups are developing small two-dimensional detectors (5 × 5
pixels) in order to measure profiles of UV laser beams [115] or synchrotron
radiation beams [116].
The major concern limiting the development of diamond detector arrays is
the highly inhomogeneous spatial response [98]. Small pixel detectors for UV
laser beam profiling have been developed by fabricating an array of squared
large – 4 mm each – coplanar contacts. This device has achieved interesting
results, even if cross-calibration is required to take into account spatial non-
uniformity. Electrical connections to the off-chip readout are still an issue,
because they are obtained through wires – one for each pixel and one common
– passing in front of the sensitive area. Moreover, these wire-bonding connec-
tions can be a practical solution for small arrays, but they are not scalable to
large array formats. Some research groups [53, 81, 117] are studying different
Fig. 12. Sketch of a CMOS imager (the lower substrate) coupled to a diamond
photosensitive layer (the upper layer ) using indium bumps. The very thin layer on
top of the sensitive layer is an open electrode
viable solutions for UV imagers. The most promising solution is a hybrid de-
tector, where a layer of sensitive material is used to absorb UV photons and a
CMOS device includes the readout electronics, as sketched in Fig. 12. An ar-
ray of square contacts is deposited on one of the two surfaces of the sensitive
material and electrically connected to a similar pixel array of readout circuits
by means of indium bumps, a technique named flip chipping. The pixel array
and the on-chip readout electronics are implemented with CMOS technology.
The reports from the RD42 group on prototypes of CMOS flip-chipped de-
tectors – pixel or microstrip detectors on diamond layers – developed in the
field of particle detection shows encouraging results. Since the early 1990s,
an explosion of activity in the area of CMOS image sensors has taken place.
Several important factors have contributed to the emergence of CMOS image
sensors at this time rather than 10 to 20 years ago. The primary factor con-
sists of recent demand for low-power, miniaturized digital imaging systems.
A second important factor is that CMOS currently offers submicron feature
sizes as well as low concentration of defects and contaminants. Moreover,
CMOS image sensors have an intrinsically panchromatic response to visible
and near-infrared photons [118,119]. One of the main drawbacks of this imag-
ing technology is the extension of such a performance in different portions
of the spectrum. An opportunity currently investigated is introducing proper
solutions and materials compatible with the CMOS process in order to ap-
ply this well-known technology in the field of UV imagers. Progress in the
field of diamond synthesis and processing for electronic devices along with
the maturity and reliability of CMOS technology leads to the development
of a complete system-on-a-chip (SOAC) UV imager. An analog signal elec-
tronic chain is often integrated on a simple chip to provide noise reduction
and removal of artefacts such as fixed-pattern noise. One of the key steps in
realizing a digital imager on a chip is developing an on-chip ADC suitable
for integration with the image sensor.
Fig. 13. The absolute quantum efficiency of diamond detectors compared with
CCD and MCP detectors
Advantages of integrating CMOS technology with diamond layers using

the flip-chipping technique are high visible rejection ratio, high efficiency in
the UV portion of the wavelength spectrum, very low noise, ultrafast frame
rate, low power consumption and high level of radiation hardness, with stable
response on a wide temporal range. In particular, radiation hardness can be
achieved even if a silicon-based technology is adopted. CMOS devices are
in fact intrinsically radiation hard because of the technology of the gate
oxide. These devices have been tested in several severe conditions and results
highlight their radiation harness to particle doses up to 10 MRad (see, for
example, [120]).
9 Applications
The comparison between the performance of diamond UV detectors and CCD

detectors [121] or MCP detectors [122], generally used to detect photons in
this spectral region, is reported in Fig. 13 and highlights the high sensitivity
of diamond detectors, especially when they are based on a single-crystal film.
Therefore, diamond detectors appear to be suitable for UV applications. The
measurements show the highest sensitivity along with a very low dark current
and the best visible rejection among the commonly used detectors (included
GaN and AlN detectors). In addition, they have very low power consump-
tion, owing to the photocurrent values obtained at low bias voltage. Table 3
Table 3. Possible applications of diamond-based UV detectors
Environment Pollution, ozone, flame, and gas monitoring

Solar UV monitoring
Photochemistry
Toxic gas detection (ammonia, hydrocarbons)
Toxic or flammable-solvent monitoring in air-ducts in industries
Toxic chemical storage sites and hazardous waste disposal areas
Germicidal water disinfection
Medical Biotechnology safety systems

HPL Chromatography
UV curing applications
Analytical and chemical instruments (disinfections that kill bacteria viruses and some cysts)
Protein crystallography
Skin cancer caused by UV radiation
Observation of fluorescence
Spectroradiometry
Dermatological diagnosis
Treatment and dental polymerisation
Societal Quantification of UV exposure to personnel

Monitoring UV radiation in workplace activity such as welding
Technological Flame monitoring in aircraft engine

Burners monitoring in gas turbines
UV spectroscopy of gas on electric discharges
Mass spectroscopy laboratory analysis
UV spectroscopy of plasma in reactors
Excimer laser applications
Industrial processes Industrial process control systems

Detection of gasses in high temperature environments such as furnaces or reactors
Flame monitoring in commercial hydrocarbon-fuelled heating systems and in furnaces
Semiconductor processing
Turbine air intake and exhaust
Control of chemical processes involving gaseous compositiary
UV analysis of material degradation and ageing
Oil industry Monitoring of flammable liquids or gases during cleaning and maintenance work on empty tanks
Oil and gas platform (offshore and onshore installations) monitoring
Petrochemical and chemical industries and refineries monitoring
Transportation of dangerous materials
Fuel and gas storage loading and distribution
Aromatic hydrocarbons (Benzene, Toluene, Xylene etc.)
Desulphurisation process in refineries, oil platforms, pipelines, refuelling stations
Military Missile threat warning

UV search and track
summarises the possible applications of diamond UV detectors. Three spe-

cific applications that would benefit from the availability of such detectors
are hereafter described in some detail.
9.1 UV Lasers
A particularly promising application of diamond UV detectors is monitor-

ing the operation of high-power UV laser systems. Excimer lasers are the
most intense sources of highly monochromatic UV radiation, and they are
currently employed in a wide range of applications, including laser PCB
drilling, high-performance micromachining, next-generation photolithogra-
phy tools and the preparation of biological materials [99, 100, 123, 124, 125].
Each laser emits intense UV light at a specific wavelength in the range 157–
248 nm and beam monitoring means controlling beam uniformity, shape and
radiation dose. Laser tubes are continuously operated for about one month,
i.e., ∼ 108 pulses, at kHz frequencies and fluences of a few mJ/cm2 [126].
A typical requirement is for reliable operation of monitoring detectors during
the laser lifetime, so that they can be replaced at the same time. CCD de-
tectors are currently used, providing accurate profiling with high resolution.
However, they suffer from some drawbacks such as low UV sensitivity, higher
sensitivity to visible light, and poor radiation hardness and operational life-
time when monitoring 50% degradation after as few as 105 pulses has been
reported [126].
The properties of diamond make it the best alternative. CVD diamond
UV photoconductors based on coplanar interdigitated electrode structures
have been developed and tested by a number of groups in the field [56, 64,
127, 128, 129, 130]. In particular, Jackman et al. have shown that they can
follow the 15 ns pulses of the 193 nm excimer laser providing sensitivity,
speed and stability matching the requirements. The first detector structures
fabricated on as-grown material showed a dependence of the dark current and
photocurrent curves on previous illumination with UV or visible light [97].
Subsequently, Jackman’s group has shown that successive treatments can
progressively modify the sensitivity and response times of photoconductors
fabricated on polycrystalline diamond. Therefore, a combination of material
treatments and careful design can lead to devices overwhelming the previous
limits, but with reduced sensitivity [4, 5, 131]. Suitably designed and treated
CVD diamond photoconductive structures exhibit excellent either long-term
performance or damage levels, with relatively modest degradation of diamond
devices exposed to beam fluences of 1−2 mJ/cm2 for ∼ 108 pulses. It is im-
portant to stress that currently no other detector can meet all the necessary
requirements, apart from CVD diamond devices. Other groups have investi-
gated the use of highly oriented diamond (HOD) for UV detectors [62, 132].
The basic idea is that, in order to limit the detrimental effects of the grain
boundaries on the recombination time of the photoexcited carriers [133, 134],
HOD films could be used as a valid alternative to the single crystals. In-
deed, the reduction of the misorientation and the grain boundary angle are
expected to reduce grain boundary effects, to increase the free-carrier mobil-
ity resulting in an increase of the detection efficiency. They fabricated UV
detectors using HOD films and measured the response of the sensors after
pulsed irradiation of an ArF excimer laser or a dye laser. Kobashi et al. have
found that detected pulses exhibit two peaks for case of the ArF laser and
three peaks after dye laser irradiation. They infer that the peaks following
the first are due to a carrier emission from trap states at grain boundaries.
Achard et al. claim an HOD film exhibiting a significantly higher photore-
sponse and a discrimination factor of 103 between above and below band gap
photon energies. They compare these results to those obtained using detec-
tor fabricated on randomly oriented polycrystalline diamond films, and they
reports superior in the performance of HOD films. This is in accordance with
the results in [64], but not with previous results from other groups, showing
that detectors based on HOD films do not perform as well as those based
on films with randomly oriented crystals when illuminated with continuous
irradiation [93, 94, 95].
9.2 Photolithography
Moore’s Law states that the gate density in silicon integrated circuits quadru-
ples every 3 years. The photolithography technique forms the device struc-
tures by projecting the desired circuit image (mask) onto the silicon wafer.
The ultimate feature size achievable, and hence the gate density, is limited
by the wavelength of the radiation used to illuminate the mask.
The target is to achieve ever-smaller feature sizes, so excimer lasers have
ruled out mercury UV lamps because they emit intense pulsed beams at
shorter wavelengths, from the 248 nm radiation available from KrF filled ex-
cimer lasers to the recently introduced 193 nm radiation from ArF lasers.
The next step is to operate at 157 nm to realise device structures of 100 nm
and below [135]. A very bright source for this wavelength is the molecular
fluorine laser [136]. This development is accompanied by the need for a suit-
able solid-state photodetector sensitive to this wavelength with a lifetime
fitting long continuous operation (107 –108 laser pulses). There is currently
no commercial silicon detector available for monitoring at 157 nm, because
their sensitivity falls down around 10%, due to the combination of the short
penetration depth of photons in silicon and its spontaneously passivated sur-
face. In addition, it is difficult to stabilise the silicon detector photoresponse,
because they are cooled to reduce the dark current and layers of condensed
contaminants formed on the sensitive area cause a decreasing efficiency.
This opens opportunities for a completely new electronic material and
hence to diamond. Owing to its extreme properties, CVD diamond is an
ideal candidate for this application. Photodetectors based on this material
should be both radiation-hard and visible-blind whilst being sensitive to the
UV wavelengths [137]. Gaudin et al. have shown that photoconductors suit-

able for laser applications can be fabricated from polycrystalline diamond
by careful tailoring of device design and surface treatments [131]. Recently,
again Jackman’s group assessed for the first time coplanar interdigitated gold
photoconductor structures on free-standing thin film diamond by exposing
them to 157 nm pulsed radiation up to the fluence value of 1.4 mJ cm2 . They
demonstrated that the detectors were capable of following the 15–20 ns laser
pulses. Device gain was found to be linear with applied bias (±30 V) with
a sensitivity of 10 V/mJ cm2 over the linear portion. Unfortunately, the long-
term lifetime tests were limited to 104 –105 pulses, but no obvious degradation
in characteristics was observed.
9.3 Space Astronomy
The UV spectral region is of particular interest for astrophysics. UV allows

the study of a rich interval of emission and absorption lines due to species
and states of ionization of atoms and molecules useful for the diagnostics of
plasmas and gases. UV observations are rather recent because atmospheric
absorption forces observations from space. This has caused major technolog-
ical problems for instrumentation developments. Presently, we are living in
an exciting transition time for UV astrophysics. Planned missions have been
launched and most are even finished. In the meantime, new missions are still
under definition. These missions are mainly solar missions (Solar Orbiter, So-
lar B, SDO), but also missions towards galactic and extragalactic regions, like
WSO/UV. That situation will allow some years for the development of new
technologies. The urgent demand for new technologies, which will push cur-
rent observation limits, has induced main space agencies (ESA and NASA) to
form European and US research groups with the aim to conceive innovative
optical and spectroscopic instruments and in particular to develop new-gen-
eration detectors. The detector is a crucial component of every astronomical
experiment since it plays a critical role in determining the overall perfor-
mance, even if it represents only a small fraction of the instrument total cost.
Therefore, detectors have been and are still very important for a remarkable
technological investment. Although UV imaging detectors currently available
represent an improvement over the last decades, they exhibit some limita-
tions due to silicon technology. These limitations become crucial in space
missions where reliability, durability, low cost, sensitivity, radiation-hardness
and solar-blindness are required. Presently, photon-counting detectors repre-
sent the only viable option, since they are radiation-hard and, by using proper
selection of the photocathode, solar-blind. Initially the quantum efficiency
in the UV of a typical photon-counting detector is about 30%, although it
decreases progressively until values around 10%, due to photocathode degra-
dation. However, a photon-counting detection system requires high voltage
and an accurate and stable coupling to the intensifier with the signal readout
system (typically CCD or anode array), which increase the satellite payload,
the complexity and the probability of failure. Another problem is related to

the sources to be observed. Many UV astronomical sources are weak and
associated with very bright visible emissions. It follows that the use of pho-
ton-counting detectors coupled with optical systems able to remove visible
signals (monochromators and light traps) is necessary. When the UV flux is
more than 102 photons/(s · cm2 ) – as from the Sun – photon-counting detec-
tors may run into signal pile-up problems and integrating detectors may be
damaged by the radiations.
A strong effort in technological research aimed at the development of UV
solid-state image sensors is therefore necessary. Diamond exhibits unparal-
leled properties with respect to other wide band gap materials. The high
crystalline quality achieved with present synthetic techniques and an activa-
tion energy of 5.5 eV makes diamond detectors intrinsically solar-blind and
possessing a very low dark current. Moreover, high carrier mobility and radia-
tion hardness makes diamond appealing for UV detector development. A few
sensors based on single-crystal diamond exhibited photoconductive gains up
to G = 700 at 200 nm with rise and fall times shorter than one second and
without evidence of persistent photocurrents [42].
The application of diamond detectors in astrophysics will bring:
(a) technological advantages over the current detectors, such as lower power
consumption, longer detector lifetime, lower electric insulation and more
compact detectors without coolers, radiation shields, optical filters.
(b) performance advantages, such as long-time stability, owing to corrosion
and radiation resistance, intrinsic solar-blindness, low dark-current levels,
and high quantum efficiency.
The comparison between these results and the performance of a CCD detec-
tor [121] and an MCP detector [122], generally used in space UV experiments,
is reported in Fig. 13 and it highlights the high sensitivity of diamond detec-
tors, especially when they are based on single crystals.
NEP values of 10−11 –10−12 W cm−2 coupled to response times on the
order of 1 s, even if not exceptional for silicon detectors, are already suitable
for astronomical observations. Diamond can compete with silicon detectors
owing to the unmatched visible rejection among the commonly used detectors
(included GaN and AlN detectors) and its performance and technological
advantages [138].
Diamond UV imagers are, we hope, the next step. Pixel arrays on diamond
substrates represent a strong technological effort involving material synthesis
and electronic processing. The advent of hybrid CMOS detectors offers the
advantages of combining layers that are very sensitive to UV with highly
integrated and radiation-hard devices, also allowing strong mass reduction,
owing to a very compact SOAC that requires only a 5 V power supply [53,
117].
10 Conclusions
Diamond-based photodetector properties have been reviewed in this article.
The electro-optical properties of the material and the electronic structures
fabricated on it have been discussed. This material is still a promising can-
didate for UV photon detection at room temperature, offering superior effi-
ciency in the range 100–300 nm and unsurpassed UV/visible discrimination.
These features pose diamond as the best UV detector currently available,
but some problems limit its full exploitation, and work is in progress to solve
them. The quality of the material is the first crucial step to maximise the
efficiency of the detectors; the results of optical characterisation and their cor-
relation with the detector performance clearly demonstrate this statement.
Considerable improvements have been achieved. Reproducible high-quality
CVD diamond can now be produced even in research laboratories, but still
polycrystalline diamond is not competitive, more than ever in view of fabri-
cating pixel array detectors.
Single-pixel detectors could be used for space applications because they
fit the typical specifications of astronomical observations. Unfortunately, this
is too small a niche for attracting any industrial developments, and single-
pixel detectors do not meet the interest of the astronomers, as the imagers
do. On the other hand, UV diamond detectors appear to be suitable for fast
detection of laser pulses, in particular for VUV lithography, where silicon
devices show poor performance.
Interest may be renewed by recent progress in the synthesis of optical-
grade single crystals. Sizes of up to 10 mm are currently available with su-
perlative electrical performance. The results of the first experiments in the
UV region confirm the expectation, so it is possible to assume that in the
near future some efforts will be focused on the development of UV detectors
based on large single crystals.
Some groups have started programs to develop pixel arrays on diamond.
The inhomogeneous spatial response observed in the polycrystalline films is
the actual barrier, and it is probably difficult to find solutions to overwhelm
it, owing to the intrinsic defective nature of this form of diamond. Single-
crystal diamond can be the solution, even if the dimensions are still rather
small. However, the device technology on diamond and the reproducibility
of material processing procedures and electronic structures are still an issue
and a number of problems remain unsolved. Therefore, replicating electric
contacts on the diamond surface does not assure identical metal/diamond
interfaces, and hence this contributes to the nonuniformity of the detector
response.
Notwithstanding present technological limitations to the full development
and exploitation of UV detectors, diamond remains the most promising can-
didate to replace silicon detectors in many applications, in particular those
requiring its unique and extreme properties.
References
[1] L. S. Pan, D. R. Kania, P. Pianetta, O. L. Landen: Appl. Phys. Lett. 57, 623
(1990) 464
[2] D. R. Kania, M. I. Landstrass, M. A. Plano, L. S. Pan, S. Han: Diam. Relat.
Mater. 2, 1012 (1993) 464, 466
[3] S. C. Binari, M. Marckywka, D. A. Koolbeck, H. B. Dietrich, D. Moses: Diam.
Relat. Mat. 2, 1020 (1993) 464, 466, 468, 476
[4] R. D. McKeag, S. S. M. Chan, R. B. Jackman: Appl. Phys. Lett. 67, 2117
(1995) 464, 466, 475, 493
[5] M. D. Whitfield, R. D. McKeag, L. Y. S. Pang, S. S. M. Chan, R. B. Jackman:
Diam. Relat. Mater. 5, 829 (1996) 464, 471, 474, 475, 493
[6] R. D. McKeag, R. D. Marshall, B. Baral, S. S. M. Chan, R. B. Jackman:
Diam. Relat. Mat. 6, 374 (1997) 464, 466
[7] S. Salvatori, R. Vincenzoni, M. C. Rossi, F. Galluzzi, F. Pinzari, E. Cappelli,
P. Ascarelli: Diam. Relat. Mat. 6, 775 (1996) 464
[8] S. Salvatori, E. Pace, M. C. Rossi, F. Galluzzi: Diam. Relat. Mater. 6, 361
(1997) 464, 476, 488
[9] E. Pace, F. Galluzzi, M. C. Rossi, S. Salvatori, M. Marinelli, P. Paroli: Nucl.
Instr. Meth. A 387, 255 (1997) 464
[10] F. Galluzzi, M. Marinelli, L. Masotti, E. Milani, E. Pace, M. C. Rossi, S. Sal-
vatori, M. Santoro, S. Sciortino, G. V. Rinati: Nucl. Instrum. Meth. A 409,
423 (1998) 464
[11] F. Foulon, P. Bergonzo, C. Jany, A. Gicquel, T. Pochet: Nucl. Instrum. Meth.
A 380, 42 (1996) 464
[12] F. Foulon, P. Bergonzo, C. Jany, A. Gicquel, T. Pochet: Diam. Relat. Mater.
5, 732 (1996) 464
[13] Centronics Ltd and Centronic House: King Henry’s Drive New Addington,
Surrey, UK 464
[14] G. Davis: Diamond (Adam Hilger, Bristol 1984) 464, 465
[15] C. P. Klages: ”The physics of diamond”, Proc. of Int. School of Physics E.
Fermi, Course CXXXV (IOS, Amsterdam 1997) p. 355 464
[16] A. T. Collins: ”The physics of diamond”, Proc. of Int. School of Physics E.
Fermi, Course CXXXV (IOS, Amsterdam 1997) p. 273 464, 466, 468
[17] R. J. Neemanich, P. K. Bauman, A. T. Sowers, B. L. Ward: Proc. of 9th
CIMITEC-WORLD Ceramic Congress and Forum on New Materials (Flo-
rence, Italy 1999) p. 217 464
[18] R. Sporl, R. Heidinger, C. Wild, R. Schwab, P. Koidl: Proc. of 9th CIMTEC-
WORLD Ceramic Congress and Forum on New Materials (Florence, Italy
1999) p. 325 464
[19] C. S. J. Pickels, J. R. Brandon, S. E. Coe, R. S. Sussmann: Proc. of 9th
CIMITEC-World Ceramic Congress and Forum on New Materials (Florence,
Italy 1999) p. 435 464
[20] H. Nakahata, et al.: Japan J. Appl. Phys. 33, 324 (1994) 464
[21] C. Jang, F. Foulon, P. Bergonzo, A. Tardieu: Proc. of the V Int. Symp. on
Diamond Materials, vol. 97–32 (The Electrochem. Soc. Inc., Pennington NJ,
USA 1998) p. 208 465
[22] M. C. Rossi, S. Salvatori, F. Galluzzi: Proc. of the V Int. Symp. on Diamond

Materials, vol. 97–32 (The Electrochem. Soc. Inc., Pennington NJ, USA 1998)
p. 255 465
[23] C. E. Nebel, M. Stutzman, F. Lacher, P. Koidl, R. Zachai: Diam. Relat.
Mater. 7, 556 (1998) 465
[24] O. Gauchin, et al.: Diam. Relat. Mater. 8, 886 (1999) 465
[25] M. Nesladek, K. Meykens, L. Stals, M. Vanecek, J. Rosa: Phys. Rev. B 54,
5552 (1996) 465
[26] M. Nesladek, L. Stals, A. Stesmans, K. Iakoubovskij, G. J. Adriaenssens,
J. Rosa, M. Vanecek: Appl. Phys. Lett. 72, 3306 (1998) 465
[27] H. Kawarada: ”The physics of diamond”, Proc. of Int. School of Physics E.
Fermi, Course CXXXV (IOS, Amsterdam 1997) p. 505 465
[28] T. Tachibana, J. Glass: Electronic device processing, in L. Pan, D. Kania
(Eds.): Diamond: Electronic Properties and Applications (Kluwer Academic,
Boston 1995) p. 319 465, 488
[29] G. Messina, S. Santangelo, A. Tucciarone: ”The physics of diamond”, Proc.
of Int. School of Physics E. Fermi, Course CXXXV (IOS, Amsterdam 1997)
p. 485 465
[30] M. Werner, et al.: Diam. Relat. Mater. 5, 723 (1996) 465, 486
[31] M. A. Plano: Electronic device processing, in L. Pan, D. Kania (Eds.): Di-
amond: Electronic Properties and Applications (Kluwer Academic, Boston
1995) p. 349 465
[32] L. S. Plano: Growth of cvd diamond for electronic applications, in L. Pan,
D. Kania (Eds.): Diamond: Electronic Properties and Applications (Kluwer
Academic, Boston 1995) p. 62 465, 469, 477
[33] D. R. Kania: ”The physics of diamond”, Proc. of Int. School of Physics E.
Fermi, Course CXXXV (IOS, Amsterdam 1997) p. 555 465, 468, 486
[34] K. Meguro, Y. Ando, et al.: Proc. of 15th Diamond Symposium, vol. 12,
NDFCT (2002) p. 125 465, 477
[35] D. Takeuchi, et al.: Phys. Stat. Sol. A 186, 269 (2001) 465
[36] M. Stammler, H. Eisenbeiss, et al.: Diam. Relat. Mater. 11, 504 (2002) 465
[37] T. Teraji, S. Mitani, et al.: Crystal Growth 235, 287 (2002) 465
[38] J. Isberg, A. J. Whitehead, S. E. Coe, G. A. Scarsbrook: Science 297, 1670
(2002) 465, 484
[39] M. Kasu, N. Kobayashi: Diam. Relat. Mater. 12, 413 (2003) 465
[40] O. A. Williams, R. B. Jackman: Diam. Relat. Mater. 13, 557 (2004) 465
[41] J. Isberg, J. Hammersberg, H. Bernhoff, D. J. Twitchen, A. J. Whitehead:
Diam. Relat. Mater. 13, 872 (2004) 465, 484
[42] R. Brescia, A. D. Sio, E. Pace, et al.: Phys. Stat. Sol. A 199, 113 (2003) 465,
474, 481, 496
[43] T. Teraji, C. Wang, S. Endo, T. Ito: Proc. of the Third International Sym-
posium on Diamond Electronic Devices (2), vol. 11, NDFCT (2001) p. 313
465
[44] J. Isberg, A. J. Whitehead, et al.: Diam. Relat. Mater. 13, 320 (2004) 465,
484
[45] J. H. Kaneko, T. Tanaka, et al.: Nucl. Instrum. Meth. A 505, 187 (2003)
465, 466
[46] O. A. Williams, R. B. Jackman: Diam. Relat. Mater. 13, 325 (2004) 465
[47] E. Pace: Recent Res. Devel. Optical Eng. 2, 317 (1999) 466
ciarone, G. V. Rinati: Proc. of the V Int. Symp. on Diamond Materials, vol.
97–32 (The Electrochem. Soc. Inc., Pennington NJ, USA 1998) p. 556 466
[49] F. Prins: “The physics of diamond”, Proc. of Int. School of Physics E. Fermi,
Course CXXXV (IOS, Amsterdam 1997) p. 411 466, 488
[50] P. Gonon, S. Prawer, D. N. Jamieson, K. W. Nugent: Diam. Relat. Mater. 6,
673 (1997) 466
[51] C. Gu, et al.: Diam. Relat. Mater. 6, 673 (1997) 466
[52] E. Pace, et al.: Proc. SPIE 4498, 121 (2001) 466
[53] J. F. Hochedez, P. Lemaire, E. Pace, U. Schühle, E. Verwichte: ESA Pro-
ceedings “Solar Encounter: The First Solar Orbiter Workshop”, vol. SP-493
(2001) p. 245 466, 475, 489, 496
[54] R. D. McKeag, R. B. Jackman: Diam. Relat. Mater. 7, 513 (1998) 466, 468,
476
[55] P. Bergonzo, D. Tromson, C. Mer: Semicond. Sci. Technol. 18, S105 (2003)
466
[56] E. Lefeuvre, M.-C. Castex, J. Achard, A. Tardieu, C. Beuille, H. Schneider:
Diam. Relat. Mater. 12, 642 (2003) 466, 478, 493
[57] F. Spaziani, M. C. Rossi, S. Salvatori, G. Conte: Diam. Relat. Mater. 13, 814
(2004) 466
[58] E. Pace, A. D. Sio: Nucl. Instrum. Meth. A 514, 93 (2003) 466
[59] A. D. Sio, et al.: Diam. Relat. Mater. 12, 1819 (2003) 466
[60] R. Brescia, A. D. Sio, E. Pace, M. C. Castex: Diam. Relat. Mat. 13, 938
(2004) 466, 475, 476, 485
[61] M. D. Whitfield, S. P. Lansley, O. Gaudin, R. D. McKeag, N. Rizvi,
R. B. Jackman: Phys. Stat. Sol. A 181, 121 (2000) 466
[62] K. Hayashi, et al.: Diam. Relat. Mater. 10, 1794 (2001) 466, 493
R. B. Jackman: Phys. Stat. Sol. A 185, 99 (2001) 466
[64] F. Foulon, et al.: J. Appl. Phys. 84, 5331 (1998) 466, 475, 478, 493, 494
[65] P. Gonon, S. Prawer, Y. Boiko, D. N. Jamieson: Diam. Relat. Mater. 6, 860
(1997) 466
[66] S. Salvatori, M. C. Rossi, F. Galluzzi, D. Riedel, M. C. Castex: Diam. Relat.
Mater. 8, 871 (1999) 466, 476
[67] C. Y. Fong, B. M. Klein: Diamond: Electronic Properties and Applications
(Kluwer Academic, Boston, USA 1995) p. 1 468
[68] P. K. Bachmann, et al.: Diam. Relat. Mater. 1, 1 (1991) 468
[69] M. Marinelli, et al.: J. Appl. Phys. 76, 5702 (1994) 468
[70] M. Marinelli, et al.: J. Appl. Phys. 89, 1430 (2001) 469
[71] M. Marinelli, et al.: Phys. Rev. B 64, 195205 (2001) 469
[72] M. Marinelli, E. Milani, M. Montuori, A. Paoletti, P. Paroli, J. Thomas: Appl.
Phys. Lett. 65, 2839 (1994) 469
[73] S. Zhao: Characterization of the electrical properties of polycrystalline dia-
mond films, Phd thesis, The Ohio State University (1994) 469
[74] H. Watanabe, D. Takeuchi, S. Yamanaka, H. Okushi, K. Kajimura,
T. Sekiguchi: Diam. Relat. Mater. 8, 1272 (1999) 470
[75] D. Takeuchi, H. Watanabe, S. Yamanaka, H. Okushi: vol. 9 (2000) p. 231 470
[76] Y. K. Vohra, A. Israel, S. A. Catledge: Appl. Phys. Lett. 71, 321 (1997) 470
[77] H. Okushi, H. Watanabe, S. Ri, S. Yamanaka, D. Takeuchi: J. Cryst. Growth
237–239, 1269 470
[78] G. S. Gildenblat, S. A. Grot, A. Badzian: Proc. IEEE 79, 647 (1991) 470
[79] T. Ito, M. Nishimura, M. Yokoyama, M. Irie, C. Wang: Diam. Relat. Mater.
9, 1561 (2000) 470
[80] E. Pace, R. D. Benedetto, S. Scuderi: Diam. Relat. Mater. 9, 987 (2000) 470,
471
[81] J. F. Hochedez, et al.: Diam. Relat. Mater. 10, 673 (2001) 470, 489
[82] M. D. Whitfield, S. P. Lansely, O. Gaudin, R. D. McKeag, N. Rizvi,
R. B. Jackman: Diam. Relat. Mater. 10, 693 (2001) 470, 478, 481
[83] P. Bergonzo, A. Brambilla, D. Tromson, C. Mer, C. Hordequin, B. Guizard,
F. Foulon, V. A. Solé, C. Gauthier: Diam. Relat. Mater. 9, 960 (2000) 470
[84] S. Salvatori, M. C. Rossi, F. Galluzzi, E. Pace, P. Ascarelli, M. Marinelli:
Diam. Relat. Mater. 7, 811 (1998) 471, 475, 476
[85] D. Palik: Handbook of Optical Constants of Solids II (Academic, New York
1991) 471, 472
[86] S. M. Sze: Physics of Semiconductor Devices, 2nd ed. (Wiley, New York 1981)
473
[87] E. Pace, R. D. Benedetto, S. Scuderi: Diam. Relat. Mater. 9, 987 (2000) 474,
475, 476
[88] E. Pace, A. Vinattieri, A. Pini, F. Bogani, M. Santoro, Y. Sato: Phys. Stat.
Sol. A 181, 91 (2000) 474
[89] J. Alvarez, et al.: Diam. Relat. Mat. 13, 881 (2004) 474, 475
[90] J. F. Hochedez, et al.: Diam. Relat. Mater. 10, 673 (2001) 475
S. Santangelo, A. Tucciarone, G. V. Rinati: Diam. Relat. Mater. 10, 1788
(2001) 475, 476
[92] M. G. Donato, G. Faggio, M. Marinelli, G. Messina, E. Milani, S. Santangelo,
G. V. Rinati: Thin Solid Films: Applications, Preparation and Characterisa-
tion, Research Signpost (Kerala 2003) pp. 1–16 475
[93] G. Faggio, G. Messina, S. Santangelo, G. V. Rinati: Microsyst. Technol. 5,
151 (1999) 475, 494
[94] G. Faggio, M. Marinelli, G. Messina, E. Milani, A. Paoletti, S. Santangelo,
G. V. Rinati: Microsyst. Technol. 6, 23 (1999) 475, 494
ciarone, G. V. Rinati: Diam. Relat. Mater. 7, 1039 (1998) 476, 494
[96] C. Jany, F. Foulon, P. Bergonzo, A. Brambilla, F. Silva, A. Gicquel, T. Pochet:
Diam. Relat. Mat. 5, 741 (1996) 476
[97] P. Gonon, S. Prawer, Y. Boiko, D. N. Jamiesson: Diam. Relat. Mat. 6, 860
(1997) 476, 493
[98] J. Hiscock, A. T. Collins: Diam. Relat. Mater. 8, 1753 (1999) 476, 480, 489
R. B. Jackman: Diam. Relat. Mater. 10, 650 (2001) 478, 493
R. B. Jackman: Diam. Relat. Mater. 10, 715 (2001) 478, 493
[101] E. Pace, A. D. Sio: Nucl. Instrum. Meth. A 514, 93 (2003) 478
[102] A. D. Sio, et al.: Diam. Relat. Mater. 12, 1819 (2003) 478, 479
[103] S. Ramo: Proc. Inst. Radio Eng. 27, 584 (1939) 481, 487
[104] S. P. Lansley, et al.: Diam. Relat. Mater. 12, 677 (2003) 481
[105] J. Achard, R. Sussmann: Private communication (2004) 484
[106] A. D. Sio, J. Achard, A. Tallaire, R. S. Sussmann, A. T. Collins, F. Silva,
E. Pace: Appl. Phys. Lett. 86, 213504 (2005) 484
[107] A. Yariv: Optical Electronics (Saunders College Publishing, USA 1991) p. 399
486
[108] M. Marckywka, S. C. Binari, P. E. Pehrsson, D. Moses: SPIE Proc. 2282, 20
(1994) 488
[109] M. W. Geis, J. A. Gregory, B. B. Pate: IEEE Trans. Electr. Dev. 38 3, 619
(1991) 488
[110] S. Salvatori, E. Pace, M. C. Rossi, F. Galluzzi: Mat. Sci. Eng. B 46, 105
(1997) 489
[111] R. Stone, et al.: Nucl. Instrum. Meth. A 473, 136 (2001) 489
[112] W. Adam, et al.: Nucl. Instrum. Meth. A 453, 141 (2000) 489
[113] W. Adam, et al.: Nucl. Instrum. Meth. A 476, 706 (2002) 489
[114] J. F. Hochedez, et al.: Diam. Relat. Mater. 11, 427 (2002) 489
[115] J. Schein, et al.: Rev. Sci. Instrum. 73, 18 (2002) 489
[116] A. Schulze-Briese, et al.: Nucl. Instrum. Meth. A 468–468, 230 (2001) 489
[117] E. Pace, et al.: ESA Proc. “Solar Encounter: The First Solar Orbiter Work-
shop”, vol. SP-493 (2001) p. 311 489, 496
[118] W. Ogiers, D. Uwaerts, et al.: Second Round Table on Micro-Nano Technolo-
gies for Space (ESA/ESTEC, Noordwijk, the Netherlands 1997) 490
[119] B. Dierickx, G. Meynants, D. Scheffer: IEEE Workshop on Charge-Coupled
Devices & Advanced Image Sensors (June 1997) 490
[120] IMEC: Esa-ascmsa report p60280-ms-rp-001 also in http://www.imec.be/
fuga/publications/publ.html (March 1997) 491
[121] G. Naletto, E. Pace, G. Tondello, A. Boscolo: Meas. Sci. Techn. 5, 1491 (1994)
491, 496
[122] K. Wilhelm, et al.: Solar Physics 162, 189 (1995) 491, 496
[123] P. Apte, M. Gower: SPIE 2375, 349 (1996) 493
[124] P. T. Rumsby, E. C. Harvey, D. Thomas, N. H. Rizvi: SPIE 3184, 176 (1997)
493
[125] M. C. Gower, J. Cashmore, C. Solomon, et al.: 3rd international symposium
on 193 nm lithography 493
[126] N. H. Rizvi, J. Cashmore, C. Solomon, P. Rumsby, M. Gower: Proc. SPIE
3183, 30 (1997) 493
[127] H. Yoneda, K. I. Ueda, Y. Aikawa, K. Baba, N. Shohata: J. Appl. Phys. 83,
1730 (1998) 493
[128] C. E. Nebel, M. Stutzmann, F. Lacher, P. Koidl, R. Zachai: Diam. Relat.
Mater. 7, 556 (1998) 493
[129] S. Salvatori, M. C. Rossi, F. Galluzzi, D. Reidal, M. C. Castex: Diam. Relat.
Mater. 8, 871 (1999) 493
[130] M. D. Whitfield, S. P. Lansley, O. Gaudin, R. D. McKeag, N. H. Rizvi,
R. B. Jackman: Proc. of the VI Int. Symp. on Diamond Materials, vol. 99–32
(The Electrochem. Soc. Inc., Pennington NJ, USA 1999) p. 322 493
[131] O. Gaudin, S. Watson, S. P. Lansley, H. J. Looi, M. D. Whitfield, R. B. Jack-
man: Diam. Relat. Mater. 8, 886 (1999) 493, 495
[132] J. Achard, et al.: Diam. Relat. Mater. 11, 423 (2002) 493
[133] H. Yoneda, K. Ueda, Y. Aikawa, K. Baba, N. Shohata: J. Appl. Phys. 83,
1730 (1998) 493
[134] S. Han, R. S. Wagner: Appl. Phys. Lett. 68, 3016 (1996) 493
[135] R. Harbison: in Proc. of the First International Symposium on 157 nm Litho-
graphy, vol. 1 (Dana Point, California, USA 2000) p. 9 494
[136] S. M. Hooker, P. T. Landsberg: Progress in Quantum Electronics 18, 227
(1994) 494
[137] M. I. Landstrass, M. A. Plano, M. A. Moreno, S. McWilliams, L. S. Pan,
D. R. Kania, S. Han: Diam. Relat. Mater. 2, 1033 (1993) 495
[138] E. Pace, et al.: SPIE Proc. 4498, 121 (2001) 496
Index
blindness diamond growth, 465, 468, 469
solar blindness, 482 diamond properties, 467, 468, 470,
visible blindness, 466, 468 472
highly-oriented diamond, 493
charge lifetime, 487 homoepitaxial diamond, 470
charge mobility, 487 HPHT diamond, 469, 470, 474
chemical vapour deposition (CVD), IIa type diamond, 469
463, 464 natural diamond, 463, 466, 468–470,
coplanar contacts, 480, 489 474, 488
coplanar structure, 485 single-crystal diamond, 463–465, 474,
CVD diamond, 464, 468, 469, 472 484, 488
homoepitaxial CVD diamond, 464, diodes
465 MIS diode, 488
single-crystal diamond, 481
external quantum efficiency (EQE),
dark current, 472 473, 480–482
defects
grain boundaries, 470 glow discharge, 469
detectors
CCD detectors, 491, 493, 496 heteroepitaxial growth, 470
CMOS detectors, 490, 496 high-pressure high-temperature
detector gain, 474 (HPHT), 465, 469
diamond-based detectors, 474, 481 homoepitaxial growth, 465, 470
imaging detectors, 463 hot filament, 469
MCP detectors, 491, 496
MOS detectors, 488 image sensors, 489
pixel-array detectors, 489, 497 CMOS image sensors, 490
single-pixel detectors, 497 UV imagers, 490
time response, 476
UV detectors, 463–465, 472, 474, 476, metal-insulator-metal (MIM), 485
481, 484, 485 metal-oxide-semiconductor (MOS), 488
diamond, 463, 467 metal-semiconductor-metal (MSM),
diamond detector array, 489 485
metal/diamond interfaces, 478 polycrystalline diamond, 463–465, 469,

474, 488, 495, 497
penetration depth, 471, 479 priming, 478
photoconductive gain, 473, 482 UV priming, 478
photoconductivity, 470
photoconductors, 472, 485 quantum efficiency, 463, 472, 473, 486
UV photoconductors, 470, 493
photocurrent, 486 response time, 482
photocurrent persistence, 476, 478,
479, 482 sandwich contacts, 479, 489
photodetectors, 468, 472, 485, 494 sandwich structure, 485
diamond-based photodetectors, 497 Schottky barrier, 486–488
photodiodes, 487, 488 Schottky contacts, 488
photon counting, 495 Schottky diode, 488
photoresponse, 475, 479 spectral response, 481
vacuum UV photoresponse, 475 surface treatment, 495
deposition (PECVD), 469 x-ray diffraction (XRD), 469
Vibrational Spectroscopy in Ion-Irradiated
Carbon-Based Thin Films
Giuseppe Compagnini1 , Orazio Puglisi1 , Giuseppe A. Baratta2,

and Giovanni Strazzulla2
1
Dipartimento di Scienze Chimiche, Università di Catania, Viale A. Doria 6,
Catania I-95125, Italy
gcompagnini@unict.it
2
INAF-Osservatorio Astrofisico di Catania, Via S. Sofia 78, I-95123 Catania,
Italy
gianni@ct.astro.it
Abstract. In this work we present and discuss some selected experiments on ion-
irradiated carbon-based thin films. Vibrational spectroscopy is used to investigate
the materials structure and to explore the mechanisms of ion beam-induced mod-
ifications in many carbon solids such as crystalline carbon and carbon alloys, hy-
drocarbon molecules and exotic carbon species.
1 Introduction
For many years the vibrational properties of carbon-based amorphous thin
films have been considered crowded with information for two main reasons.
First of all, the position and relative intensities of the bands observed in
Raman scattering or infrared absorption measurements are indicative of the
bonding state of the carbon atoms and allow the study of the materials in
which heteroatomic bonds are included as constituents (carbon alloys, hy-
drogenated and nitrated systems). Second, the shape and width of the bands
are directly connected with structural relaxation phenomena or with the ex-
tension of the nanocrystalline domains that are typically found in the amor-
phous microscopic structure. If we consider additional benefits such as the
nondestructive nature of the techniques and their accessibility and low cost
for many laboratories, it is easy to understand why they are among the most
used for carbon materials. In the case of pure amorphous carbons, the Raman
technique has a dominant position [1]. This is because it is more suitable in
the detection and characterization of homonuclear C–C bonds, even though
the relaxation of the selection rules typical of the amorphous state frequently
gives the possibility to observe other kinds of vibrations. On the other hand,
infrared absorption is very useful in the detection of heteronuclear bonds.
This is the case of the so-called hydrogenated and nitrated carbon materials
or of carbon alloys such as a-Si1−x Cx systems.
Particularly interesting is the study of carbon materials subjected to ion
irradiation or other particle bombardment. General features of the ion bom-
bardment are the formation of metastable phases and the introduction of
506 Giuseppe Compagnini et al.
foreign species in a controlled thickness range between a few nanometers and

few microns, depending on the ion species and their energy. The first effect
is due to the unusual energy release that is typical of the ion collision cas-
cade, while the second can be used either for semiconductor doping or for
the formation of new thin films or buried layers. In the case of crystalline
carbon, many experiments on ion implantation in diamond have been per-
formed for the last three decades. One objective has been its doping [2, 3, 4].
Others focused on the transformation of diamond (sp3 hybridization) into
a graphite-like sp2 -bonded amorphous carbon [5, 6] in which the threshold
for the sp3 –sp2 conversion and the resulting amorphous carbon properties
were found to strongly depend on the implantation temperature. In any case,
depending on the nature, energy and flux of the particles, the properties of
the irradiated carbon samples can be changed in different ways. It has been
found that damage beyond the amorphization threshold in diamond leads,
upon thermal annealing, to relaxation of the diamond structure to graphite
for sub-MeV ion irradiation, while for MeV alpha particles irradiation an-
nealing restores the original diamond structure [7].
Other important fields in which keV–MeV irradiation of carbon-based
solids find applications are: polymer irradiation in ion beam lithography [8,9],
high-energy graphite particle irradiation studies for applications into nuclear
fusion materials [10], formation of biocompatible materials by ion beam sur-
face modifications of amorphous carbons [11] and the application of all the
irradiated materials’ features to the astrophysical field. In this last case the in-
terest is due to the supposed presence of carbon species in the interstellar and
circumstellar media as well as in many objects of the solar system [12, 13, 14].
Observed and/or predicted carbon-bearing solids (or large molecules) include
species with different hybridizations (sp1 , sp2 , sp3 ) such as amorphous car-
bons, polycyclic aromatic hydrocarbons, fullerenes, nanodiamonds, graphite
and linear carbon chains.
A complete overview of these fields is impossible; however, in the follow-
ing we report a series of selected experiments for large class of carbon-based
species from crystalline carbon forms and carbon alloys, to hydrocarbon mole-
cules and unusual carbon species. The aim is to explore the mechanisms of
ion irradiation-induced modifications in many carbon-based solids.
2 Irradiation of Crystalline Carbon and Carbon Alloys

One of the best ways to produce amorphous materials in the form of thin films
is to induce amorphization by using ion bombardment. This procedure has
been successfully used for several elemental solids and compounds [15, 16].
For the covalent group IV materials like silicon or germanium, ion beam
amorphization produces a so-called continuous random network of tetrahedral
sites [17]. In the case of carbon the situation is complicated by the different
bonding possibilities. Moreover, amorphous carbon (a-C) can be obtained by
Vibrational Spectroscopy in Ion-Irradiated Carbon-Based Thin Films 507
Fig. 1. I D /I G intensity ratio in ion-damaged HOPG as a function of the deposited

energy. The inset shows a typical Raman spectrum for a keV ion-produced amor-
phous carbon
ion irradiation starting from graphite, diamond or other crystalline forms.

Raman spectroscopy is able to characterize the irradiation effects starting
from early stages to the final amorphization. If one consider the irradiation
of a highly oriented pyrolitic graphite (HOPG) sample at low fluences, Raman
spectra show the formation of a feature at 1360 cm−1 , called the D line, which
is accompanied by the formerly present 1580 cm−1 peak (G line) [18, 19].
It is now clear that the G line is due to the bond stretching of all pairs
of sp2 atoms in both rings and chains. The D peak is attributable to the
breathing mode of some pieces of graphene sheets formed by ballistic collisions
during the irradiation [20, 21]. Once the ion fluence is increased to overcome
few eV/atom, a collapse to the amorphous phase is observed with a typical
Raman spectrum reported in the inset of Fig. 1. Again, D and G bands
are present even though the deconvolution of the entire signal is frequently
ambiguous. Figure 1 also shows the evolution of the I D /I G intensity ratio
for the irradiation of graphite in the early stages of the process. The data
have been reported for different irradiating ions and fluences (circles refer to
1015 ions/cm2 , squares to 1014 ions/cm2 ) at an ion energy of 100 keV [22].
These data are reported as function of the energy lost by the beam through
ballistic collisions, considered the fundamental parameter which determines
the vibrational and electronic properties of the final amorphous layer. Even
though ballistic collisions are considered random and are produced in a high-
density collision cascade (typical values for 400 keV He+ irradiation are 4 ×
10−14 eV/(atom cm−2 )), rearrangements following these processes produce
materials which are not purely sp2 hybridized.
Their electronic nature can be studied through electron energy loss spec-
troscopy experiments. In particular, it has been shown that amorphous car-
bons obtained by keV ion irradiation of diamond or graphite at room temper-
ature possesses nearly 80% sp2 bonding arrangement so that the vibrational
features can be mainly assigned to the trigonal carbon component. The final
result is also independent of the initial crystalline structure of the irradiated
carbon. A careful study of the D and G lines and their significance for the
amorphous carbon structure has been recently given by Ferrari et al. [21].
These authors pointed out that there is a strict correlation between the in-
tensity ratios and the extension of the graphite domains found in a-C and
nanocrystalline graphite. These considerations on the degree of order have
been successfully applied in the study of carbonaceous materials that have
been identified in the so-called interplanetary dust particles (IDPs). Indeed,
since fast-particle bombardment is one of the most important processes suf-
fered by materials (dust, asteroids, comets, etc.) present in the solar sys-
tem, the Raman spectra of a number of different ion-irradiated carbons have
been compared with those obtained for IDPs caught on airbone impacting
collectors at an altitude of 18–20 km and believed to be extraterrestrial in
origin [12]. The different degrees of order observed could be indicative of
different irradiation doses by solar wind particles suffered by IDPs in the
interplanetary medium before the collection in the Earth’s atmosphere.
Let us now consider what happens when elements other than carbon par-
ticipate in the structure of the final amorphous layer. It has been observed
that keV ion irradiation produces variable amounts of trigonal carbons when
different amounts of hydrogen are introduced in the sample structure. This
can be done either starting from a hydrogenated material (as in the case of the
amorphization of hydrocarbons, see next section) or by introducing hydrogen
by H+ irradiation. In particular, an increase of the hydrogen concentration
favors the formation of tetrahedral bonds. This increase of tetrahedral sites is
also responsible for an increase of the optical energy gap and of the resistiv-
ity, giving materials frequently termed as hydrogenated amorphous carbons.
The presence of other group IV elements also deeply influences the bonding
state. One of the most interesting examples related to this topic is the study
of binary amorphous carbon-based alloys like a-Si1−x Cx with different stoi-
chiometry. Again, vibrational spectroscopy is an excellent tool to monitor the
behaviour of the local order in covalently bonded solids through the analy-
sis of the properties of their fundamental units [23]. In particular, Raman
spectroscopy provides good detection and characterization of homonuclear
C–C and Si–Si vibrations through their intense signals located in the range
1000–2000 cm−1 and 300–500 cm−1 , respectively. Moreover, the relaxation
Fig. 2. Combined Raman and infrared data for crystalline and ion-irradiated stoi-
chiometric SiC
of the selection rules typical of the amorphous state gives the possibility to
also observe vibrations coming from heteronuclear bonds like Si–C (between
700 and 800 cm−1 ).
Figure 2 shows a collection of Raman and infrared data in crystalline
and amorphous SiC. Amorphization has been induced by Ar+ 300 keV ion
irradiation. It is straightforward to observe that the Raman spectrum of
the a-SiC shows all the three signals obtainable in a random mixture of
silicon and carbon atoms, that is, Si–Si (500 cm−1 ), Si–C (800 cm−1 ) and C–
C (1430 cm−1 ) vibrations. Heteronuclear bond vibrations are also observed
in the infrared absorption spectrum with a relevant increase in width and
decrease in intensity for the case of the amorphized layer.
Interestingly, the position and width of the C-C signal in a-Si1−x Cx al-
loys is quite different with respect to those commonly found in all the hydro-
genated and unhydrogenated amorphous carbons (see also the inset in Fig. 1),
since it is symmetrical and only 170 cm−1 wide [24]. The difference can be
ascribed to the chemical environment of the carbon subsystem, embedded
into a silicon matrix that favors the formation of sp3 -arranged carbon atoms,
considered predominant in these amorphous silicon carbon alloys. Similar re-
sults are obtained in a-GeC samples [25]. Analogous considerations can be
done for the Si–Si signal at around 480 cm−1 . In this respect it has been
recently reported that [26] the width of this band is almost twice that of ion-
implanted amorphous silicon. This can be explained with the differences in

bond angle distortion (∆θ). In the case of a-SiC samples this has been eval-
uated as ∆θ = 22.5◦ . Moreover, upon annealing ∆θ increases up to 27◦ at
700◦ C, while for amorphous silicon it decreases from 12.3◦ to 9.5◦ . Relaxation
of the a-SiC matrix provides the energy required to increase the distortion of
the silicon clusters.
Annealing experiments conducted at 1000◦C have also shown that the
disorder-induced 1430 cm−1 signal is reversible. This occurs in all the alloys
with x ≤ 0.5. The observation is consistent with the fact that sp2 graphite-
like rings (strong stability) are never formed in substoichiometric alloys and,
as a consequence of ion irradiation, all the carbon atoms are “used” to obtain
silicon carbide crystalline grains eventually embedded in a silicon excess. The
experimental observations regarding the behaviour of the C–C Raman signal
are reported in Fig. 3. Different alloys are obtained by implanting different
amounts of carbon atoms into a silicon layer, in such a way to have a uniform
thin film. In all of the as-prepared samples (Fig. 3a) the C–C signal is ob-
servable if the carbon concentration is above a bonding percolation threshold
(around x = 0.4). This signal disappears after 1000◦ C annealing for all the
substoichiometric samples, while for higher carbon concentrations, sp2 clus-
tering is obtained (well-resolved D and G components) with the precipitation
of pure carbon phases embedded into a SiC matrix [27].
Combination of Raman scattering and infrared absorption data has also
permitted [23] researchers to quantitatively determine the fraction of C–C
bonds as a function of the carbon concentration, revealing that there is an
onset in the formation of C–C bonds just at x = 0.4, consistent with the
qualitative behavior seen in Fig. 3. The increase of the signal intensities with
increasing carbon concentration is in agreement with the known theories of
random bonding in amorphous alloys.
3 Irradiation of Hydrocarbons (Oligomers, Polymers

and Frozen Gases)
The effect of ion irradiation on organic molecules is strongly related to the
amount of the deposited energy and to the nature of the species involved.
A first effect of the ionizing radiation is to change the molecular structure
through a series of nonconventional chemical reactions. Typical examples are
the effects of cross-linking [28] and scission [29] observed in several irradiated
polymer chains at moderate or low energy releases (below 0.1 eV/atom).
At these stages ion-induced enhanced chemical processes are also observed
such as polymer unzipping effects typical of PMMA [30, 31]. These processes,
which occur naturally at temperatures of about 360◦ C at which methacry-
late monomer is the most abundant product of the PMMA pyrolysis, are
observed at lower temperatures (120–200◦C) if the polymer is irradiated.
Other reported phenomena are the evolution of simple gaseous molecules like
Fig. 3. Raman data in the region of C–C signals for a-Si1−x Cx samples as a function
of the carbon content. As-deposited and 1000◦ C-annealed samples are shown
H2 , CH4 or more complex species [29]. A useful parameter to characterize ion

beam effects on polymers is the so-called chemical yield G (number of events
by 100 eV if ion deposited energy). This quantity depends on the type of
polymer and on the ion energy. For instance, alkane chains, like those which
constitute the polyethylene structure, release a large amount of H2 mole-
cules with a G = 2 H2 /100 eV accompanied with the formation of cross-
links (cl) with a G = 1.2 cl/100 eV. The analysis of these irradiated targets
by vibrational spectroscopy can be correlated to the change of some important
physico-chemical properties such as the solubility change or the wetting re-
sponse of the irradiated surface. As an example we cite a series of experiments
on the irradiation of assembled aliphatic molecules grafted to a silver metal
surface through S–Ag bonds [32, 33]. The monolayers (CH3 (CH2 )11 S–Ag) are
irradiated by 100 keV H+ ions at fluences in the range 1012 –1014 ions/cm2 and
are studied by surface-enhanced Raman (infrared) spectroscopy [34], which
ensures a cross-section magnification (105 –108) with respect to conventional
vibrational techniques.
Fig. 4. Correlation between wetting response and surface composition in ion-irra-

diated alkane self-assembled monolayers
The observation of the CH stretching regions both in the Raman spectra

and in the infrared ones, before and after the irradiation, is in agreement
with above-mentioned hydrogen loss:
– There is a decrease in the CH3 /CH2 stretching intensity ratio.
– There is the formation of olefinic groups with CH stretching signals above
3000 cm−1 .
Figure 4 reports the correlation between the wetting response of the sur-
face as obtained by measuring the water liquid contact angle, and the change
in the chemical structure induced by irradiation. Wetting data indicate that
ion irradiation changes the ambient interface from a layer of CH3 groups
to a more polar one, thus decreasing the contact angle. Then, hydrocarbon
molecules have lost a significant amount of the original aliphatic character to
become a rather carbonaceous material.
At higher irradiation fluences (higher deposited energies) the formation
of a carbonaceous layer (whose thickness depends on the irradiation para-
meters) is a dominating feature. This can be observed either by irradiating
large molecules, polymers or frozen gases. In the case of the irradiation of ices,
most of the studies have been centered on the so-called solid residue, obtained
once the irradiated sample has been annealed at room temperature. Indeed,
among the effects induced by fast ions, the formation of an organic residue
evolving at high doses towards an ion-produced hydrogenated amorphous
carbon (IPHAC) is particularly interesting. IPHAC formation has been ob-
served in several kinds of carbon-containing ices (C6 H6 , CH4 , C4 H10 , etc.) for
a combination of bombarding ions (H, He, Ar, etc.) and ion energies (ranging
between a few keV and MeV).
It has been shown, by in situ Raman spectroscopy, that 3 keV He+ irra-
diation of a frozen benzene film leads to the formation of an IPHAC even at
low temperature (T = 77 K) [35]. In that experiment the penetration depth

of ions (about 0.04 µm) was much less than the thickness of the benzene
film (about 2 µm thick). After warm up to room temperature, no feature
due to the benzene ice was present in the spectrum because the underlying
unirradiated ice has sublimated, passing through the porous structure of the
residue. Only the C=C 1560 cm−1 Raman band was observed. In situ IR
spectroscopy has also evidenced the presence of both sp1 (CH stretching in
monosubstituted acetylenes) and sp3 carbon hybridizations (CH2 aliphatic
stretching) in addition to the sp2 hybridization. Further irradiation of the
organic residues at room temperature (Fig. 5) determines an increase of the
I D /I G ratio with the dose up to a saturation value, while the H/C ratio
measured by elastic recoil detection and the optical gap decreases [35].
Figure 5 is divided into three regions delimited by broken vertical lines.
Up to a total dose of 10 eV/atom, the deposited frozen film remains a mole-
cular solid (no optical gap can be measured). Here the relative amount of H
atoms bound to sp2 carbon sites decreases strongly with respect to sp1 and
sp3 sites (see the two bottom lines). No amorphous carbon feature can be
clearly seen in the Raman spectrum. In the second region (10–25 eV/C atom),
the ratios of different hybridization remain constant, a strong H loss is ob-
served and the amorphous carbon Raman feature become observable: the
material has changed into a polymer-like compound (an optical gap can be
measured). The D line is almost absent in the Raman spectrum. In the third
region (≥ 25 eV/C atom) the relative number of H atoms bound to sp2 sites
decreases strongly with respect to sp1 and sp3 sites (note, however, that all
the C–H bands become weaker and weaker due to the H loss), the ratio of
the D to the G line increases: the material has changed in an amorphous
carbon film with a decreasing optical gap. These results were interpreted in
terms of an increasing average sp2 cluster size attributed to the preferential
hydrogen loss from sp2 sites with respect to sp3 sites (see [35] for further
details). A similar trend has been obtained on an aliphatic (C4 H10 ) ice at 10
K.
Recently, it has been shown that ion irradiation can drive the formation
of amorphous carbon also in a pure methane ice and in the case of a mixtures
containing several ices [36]. In particular, 30 keV He+ irradiation of an H2 O :
CH4 : N2 = 1 : 6 : 3 icy mixture at 12 K leads to the formation of a
hydrogenated amorphous carbon. Even in this case the Raman results are
consistent with an increasing size or number of the sp2 clusters. This supports
the idea that the process of IPHAC formation is general. Anyway, there seems
to be great difference in the efficiency of the process itself [36].
4 Irradiation of sp-Rich Amorphous Carbon Phases

Even though many papers have been published on the effects of ion irradiation
on diamond and diamond-like materials as well as on organic carbon-based
+
3 keV He on C6H6
2 ID
(Raman)
IG
1
0
0 20 40 300 600 900
H
1 (ERDA) 27 eV/C-atom
C
0 180 eV/C-atom
Ratio of peaks absorbance
Damaged polymer IPHAC

molecular like
2 solid
(IR: ice) 280 eV/C-atom
1
1 ≡ C⎯H sp
2 1000 eV/C-atom
C⎯H sp
G
0 D
(IR: ice) 1200 1400 1600 1800

-1
2 Ram an shift (cm )
3
CH2 sp
1
2
C⎯H sp
0
0 20 40
Dose (eV/C-atom)
Fig. 5. Overview of experimental results obtained with various techniques (Raman,
IR and elastic recoil detection) after ion irradiation of frozen benzene
systems and graphitic-like entities (included fullerenes, nanotubes and carbon

onions), there is still a lack in the literature regarding the irradiation of linear
carbon species. These materials are essentially constituted by sp-hybridized
carbon atoms to form linear carbon chains frequently termed as carbynes [37].
Several authors claim to be able to obtain a relevant percentage of carbynoid
species in their produced amorphous carbons, but these species easily react
or cross-link once exposed to air [38, 39, 40].
Methods of preparation are based both on physical processes and chemical
routes. In the first case ultrafast temperature treatments in confined geome-
tries are needed to reach the desired point in the carbon phase diagram where
Fig. 6. Typical Raman spectrum for an amorphous carbon thin film containing
carbynoid features. The inset shows the carbon triple-bond signal as observed with
Raman and infrared spectroscopies
carbynes seem to be stable [38]. In the second case several approaches are used
such as arc discharge between graphite electrodes submerged in various sol-
vents [41], dehydrogenation of polymers [37], electrochemical carbonization or
condensation of end-capped chain molecules produced in the gas phase [42].
Consistent amounts of sp species are also found in carbon clusters obtained
by adiabatic condensation of carbon plasmas [43, 44]. These species can be
deposited onto suitable substrates once seeded into a buffer gas which under-
goes a supersonic expansion [45].
A characteristic vibrational feature of carbyne systems is to present C–
C stretching vibrations located at higher wavenumbers (around 2100 cm−1 )
with respect to those detected for trigonal and tetrahedral carbon bonds.
A Raman spectrum in which the signal at 2100 cm−1 is evidenced for cluster
beam deposited thin films is reported in Fig. 6 as an inset [40]. Frequently
this signal is considered as the convolution of those coming from polyynes
(alternate single and triple bonds, at 2100 cm−1 ) and polycumulenes (double
bonds at 2000 cm−1 ).
The irradiation by means of 200 keV Ar+ beams produces effects on both
graphite-like and sp signals. The first is modified by changing the D/G inten-
sity ratio as already observed, while the intensity of the sp signal decreases by
increasing the ion fluence as reported in Fig. 6. Assuming that resonance en-
hancements for sp-bonded species are independent of the structural changes
associated with ion bombardment, this decrease indicates that ion irradiation
destroys the carbyne chains in terms of the disappearance of triple carbon
Fig. 7. The evolution of the carbyne features as observed using infrared absorption.
Left scale refers to the sharp 2110 cm−1 component, while right scale reports the
wide component
bond signals. At the same time, the shape of the 2100 cm−1 band shows a
mild change.
While one should expect similar trends in the analysis of the infrared ab-
sorption data, something different has been revealed as reported in Fig. 7.
As the ion fluence increases, the broad sp C–C stretching carbon band grad-
ually changes in a sharper peak located at 2110 cm−1 (inset of Fig. 7). Then,
despite the fact that sp carbon bonds are known to be highly unstable, when
irradiation is performed a new sharp component at 2110 cm−1 appears, even
though the intensity of the above-mentioned broad band decreases at increas-
ing fluences.
The rest of Fig. 7 gives quantitatively the decrease of the wide component
and the increase of the sharp one as a function of the ion fluence (φ). These
data are obtained by a proper deconvolution of the overall signal into a sharp
component (2110 cm−1 ) and two contributions related to the broad band
(polycumulene and polyyne). It is interesting to observe that fits of these
experimental data with exponential curves such as:
– 1− e−σφ (growth of the 2110 cm−1 sharp line)
– e−σφ (decrease of the broad triple bond stretching band)
give similar cross sections (σ). In particular, these are 3.7 ± 1 × 10−14 cm2 for
the wide component and 4.3 ± 1×−14 cm2 for the sharp one, suggesting a cor-
relation between the two phenomena. Similar analysis on the already reported
Fig. 8. Correlation between the 2110 cm−1 line and the observed 3300 cm−1 CH
stretching component, observed after ion irradiation
Raman data gives a cross section five times greater (see solid line in Fig. 6).
This discrepancy can be attributed to differences in the processes involved
in obtaining the vibrational features (that is, differences in the spectroscopic
processes, i.e., between Raman scattering and infrared absorption).
Following a wide consensus present in the literature, it is easy to attribute
the sharp 2110 cm−1 signal to the formation of monosubstituted acetylenic,
diacetylenic or oligoacetylenic structures which our results indicate to be
formed upon irradiation [40]. The similarity between the mentioned cross
sections confirms that these structures are formed as a consequence of the
fragmentation of the long carbyne chains induced by ion irradiation. More-
over, the decrease of the carbyne signal in the Raman spectrum can be due
to the formation of polar molecular structures, more easily revealed with IR
spectroscopy, than with Raman scattering (mutual exclusion rule).
The presence of hydrogen species in the carbon residue-containing sp
species also induces the formation of hydrogenated acetylenic molecules, as
observed looking at the CH stretching region. The general effect of the ion
beam is to decrease the intensity of the aliphatic CH stretching bonds and
to induce the appearance of a signal located at 3300 cm−1 , attributed to
alkynic CH bonds. Their formation can be related to the known effect of
ion irradiation to induce hydrogen loss in hydrogenated carbon structures
(see previous section). Such hydrogen molecules are then readily available to
interact with sp hybridized carbons. Figure 8 shows the behaviour of 2100

cm−1 and 3300 cm−1 bands as a function of ion fluence. It is straightforward
to confirm a correlation between them, indicating the formation of short
monosubstituted sp structures which are hydrogen-terminated on one side.
References
[1] R. Merlin: Raman Scattering in Material Science (2000) 505
[2] J. Orwa, K. W. Nugent, D. N. Jamieson, et al.: Phys. Rev. B 62, 5461 (2000)
506
[3] S. Prawer, R. Kalish: Phys. Rev. B 51, 15711 (1995) 506
[4] J. F. Prins: Semicond. Sci. Technol. 18, S27 (2003) 506
[5] R. Kalish, A. Reznik, S. Prawer, D. Saada, J. Adler: Phys. Status Solidi A174,
83 (1999) 506
[6] R. Brunetto, G. Baratta, G. Strazzulla: J. Appl. Phys. 96, 380 (2004) 506
[7] S. Prawer, R. Kalish: Phys. Rev. B 51, 15711 (1995) 506
[8] A. Taniike, N. Kubota, M. Takeuchi, Y. Furuyama, A. Kitamura: J. Appl.
Phys. 92, 6489 (2002) 506
[9] T. C. Sum, A. A. Bettiol, J. A. van Kan, F. Watt, E. Y. B. Pun, K. K. Tung:
Appl. Phys. Lett. 83, 1707 (2003) 506
[10] V. Philipps, M. Stamp, A. Pospieszczyk, A. Huber, A. Kirschner, E. Vietzke:
J. Nucl. Mat. 313-316, 354 (2003) 506
[11] K. Walachova, V. Svorcik, L. Bacakova, V. Hnatowicz: Biomaterials 23, 2989
(2002) 506
[12] G. A. Baratta, V. Mennella, J. R. Brucato, L. Colangeli, G. Leto,
M. E. Palumbo, G. Strazzulla: J. Raman Spec. 35, 487 (2004) 506, 508
[13] F. Salama, C. W. Bauschlicher: Molecular spectroscopy in astrophysics 57
(2001) 506
[14] W. W. Duley: Astrophys. J. 528, 841 (2000) 506
[15] E. Rimini: Ion Impantation Technology: Basics to Devices (Kluwer 1995) 506
[16] A. Battaglia, S. Coffa, F. Priolo, G. Compagnini, G. A. Baratta: Appl. Phys.
Lett. 63, 2204 (1993) 506
[17] S. Roorda, J. M. Poate, D. C. Jacobson, B. S. Dennis, S. Dierker, W. C. Sinke:
Appl. Phys. Lett. 56, 2097 (1990) 506
[18] G. Compagnini, G. A. Baratta: Appl. Phys. Lett. 61, 1796 (1992) 507
[19] B. S. Elman, M. S. Dresselhaus, G. Dresselhaus, E. W. Maby, H. Mazurek:
Phys. Rev. B 24, 1027 (1981) 507
[20] F. Tuistra, J. L. Koenig: J. Chem. Phys. 53, 1126 (1970) 507
[21] A. C. Ferrari, J. Robertson: Phys. Rev. B 61, 14095 (2001) 507, 508
[22] G. Compagnini, L. Calcagno: Mat. Sci. Eng. R 13, 193 (1994) 507
[23] G. Compagnini, G. Foti, A. Makhtari: Europh. Lett. 41, 225 (1998) 508, 510
[24] G. Compagnini, G. Foti: Nucl. Instr. Meth. B 127/128, 639 (1997) 509
[25] G. Compagnini: The physics of diamond, in A. Paoletti, A. Tucciarone (Eds.):
Proceedings of the International School of Physics “E.Fermi” Course CXXXV
(1997) p. 255 509
[26] P. Musumeci, F. Roccaforte, R. Reitano: Europh. Lett. 55, 674 (2001) 509
[27] L. Calcagno, G. Compagnini, G. Foti, M. G. Grimaldi, P. Musumeci: Nucl.

Instr. Meth. B 120, 121 (1996) 510
[28] L. Calcagno, G. Compagnini, G. Foti: Phys. Rev. B 46, 10573 (1992) 510
[29] M. E. Fragalá, G. Compagnini, A. Licciardello, O. Puglisi: J. Pol. Sci. 36, 655
(1998) 510, 511
[30] M. E. Fragalá, G. Compagnini, O. Puglisi: J. Material Research 14, 228 (1999)
510
[31] A. Raudino, M. E. Fragalá, G. Compagnini, O. Puglisi: J. Chem. Phys. 111,
1721 (1999) 510
[32] G. Compagnini, S. Pignataro, O. Puglisi: Nanostr. Mater. 12, 345 (1999) 511
[33] G. Compagnini, C. Galati, S. Pignataro: Phys. Chem. Chem. Phys. 1, 2351
(1999) 511
[34] G. Compagnini, B. Pignataro, B. Pelligra: Chem. Phys. Lett. 272, 453 (1997)
511
[35] G. Strazzulla, G. A. Baratta: Astr. Astroph. 266, 434 (1992) 513
[36] G. Ferrini, G. A. Baratta, M. E. Palumbo: Astr. Astroph. 414, 757 (2004) 513
[37] L. Kavan, J. Kastner (Eds.): Carbyne and Carbynoid Structures (Kluwer, Dor-
drecht 1999) 514, 515
[38] J. A. Lens, C. A. Perottoni, N. M. Balzaretti, J. A. H. da Jornada: J. Appl.
Phys. 89, 8284 (2001) 514, 515
[39] L. Kavan: Chem. Rev. 97, 3061 (1997) 514
[40] G. Compagnini: Materials Research Society Symposium Proceedings 851, 41
(2005) 514, 515, 517
[41] F. Cataldo: Carbon 41, 2671 (2003) 515
[42] S. E. Evsyukov: Chemical Dehydrogenation of Polymers in Carbyne and Car-
bynoid Structures (Kluwer, Dordrecht 1999) 515
[43] E. A. Rohlfing: J. Chem. Phys. 93, 7851 (1990) 515
[44] J. R. Hearth, Q. Zhang, S. C. O’Brien, R. F. Curl, H. W. Kroto, R. E. Smalley:
J. Am. Chem. Soc. 109, 359 (1987) 515
[45] L. Ravagnan, F. Siviero, C. Lenardi, P. Piseri, E. Barborini, P. Milani,
C. S. Casari, A. Li Bassi, C. E. Bottani: Phys. Rev. Lett. 89, 285506 (2002)
515
[46] L. J. Bellamy: The Infra-Red Spectra of Complex Molecules (Chapman and
Hal, London 1975)
Index
a-C, 506 carbyne, 515
a-Si1−x Cx , 508 clustering, 510
sp1 hybridisation, 514 cross-links, 510, 511, 514
amorphization, 509 diamond, 506

film stoichiometry, 508, 510
biocompatibility, 506
highly oriented pyrolytic graphite
carbon-containing ices, 512 (HOPG), 507
ices, 512 organic molecules, 510

infrared absorption, 505, 509–512,
515–517 polycumulenes, 515
interplanetary dust particles (IDPs), polymers, 510
508 polymer chains, 510
ion beam, 505 polyynes, 515
ion implantation, 506
ion irradiation, 510 Raman spectroscopy, 507–512, 514–517
ion produced hydrogenated amorphous
carbon (IPHAC), 512 surface enhanced Raman spectroscopy
(SERS), 511
linear carbon species, 514
monosubstituted sp structures, 518 vibrational spectroscopy, 505, 508, 511
oligomers, 510 wetting response, 511

Index
π states, 95–97, 99, 100, 102, 103, 423, ta-C, 98, 138, 149, 162, 163, 360,
426, 452, 456 362–365, 367, 373, 374, 376, 378,
σ states, 95–97, 427 403, 428
a-C, 55, 83, 98, 137, 149, 175, 176, 181, ta-C:H, 98, 149
428, 447, 506 ta-C:N, 149
a-C:H, 138, 218–220, 222–224, 227–230,
232–234, 426, 428–443, 447 absorption coefficient, 404, 409, 411,
a-C:H:F, 219–224, 226, 228–234 413–418
a-C:H:F:N, 219, 233 acceptor
a-C:H:N, 1–5, 10, 11, 13, 14, 16, 17, acceptor states, 304
218, 220–224, 226, 228–230, 233 donor acceptor pair recombination,
a-Si1−x Cx , 508 304
sp2 -bonded clusters, 5, 10, 18, 38, 95, accommodation length, 100, 101, 103
98–103, 137, 139, 142, 179–181, acoustic phonon, 154
183, 217, 426–429, 435, 437–439, acoustic waves, 153
442, 443 acoustic wave velocity, 154, 155, 157,
cluster conjugation level, 427 159, 160, 163
cluster size, 5, 8, 10, 11, 18, 103, surface acoustic waves (SAWs), 154,
139–141, 144, 147, 226, 359, 372, 155, 157–159, 162, 164, 168
376, 377, 426–429, 435 pseudo surface waves (PSAWs), 159
distorted clusters, 102, 103, 142 SAW velocity, 154, 155, 159
intercluster interaction, 137, 147 activation energy, 128–132
intracluster interaction, 137, 142, 150 alkaline phosphatase (ALP), 58–61, 66,
undistorted clusters, 97, 142 67, 70, 71
sp2 -chains, 139, 427, 431, 432, 435–439, amorphization, 509
443 amorphous carbon, 95, 98
chain length, 439, 443 characteristic lengths, 99, 100
sp2 -rings, 139, 141, 145, 426, 427, 431, amorphous carbon thin films, 137, 163,
434, 435, 437, 438, 442, 443 423, 505
distorted rings, 435, 437, 438, 442 annealing, 41, 175–180, 184, 228, 382,
five-fold rings, 145 403, 417, 418
six-fold rings, 145, 426 approximated solutions, 1, 2, 10, 11, 13
sp1 hybridisation, 96, 97, 514 arc discharge, 82
sp2 hybridisation, 96, 97 atomic force microscopy (AFM), 229,
sp3 backbone, 98–103 233, 234, 361, 371–373
sp3 hybridisation, 96, 97 atomic structures, 187
sp3 /sp2 bonding ratio, 17, 62, 97–99, Auger emission spectroscopy, 449–455,
101, 293, 295, 360, 374, 427 457
522 Index
background gas, 371, 372, 376, 377 C84 D2 , 198, 200

bioactivity, 55, 58, 60, 65, 73 C84 D2d , 200
biocompatibility, 55–58, 64, 71, 506 C84 Td , 198
biomaterials, 57, 64, 73 C84 Td , 200
biomedical implants, 55–57, 64, 73 carbon chemical bonding, 290
blindness carbon clusters, 194
solar blindness, 482 endohedral clusters, 192
visible blindness, 466, 468
carbon coordination, 17, 98–101, 230,
bond length alternation (BLA), 31
291
effective BLA (EBLA), 31, 34
threefold and fourfold coordination,
bone sialoprotein (BSP), 57, 58, 61, 66,
17, 292, 296
67, 70–72
carbon-containing ices, 512
Brillouin scattering, 153–159, 161–164,
167–170 carbyne, 515
surface Brillouin scattering, 157, carrier mobility, 110, 116
163–165 catalytic chemical vapour deposition
buckminsterfullerene, 78, 79 (CVD), 83
cathodoluminescence (CL), 251, 304
C168 , 198, 199 charge collection distance (CCD), 110,
C180 , 200, 201 117, 118, 130, 244, 245, 247, 249,
C186 , 198 256, 263, 306–309, 315
C20 , 200 charge collection efficiency (CCE),
C24 , 363 240, 242, 244–254, 258, 260–263,
C2610 , 196 280–285
C264 , 196 CCE maps, 242, 244
C276 , 198, 199 charge collection spectrum, 111–113,
C276 :C84 , 198, 199 125
C28 , 196 charge lifetime, 487
C36 , 190, 200 charge mobility, 487
C408 , 198, 199 chemical shift, 290
C60 , 78, 79, 187, 191–194, 198, 200, 359 chemical vapour deposition (CVD), 83,
C60 :Li, 207 107–112, 269, 329, 463, 464
C60 :N, 192 CVD heteroepitaxial growth
C60 :Rb, 192 (CVDHG), 269
C60 with hydrogen, 207–211
CVD heteroepitaxial growth
doped C60 , 191
(pCVD), 269
C67 , 203
radio frequency CVD, 429
C70 , 79, 190, 194
chiral vector, 79
C72 , 194
C74 , 192–195, 203 cluster assembled (CA) films, 362,
C74 :Ca, 194 368–375, 377
C74 :Eu, 194 cluster model, 426
C74 :Gd, 194 cluster–cluster interaction, 100
C74 :La, 194 cluster-assembled (CA) films, 372, 376,
C74 :Si, 192–195 378
C76 , 194 cluster-cluster interaction, 147
C82 , 192 cluster-cluster separation, 149
C84 , 192, 198, 199 clustering, 97, 137, 145, 146, 360,
C84 :C24 , 197 426–429, 435, 437, 438, 510
Index 523
graphite-like clusters, 175, 180, 181, defects, 27, 28, 33–35, 108, 142, 144,
183 148, 383, 395
CNx thin films, 289, 290, 296, 298 defects distribution, 108, 110–113,
coatings, 56–58, 67, 71–73, 360 115–117, 119, 133, 279, 280
bioactive coatings, 55 grain boundaries, 108, 109, 112, 113,
biomedical coatings, 57, 360 115, 116, 118–120, 333, 341, 345,
protective coatings, 218 346, 470
coherence length, 372, 377, 428 grain boundaries distribution, 112,
columnar film, 367, 368, 371 113, 116, 118
node-like film, 367, 368, 371 in diamond, 329, 333, 334, 341, 347,
columnar growth, 109, 111–113, 115, 352–356
116, 118 band A, 304
confinement, 28, 36, 38, 39, 382–384, band A, 308, 310, 311, 315, 316
392–395, 397, 399, 400 band B, 308–311, 313, 321–323
conjugation length, 30, 383 neutron irradiation, 313, 315, 324
effective conjugation length (ECL), in-grain defects, 112, 113, 115, 116,
30 118, 119
coplanar contacts, 480, 489 in-grain defects distribution, 112,
coplanar structure, 485 113, 115, 118
core level photoelectron peak, 290 dendritic film, 367, 371
cross-links, 510, 511, 514 density, 224, 227–229, 293
crystallite size, 428 density functional theory (DFT), 32,
cumulene, 32 386, 390–393
current detector efficiency, 110
persistent, 304, 308, 317, 319–324 detectors, 108, 240, 241, 244, 248, 253,
radiation induced, 304, 306, 308, 311, 254, 256, 261, 263
313 CCD detectors, 491, 493, 496
sensitivity, 311 CMOS detectors, 490, 496
CVD diamond, 242, 244, 254, 263, detector gain, 474
345–348, 464, 468, 469, 472 diamond-based detectors, 346, 474,
homoepitaxial CVD diamond, 464, 481
465 imaging detectors, 463
single-crystal diamond, 481 MCP detectors, 491, 496
cytotoxicity, 58, 59, 64, 65, 71 MOS detectors, 488
pixel-array detectors, 489, 497
D band, 36–38, 63, 145, 149, 177, 179, single-pixel detectors, 497
180, 224, 225, 364, 384, 391, 406, time response, 476
427–432, 434–438, 440 UV detectors, 463–465, 472, 474, 476,
D-band dipersion, 37 481, 484, 485
D-band, 38 detrapping, 123–129, 131
D/G band intensity ratio, 177, 179–184, diamond, 26, 41, 96, 97, 107, 161, 217,
225, 226, 406, 409, 416, 417, 428, 329, 359, 426, 427, 463, 467, 506
431, 432, 438, 441 defects in diamond, 269, 270
D/G-band intensity ratio, 145, 149 diamond detector array, 489
D/G band intensity ratio, 295, 366, 375 diamond detectors, 267–273, 277, 278
D/G band intensity ratio, 295, 298–300, diamond growth, 465, 468, 469
364, 372 diamond nucleation, 329, 336, 338,
dark current, 472 339
524 Index
diamond properties, 345, 467, 468, elastic constants, 155, 156, 158,
470, 472 160–162, 164, 165, 168
high-quality diamond, 347, 348, elastic moduli, 155, 159–161, 163, 166
351–357 elastic properties, 154, 155, 170
highly-oriented diamond, 493 elastic scattering, 153
homoepitaxial diamond, 345–348, electrical conductivity, 102, 103
352, 355–357, 470 electrical properties, 63, 218, 219
HPHT diamond, 346, 469, 470, 474 electroluminescence, 304
IIa type diamond, 469 electron diffraction (ED), 176–181, 183,
IIb type diamond, 348, 352 184
natural diamond, 348, 352, 463, 466, electron energy loss spectroscopy
468–470, 474, 488 (EELS), 224, 226, 361, 362, 367,
nitrogen contamination, 347, 353–356 368, 449–452, 458–460
silicon contamination, 347 electron energy loss spectroscopy
single-crystal diamond, 345, 463–465, (EELS), 455
474, 484, 488 electron paramagnetic resonance
transport in diamond, 273–279 (EPR), 142
Hecht’s model, 275 electron spin resonance (ESR), 142–144,
diamond-like carbon (DLC), 25, 56, 57, 146, 150
97, 99, 138, 139, 144, 147, 149, ESR signal, 102, 103
150, 162, 164–166, 175, 218, 224, electron–phonon coupling, 28, 30, 38,
228–230, 234, 294, 359, 360, 362, 382, 385, 387–393, 395, 398, 399
363, 365, 368, 376, 403, 404, 406, electron-phonon coupling, 384, 425
412, 413, 415, 416, 418, 428, 430, electron-photon coupling, 425
431, 437, 443 electronic properties, 95
diamondlike carbon (DLC), 217, 293 external quantum efficiency (EQE),
dimensionless arguments, 4–6, 8 473, 480–482
dimerisation, 386, 390, 393, 394
bonds dimerisation, 395 fading
dimerisation parameter, 387 optical fading, 305
diodes thermal fading, 310, 317, 319
MIS diode, 488 fast component, 123, 128
disorder, 100, 101, 137, 139–141, 150 fast photography, 287–289, 296, 299
homogeneous disorder, 140 figure of merit for growth process, 17
inhomogeneous disorder, 140, 141 film stoichiometry, 3, 11, 12, 14–17,
dispersion, 166, 383, 384, 386, 398, 399, 508, 510
423, 428, 432, 435–437, 439, 440, first order kinetics, 310, 316, 319
443 first principle calculations, 24, 28, 30,
phonon dispersion, 26, 388, 391 32, 34, 386, 398, 400
distortions, 98–103 floppiness, 98
donor fluorine incorporation, 217–219, 223,
deep acceptor centers, 308 224, 226–228, 230, 231
deep donor level, 305, 308 free expansion regime, 297
donor acceptor pair recombination, frequency, 316
304 friction coefficient, 56, 57, 219, 229,
D band, 293, 294, 296, 298 232, 233, 360
fullerene, 25, 41, 187–192, 218, 359, 383
effective conjugation coordinate theory fullerene materials, 188, 189, 207
(ECCT), 38, 385, 388, 390 giant fullerene, 192, 199
Index 525
fullerite, 162 hydrogen desorption, 14, 16, 42,

455–458, 460
G band, 3, 4, 8, 10, 11, 17, 18, 37, hydrogen incorporation, 218
63, 140, 141, 177–180, 224, 225, hydrogen storage, 44, 207–209, 211
364, 395–398, 406, 416, 417, 423, hydrogenated amorphous carbon, 162
427–432, 435, 437, 438, 440 hydrogenated amorphous carbon-
dispersive G band, 435–437, 439, 440 nitrides, 1
non-dispersive G band, 435, 437, 439, hydrogenated diamond-like carbon
440 (DLC:H), 99, 138, 142, 149
G-band, 150 hydrogenated graphite-like carbon
G-band, 145, 149 (GLC:H), 99
glassy carbon, 98 hydrogenated polymer-like carbon
glow discharge, 84, 469 (PLC:H), 99
graphene, 78–80, 97, 102, 162, 218, hydrophobicity, 219, 231, 232, 235
381–384, 390, 392, 393, 396, 427
graphite, 28, 41, 96, 97, 161, 217, 359,
ices, 512
426–428, 430, 431, 435
image sensors, 489
graphite-like carbon (GLC), 25, 139,
CMOS image sensors, 490
142, 175, 218, 287
UV imagers, 490
graphitization, 41, 147, 175, 179, 180,
inelastic scattering, 153
184, 205, 230, 234, 292, 293, 299,
430, 438, 443 infrared (IR) spectroscopy, 24, 25,
33–36, 42, 224, 227, 229, 361–364
G band, 293, 294, 298
infrared absorption, 505, 509–512,
Hückel theory, 382, 384, 386, 387, 389, 515–517
391, 392, 394, 399 interference of light, 409
Hückel approximation, 426 internal stress, 217, 218, 229–231, 234
haemocompatibility, 56, 57 interplanetary dust particles (IDPs),
hardness, 56, 57, 218, 229–231, 233, 508
234, 360 ion beam, 505
Hecht theory, 110, 111, 114, 116, 118, ion beam analysis (IBA), 220, 224
275, 278 ion beam induced charge (IBIC),
heteroepitaxial growth, 329, 346, 470 241–243, 245–252, 257, 258, 260,
high-pressure high-temperature 261, 263
(HPHT), 107, 346, 465, 469 frontal IBIC, 242–244, 262, 263
high-resolution electron microscopy IBIC maps, 244, 245, 252, 259–263
(HREM), 191, 192 lateral IBIC, 242, 243, 245–250, 253,
highest occupied molecular orbital 259, 262, 263
(HOMO), 194 ion beam induced luminescence (IBIL),
highly oriented pyrolytic graphite 240–242, 251–256, 263
(HOPG), 383, 507 frontal IBIL, 253, 254, 256
HOMO–LUMO gap, 29, 30 IBIL maps, 251–253, 255, 256, 263
HOMO-LUMO gap, 194, 203 lateral IBIL, 252, 253
homoepitaxial growth, 329, 346, 465, ion beam-induced charge (IBIC), 240
470 ion beam-induced luminescence (IBIL),
hot filament, 469 240
hybrid arc-discharge, 187, 189, 190, ion bombardment (IB), 176, 181–184
192, 203 ion implantation, 506
hydrogen adsorption, 42, 44, 45 ion irradiation, 416, 510
526 Index
ion produced hydrogenated amorphous nanocapsules, 187, 190, 192, 200,

carbon (IPHAC), 512 202–205, 207–211
self-organization, 187, 209
Kohn anomaly, 28, 32, 392 one-dimensional self-organization,
209
laser vaporisation, 82 nanofibres, 80, 82
lattice dynamics, 24, 27, 29, 33 herringbone nanofibres, 81
light scattering, 154 stacked nanofibres, 81
limited hyperbolic function, 321 nanohorns, 187, 202, 203
linear carbon species, 514 nanoparticles, 210
linear model, 308, 310 gold nanoparticles, 210, 211
low-frequency Raman, 169 nanostructured carbon, 162
lowest unoccupied molecular orbital nanostructures
(LUMO), 194 hollow-cage-nanostructures, 187
nanotribology, 229
magnetron sputtering, 176 nanotubes, 25, 28, 41, 77–85, 87–89,
mass spectrometry, 84, 85 166, 187, 192, 196, 201–203,
mass-separated ion beams (MSIB), 208, 210, 211, 218, 359, 381–383,
403–405, 407–414, 416–418 395–399
mechanical properties, 98, 99, 101–103, armchair nanotubes, 79, 80, 398
153, 217–219, 229 chiral nanotubes, 79
metal-insulator-metal (MIM), 485 chiral vector, 79
metal-oxide-semiconductor (MOS), 488 chirality, 397
metal-semiconductor-metal (MSM), chiral vector, 396
485 growth mechanism, 85
metal/diamond interfaces, 478 mininanotube, 200
metallofullerene, 187, 190, 192 multi-walled nanotubes (MWNTs),
microscopic growth variables, 12–15 200
microstructure, 175, 176, 178–181, 183, multiwalled nanotubes (MWNTs),
184, 217, 219, 224, 228, 233 77, 79, 82, 84, 206
graphite-like structure, 179, 183 single-walled nanotubes (SWNT),
mineralization, 57, 60, 64–67, 71 166, 167
Mo substrates, 329–341 single-walled nanotubes (SWNTs),
molecular approach, 24, 29, 38, 384, 79, 81, 82, 156, 162, 167–169, 200,
391 395–397, 399, 400
molecular dynamics, 24, 27, 33, 35, 207, zig-zag nanotubes, 79, 80, 398
390 nanowires, 200, 203, 210, 211
monosubstituted sp structures, 518 nitrogen, 374, 376, 377
Monte Carlo simulation, 113 nitrogen incorporation, 10, 12–17, 147,
MTT test, 65, 67 217–219, 221, 224, 226, 228, 230,
multi-wavelength Raman spectroscopy 232, 234
(MWRS), 428 normal modes, 424
multiwavelength Raman spectroscopy nucleation, 329–341
(MWRS), 37–40 fast phase, 338, 339
nucleation probability, 339
nano-glassy carbon, 372, 377 nucleation rate, 338
nanocages, 187, 189, 207 slow phase, 338
nanocage materials, 189 statistical study, 336–339
Index 527
surface nucleation density, 338 plasma-enhanced chemical vapour

deposition (PECVD), 77, 83, 84,
oligomers, 25, 27, 28, 30–32, 382, 390, 87, 89, 138, 142, 218, 219, 228,
510 233, 345, 469
onions, 187, 196–199 plasmon energy, 451–453
tetrahedral onions, 196–198 polarizability, 388, 390, 391, 395, 397,
optical characterization, 403, 413 424, 425
optical gap, 102, 103 bond–bond polarizability, 398
optical microscopy (OM), 345, 347–349, bond-bond polarizability, 386, 388
355 polarizability tensor, 424
optical properties, 59, 102, 137, 218, polyacetylene, 382–384, 386–390, 392,
360, 404, 410, 411, 413, 414, 418 393
optoelectronic properties, 99, 101–103 long chains, 385, 388
organic molecules, 510 short chains, 385
polyacetylene (PA), 28, 29, 32, 41
polyconjugated systems, 31
Pócsik diagram, 37, 39, 40
polycrystalline diamond, 108, 109, 112,
Peierls distortion, 29, 32, 39, 389
116, 345–347, 352, 354, 356, 357,
penetration depth, 112, 471, 479
463–465, 469, 474, 488, 495, 497
percolation, 98, 101, 102
polycumulenes, 515
percolation threshold, 98 polycyclic aromatic hydrocarbons
phonons, 423, 425 (PAHs), 36–38, 42, 382, 390, 393
zone centre phonons, 424 polyenes, 383, 385, 386, 389, 390, 392
photoconductive gain, 473, 482 polymer-like carbon (PLC), 138, 139,
photoconductivity, 103, 304, 305, 308, 142, 144, 147, 224, 227–229, 234
470 polymers, 29, 30, 33, 34, 217, 383, 384,
photoconductors, 472, 485 510
UV photoconductors, 470, 493 polyconjugated polymers, 30, 31, 34,
photocurrent, 486 383
photocurrent persistence, 476, 478, polymer chains, 510
479, 482 polymethylene, 25
photodetectors, 468, 472, 485, 494 polyyne, 32, 33, 41
diamond-based photodetectors, 497 polyynes, 515
photodiodes, 487, 488 priming, 478
photoemission spectroscopy, 450, UV priming, 478
453–461 process approximation, 2, 13–15, 18
photoluminescence (PL), 187, 189, process modelling, 13–15
204, 329, 331, 333–335, 345, 347, proliferation, 57, 59, 64, 65, 71
352–355 pulsed cathodic arc discharge (PCAD),
micro-PL analysis, 333–336, 348, 404, 406, 408–412, 416, 418
352–356 pulsed laser ablation (PLA), 288, 296,
photon counting, 495 299
photoresponse, 475, 479 pulsed laser deposition (PLD), 360,
vacuum UV photoresponse, 475 361, 376
physical approximants, 1, 10, 11, 18 pyrolytic carbons, 37, 42
plasma, 83
plasma expansion, 296, 299 Q-arguments, 1, 6, 8, 9
plasma plume, 361, 373 quantum chemistry, 38, 382
plume stopping, 297 quantum efficiency, 463, 472, 473, 486
528 Index
quasi equilibrium (QE) approximation, modified Rayleigh wave velocity,

321 160, 163
quasi- elastic scattering, 154 Rayleigh wave velocity, 154, 161, 162,
quasi-equilibrium (QE) approximation, 164
312 reactive sputtering, 2, 13
reflection electron energy loss spec-
Raleigh scattering, 424 troscopy (REELS), 288–292, 295,
300
Raman dimensionless arguments, 6–8
refractive index, 63, 403, 404, 409–411,
Raman parameters, 406–409, 414
413, 414, 416, 418
Raman scattering, 35, 154, 350, 351,
residual gas analysis (RGA), 84, 85, 88,
354, 399, 409, 418, 423
89
anti-Stokes scattering, 423
resistivity, 175–178, 180, 181, 183
first-order Raman scattering, 350
resistivity anisotropy, 175–181, 184
infrared excitation, 423, 429,
response time, 482
432–435, 437–439
rigidity, 98, 100
Raman activity, 385, 386, 388–390,
393, 400 sandwich contacts, 479, 489
Raman cross section, 29, 36 sandwich structure, 485
Raman intensity, 34, 386, 410, 414, scaling laws, 1, 2, 15, 17, 18
415, 426 derivation method, 1, 4, 7–9, 17
Raman spectrum, 426 validity range, 9
Raman tensor, 424, 425 scanning electron microscopy (SEM),
resonant Raman scattering, 28, 30, 60, 65, 68, 69, 71, 330, 333, 334,
383, 386, 425, 428 336, 345, 347–350, 355, 361, 362,
second-order Raman scattering, 351, 367, 369–371
352 scanning tunneling microscopy (STM),
selection rules, 28, 384, 393, 425 147
breakdown of the selection rules, Schottky barrier, 486–488
426 Schottky contacts, 488
Stokes scattering, 423 Schottky diode, 488
visible excitation, 423, 424, 427, 429, Sezawa wave, 155, 159, 160, 162
438 shock wave, 296, 298–300
Raman spectroscopy, 1–3, 17, 24–28, SiC, 200, 202–205
32–36, 42, 63, 137, 140–142, 144, silicon incorporation, 218
145, 149, 176, 179–181, 224–226, slow component, 123–125, 127–129
229, 288, 289, 293–295, 298–300, spectral response, 481
329–333, 335–337, 345, 347, 348, spin density, 102, 103
350–352, 355, 356, 362–365, 371, subplantation model, 217, 218, 374
373, 381, 383–386, 390–392, 395, substrate bias, 181
403–406, 408, 409, 418, 507–512, surface enhanced Raman spectroscopy
514–517 (SERS), 32, 404, 411, 511
micro-Raman analysis, 331–333, 337, surface treatment, 495
348, 351–356
Ramo’s theorem, 110, 123–125, T band, 364, 365
245–247, 274 T/G band intensity ratio, 366
rate equations, 303, 309, 312, 313 T/G band intensity ratio, 365
Rayleigh wave, 154, 155, 159, 162, 164 Tauc gap, 137, 139–141, 144, 147, 150
modified Rayleigh wave, 159, 160 tetrahedral amorphous carbon, 98, 139
Index 529
tetrahedral hydrogenated amorphous Urbach energy, 137, 139–141, 150

carbon, 98 UV photoemission spectra (UPS), 450,
thermal stability, 218, 219, 228, 417 461, 462, 524
thermal-induced gas effusion, 228
time constant, 123, 126, 128–132 valence force field, 25–27, 30, 382,
transmission electron microscopy 385–387, 390, 392, 396–398
(TEM), 176, 177, 181–183, 361, variable energy method (VEM),
373, 377 119–122, 130
trap, 108 variable incidence method (VIM),
density per unit energy, 310, 315 119–122, 130
depumping, 108, 128, 129, 307, 310 viability, 58, 59, 65, 71
priming, 108, 111–113, 115, 117 vibrational spectroscopy, 505, 508, 511
pumping, 108, 111, 113, 115, 117,
122, 124, 125, 128–130, 306, 307, wear resistance, 56, 57, 229, 233, 234
311 wettability, 219, 232
trap filling, 307, 311
wetting response, 511
trapping, 110, 113, 123–131
tribological properties, 56, 217, 219,
X-ray beam induced charge (XBIC),
229
257–263
ultra-large scale integrated circuits XBIC maps, 258–263
(ULSI), 209, 219, 229 x-ray diffraction (XRD), 469
self-assembled ULSI, 212 X-ray photoemission spectroscopy
ultra-thin carbon films, 156, 162–166 (XPS), 288, 291, 292, 295,
ultrananocrytalline diamond (UNCD), 298–300
148, 149 x-ray photoemission spectroscopy
unipolar conduction, 308–310 (XPS), 224, 227, 289–291, 449

Plasma-Surface Modification of Biomaterials

Uploaded by

Document Information

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

Plasma-Surface Modification of Biomaterials

Uploaded by

Copyright:

Available Formats

Topics in Applied Physics

With 245 Figures, 6 in Color, and 24 Tables

Library of Congress Control Number: 2006924284

Physics and Astronomy Classiﬁcation Scheme (PACS):

ISSN print edition: 0303-4216

© Springer-Verlag Berlin Heidelberg 2006

To my sons Gloria and Alfonso,

Carbon is a surprisingly versatile element, able to hybridise in three diﬀer-

and those of graphite (graphite-like carbon, GLC) can be deposited by the

tion methods. Surprisingly, however, there are even manifold applications of

the future challenges consists in the creation of highly integrated CNT-based

It appears now clear that carbon-based materials constitute a topic of

etc.) or to contribution type (fundamental studies, technological applications,

Reggio Calabria, Giacomo Messina

Aid of Scaling Laws in the Achievement of a Well-Controlled

2.4 Cytotoxicity Test . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59

3 Analysis of the Charge Collection Spectrum . . . . . . . . . . . . . . . . . . . . 111

Elastic and Structural Properties of Carbon Materials

3.2 Substrate Bias Eﬀects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 181

Measurements of Defect Density Inside CVD Diamond Films

2 A General Model for Transport Properties of pCVD Diamond:

Pulsed Laser Deposition of Carbon Films: Tailoring Structure

Raman Spectroscopy of CVD Carbon Thin Films Excited by

9.3 Space Astronomy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 495

Giacomo Messina and Saveria Santangelo

Dipartimento di Meccanica e Materiali, Università “Mediterranea” di Reggio

Table 1. Values of the dimensionless argument Q corresponding to the growth

Film Q ΦN2 ΦH2 ΦHe ΦAr p W

Fig. 1. Raman spectra of a-CN:H

This, as demonstrated in Fig. 2, where the G band frequency position (ωG )

3.2 Operative Scaling Law Derivation Procedure

A simpler picture might derive from ﬁnding a dependence of Raman de-

Film ωG ID /IG LC C content H content N content O content

Fig. 2. Dependence of the G band frequency position (ωG ) on rf power (W ) and

ate Raman descriptive parameters with appropriate dimensionless Raman

Stage 3. A new function of Q is chosen for approximating the experimental

3.3 Choice of Q-arguments

According to the procedure of Sect. 3.2, the G band frequency position is

3.4 The Scaling Law for the Growth Conditions

At the end of the above described procedure, the scaling law,

3.5 Resulting Simpliﬁcation of the Process Picture

As demonstrated in Fig. 4, a progressive process picture simpliﬁcation is

xN (%) 7.26 + 0.85 exp[((Q/0.8)0.1 )].

Fig. 5. Dependence of ﬁlm N content (xN ) on the dimensionless argument Q. Open

ωG (cm−1 ) 1565.8 − 3.0Q + 9.5 exp[−(Q/0.8)8]

whose form, empirically demonstrated by ﬁtting the experimental data, is

Fig. 6. Dependence of the G band frequency position (ωG ) on the dimensionless

Fig. 8. Average ﬁlm growth rate as a function of the dimensionless argument Q.

of such an eﬀective combination suggests the possibility that, upon variation

4.2 Modelling of the Growth Process

It is important to remember that an approximate, rather than an exact solu-

Q = (Ei,r )a · (FC→s )c , (2)

with a and c exponents to be empirically determined. Having elementarily

is due to the diﬀerent sticking coeﬃcients of N- and H-containing precur-

Table 3. Proportionality relationships between microscopic (internal) growth vari-

Internal variable Proportionality with external parameters From refs.

precursors incorporated. Accordingly, the ﬂux of precursor species reaching

FC→s ∝ (FAr→t )b · (FAr→s )−d · (Frg→s )f · (trg )h · (sn,h )−l , (3)

with b, d, f , h and l positive exponents to be empirically determined. FAr→s

4.3 Physical Meaning of the Scaling Law

As accurately discussed in [54], in a-CN:H preparation by rf diode systems,

4.4 The Scaling Law for the Microscopic Growth Variables

As hypothesised, Q depends inversely on FAr→s and sn,h . This ﬁnding indi-

4.5 A Figure of Merit for a-CN:H Film Deposition