Professional Documents
Culture Documents
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
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
springer.com
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
Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 521
Aid of Scaling Laws in the Achievement
of a Well-Controlled Film Deposition Process
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)
© Springer-Verlag Berlin Heidelberg 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
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.
4 Giacomo Messina and Saveria Santangelo
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
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
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
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.
Scaling Laws for Film Deposition 9
(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.
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
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
12 Giacomo Messina and Saveria Santangelo
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.
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)
16 Giacomo Messina and Saveria Santangelo
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
The existence of the latter reflects onto (and, thus, explains fully and defi-
nitely) the existence of the former.
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
18 Giacomo Messina and Saveria Santangelo
Acknowledgements
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
Scaling Laws for Film Deposition 19
[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
Scaling Laws for Film Deposition 21
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
1 Introduction
G. Messina, S. Santangelo (Eds.): Carbon, The Future Material for Advanced Technology Ap-
plications, Topics Appl. Phys. 100, 23–53 (2006)
© Springer-Verlag Berlin Heidelberg 2006
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.
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.
2.1 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:
A Spectroscopic Approach to Carbon Materials for Energy Storage 27
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
28 Giuseppe Zerbi et al.
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
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
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.
32 Giuseppe Zerbi et al.
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].
A Spectroscopic Approach to Carbon Materials for Energy Storage 33
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].
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
36 Giuseppe Zerbi et al.
(a) (b)
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).
40 Giuseppe Zerbi et al.
(a)
(b)
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.
A Spectroscopic Approach to Carbon Materials for Energy Storage 43
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 –
the TRIO band, thus indicating that the sizes of the clusters have increased.
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
A Spectroscopic Approach to Carbon Materials for Energy Storage 45
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.
A Spectroscopic Approach to Carbon Materials for Energy Storage 47
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
48 Giuseppe Zerbi et al.
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
A Spectroscopic Approach to Carbon Materials for Energy Storage 49
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
1 Introduction
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
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].
The films deposited on silicon substrates were used to obtain the thickness us-
ing a DEKTAK profilometer. The optical absorption in the ultraviolet–visible
Biocompatibility, Cytotoxicity and Bioactivity 59
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.
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
60 Sandra E. Rodil et al.
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
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.
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
Biocompatibility, Cytotoxicity and Bioactivity 61
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.
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.
62 Sandra E. Rodil et al.
3 Results
3.1 Film Properties
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.
64 Sandra E. Rodil et al.
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.
Biocompatibility, Cytotoxicity and Bioactivity 65
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
66 Sandra E. Rodil et al.
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
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,
68 Sandra E. Rodil et al.
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
Biocompatibility, Cytotoxicity and Bioactivity 69
Fig. 10. Western blot analysis of bone sialoprotein (top) and densitometric analysis
of the bands, showing the intensity per µm2 (bottom)
Biocompatibility, Cytotoxicity and Bioactivity 71
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
72 Sandra E. Rodil et al.
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
Biocompatibility, Cytotoxicity and Bioactivity 73
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
74 Sandra E. Rodil et al.
[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
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-
G. Messina, S. Santangelo (Eds.): Carbon, The Future Material for Advanced Technology Ap-
plications, Topics Appl. Phys. 100, 77–93 (2006)
© Springer-Verlag Berlin Heidelberg 2006
78 Martin S. Bell et al.
A
B
Fig. 1. AB stacking in graphite
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. 4. “Armchair (5,5)” and “zig-zag (9,0)” nanotubes (top and bottom, respec-
tively)
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 .
+
catalyst C xH y
particle C
C C
C C C xH y catalyst
Cx Hy
particle
barrier layer barrier layer
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
Growth Mechanism During PECVD of Multiwalled Carbon Nanotubes 87
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
90 Martin S. Bell et al.
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
Growth Mechanism During PECVD of Multiwalled Carbon Nanotubes 91
Index
a-C, 83 C60 , 78, 79
C70 , 79
arc discharge, 82
catalytic chemical vapour deposition
buckminsterfullerene, 78, 79 (CVD), 83
Growth Mechanism During PECVD of Multiwalled Carbon Nanotubes 93
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.
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-
Correlation Between Local Structure and Film Properties 99
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.
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
102 Giovanni Fanchini and Alberto Tagliaferro
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.
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.
104 Giovanni Fanchini and Alberto Tagliaferro
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
Correlation Between Local Structure and Film Properties 105
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-
G. Messina, S. Santangelo (Eds.): Carbon, The Future Material for Advanced Technology Ap-
plications, Topics Appl. Phys. 100, 107–135 (2006)
© Springer-Verlag Berlin Heidelberg 2006
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.
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.
110 Marco Marinelli et al.
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-
Defects in CVD Diamond Films 113
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
114 Marco Marinelli et al.
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.
Defects in CVD Diamond Films 115
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
116 Marco Marinelli et al.
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:
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.
Defects in CVD Diamond Films 117
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
118 Marco Marinelli et al.
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)
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.
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.
120 Marco Marinelli et al.
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
Defects in CVD Diamond Films 121
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
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
4 Time-Domain Analysis
4.1 The Background
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
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
Defects in CVD Diamond Films 127
Fig. 10. Comparison of measured pulses with the simulation based on the trap-
ping–detrapping model in the as-grown and pumped states
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
Defects in CVD Diamond Films 129
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
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.
130 Marco Marinelli et al.
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
Defects in CVD Diamond Films 131
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
132 Marco Marinelli et al.
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.
Defects in CVD Diamond Films 133
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
134 Marco Marinelli et al.
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
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.
G. Messina, S. Santangelo (Eds.): Carbon, The Future Material for Advanced Technology Ap-
plications, Topics Appl. Phys. 100, 137–152 (2006)
© Springer-Verlag Berlin Heidelberg 2006
138 J. David Carey and S. Ravi P. Silva
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.
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.
142 J. David Carey and S. Ravi P. Silva
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])
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
144 J. David Carey and S. Ravi P. Silva
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
Effects of Nanoscale Clustering in Amorphous Carbon 145
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
146 J. David Carey and S. Ravi P. Silva
Fig. 6. Schematic of how the different bias growth conditions can give rise to a
change in the size and concentration of sp2 clusters
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
Effects of Nanoscale Clustering in Amorphous Carbon 151
[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
152 J. David Carey and S. Ravi P. Silva
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.
G. Messina, S. Santangelo (Eds.): Carbon, The Future Material for Advanced Technology Ap-
plications, Topics Appl. Phys. 100, 153–174 (2006)
© Springer-Verlag Berlin Heidelberg 2006
154 Marco G. Beghi et al.
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].
156 Marco G. Beghi et al.
2 Experimental Technique
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 ,
Brillouin Light Scattering 157
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)
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.
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.
Brillouin Light Scattering 161
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.
Brillouin Light Scattering 165
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.
166 Marco G. Beghi et al.
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-
Brillouin Light Scattering 167
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
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,
Brillouin Light Scattering 169
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].
170 Marco G. Beghi et al.
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
Brillouin Light Scattering 171
[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
Brillouin Light Scattering 173
[62] C. Mapelli, C. Castiglioni, G. Zerbi, K. Müllen: Phys. Rev. B 60, 12710 (1999)
169
174 Marco G. Beghi et al.
Index
ta-C, 162, 163 graphite, 161
Alexei A. Onoprienko
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.
G. Messina, S. Santangelo (Eds.): Carbon, The Future Material for Advanced Technology Ap-
plications, Topics Appl. Phys. 100, 175–186 (2006)
© Springer-Verlag Berlin Heidelberg 2006
176 Alexei A. Onoprienko
Fig. 1. Schematic draw of specimen with Ni contacts (1–1 and 2–2 ) and carbon
film for resistance measurements
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
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
180 Alexei A. Onoprienko
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.
Fig. 4. TEM micrographs of carbon films subject to (a) and not subject to (b) in-
tense IB during deposition
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
Resistivity and Structure of Magnetron-Sputtered Carbon Films 185
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
186 Alexei A. Onoprienko
[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
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
G. Messina, S. Santangelo (Eds.): Carbon, The Future Material for Advanced Technology Ap-
plications, Topics Appl. Phys. 100, 187–216 (2006)
© Springer-Verlag Berlin Heidelberg 2006
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
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
190 Takeo Oku et al.
Fig. 2. HREM images of (a) C36 and (b) C70 clusters. Atomic structural models
of (c) C36 and (d) C70 clusters
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.
Carbon Nanocage Materials 193
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
194 Takeo Oku et al.
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 ,
Carbon Nanocage Materials 195
Fig. 5. Energy level diagram and density of states of (a) C74 , (b) Si@C74 , and
(c) a single Si atom
196 Takeo Oku et al.
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.
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
198 Takeo Oku et al.
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-
Carbon Nanocage Materials 199
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
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,
Carbon Nanocage Materials 201
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
202 Takeo Oku et al.
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
Carbon Nanocage Materials 203
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-
204 Takeo Oku et al.
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].
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
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
Carbon Nanocage Materials 209
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.
Table 3. Energy of C60 cluster with hydrogen. A C–C bond was broken in case of
8 chemisorbed hydrogen atoms (∗ )
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
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
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
Carbon Nanocage Materials 213
[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
216 Takeo Oku et al.
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
G. Messina, S. Santangelo (Eds.): Carbon, The Future Material for Advanced Technology Ap-
plications, Topics Appl. Phys. 100, 217–238 (2006)
© Springer-Verlag Berlin Heidelberg 2006
218 Fernando L. Freire Jr.
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
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
222 Fernando L. Freire Jr.
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
Hard Amorphous Hydrogenated Carbon Films and Alloys 223
Fig. 3. Fluorine and hydrogen content in a-C:H:F films as a function of the self-bias
voltage
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
Hard Amorphous Hydrogenated Carbon Films and Alloys 225
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
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
Hard Amorphous Hydrogenated Carbon Films and Alloys 227
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-
Hard Amorphous Hydrogenated Carbon Films and Alloys 231
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
232 Fernando L. Freire Jr.
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
Hard Amorphous Hydrogenated Carbon Films and Alloys 233
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
234 Fernando L. Freire Jr.
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
5 Conclusion
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
236 Fernando L. Freire Jr.
[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.:
Diamond Relat. Mater. 12, 2037 (2003) 219
[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:
Diamond Relat. Mater. 10, 125 (2001) 219
[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
Hard Amorphous Hydrogenated Carbon Films and Alloys 237
[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
238 Fernando L. Freire Jr.
Claudio Manfredotti
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.
G. Messina, S. Santangelo (Eds.): Carbon, The Future Material for Advanced Technology Ap-
plications, Topics Appl. Phys. 100, 239–265 (2006)
© Springer-Verlag Berlin Heidelberg 2006
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
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
242 Claudio Manfredotti
Fig. 2. Details of the scattering chamber used in LNL for IBIC/IBIL measurements
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
244 Claudio Manfredotti
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
Ion Microscopy on Diamond 245
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).
246 Claudio Manfredotti
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
Ion Microscopy on Diamond 247
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
Ion Microscopy on Diamond 249
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
250 Claudio Manfredotti
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.
Ion Microscopy on Diamond 251
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-
252 Claudio Manfredotti
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
Ion Microscopy on Diamond 253
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
254 Claudio Manfredotti
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
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)
256 Claudio Manfredotti
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.
Ion Microscopy on Diamond 257
Fig. 22. The ID21 beam line of ESRF and the scanning X-ray microscope, with a
layout of XBIC
5 XBIC
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
Ion Microscopy on Diamond 259
Fig. 25. XBIC photocurrent maps obtained at different (positive and negative)
bias voltages over the region indicated by a square in Fig. 24
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
Ion Microscopy on Diamond 261
Fig. 27. IBIC map obtained from the same data as Fig. 6 by a suitably selected
grey scale
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
262 Claudio Manfredotti
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.
Ion Microscopy on Diamond 263
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
264 Claudio Manfredotti
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
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
Ion Microscopy on Diamond 265
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
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.
a)
b)
-1
2.4 cm
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.
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
Defect Density Measurements Inside CVD Diamond Films 271
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
272 Renato Potenza and Cristina Tuvé
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
Defect Density Measurements Inside CVD Diamond Films 273
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
0.1
0.0
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
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 .
−
274 Renato Potenza and Cristina Tuvé
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
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.
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
276 Renato Potenza and Cristina Tuvé
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
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)
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
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)
278 Renato Potenza and Cristina Tuvé
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 .
±
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
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.
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
10
5 a
b
0
0 10 20 30 40 50 60 70 80 90 100
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.
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
284 Renato Potenza and Cristina Tuvé
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
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:
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.
Defect Density Measurements Inside CVD Diamond Films 285
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
[10] M. Marinelli, E. Milani, A. Paoletti, A. Tucciarone, G. V. Rinati, M. Angelone,
M. Pillon: Diamond Relat. Mater. 10, 645 (2001) 275
286 Renato Potenza and Cristina Tuvé
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
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
G. Messina, S. Santangelo (Eds.): Carbon, The Future Material for Advanced Technology Ap-
plications, Topics Appl. Phys. 100, 287–302 (2006)
© Springer-Verlag Berlin Heidelberg 2006
288 Enza Fazio et al.
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.
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].
Laser Ablation-Deposited CNx Thin Films 291
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
292 Enza Fazio et al.
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
(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:
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
Laser Ablation-Deposited CNx Thin Films 293
PN =0.1 Pa PN =6.5 Pa
2 2
Intensity (arb. un.)
P =66.7 Pa
N
PN =17 Pa 2
2
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
[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
Laser Ablation-Deposited CNx Thin Films 295
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
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
296 Enza Fazio et al.
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
Fig. 7. Plot of the distance vs. time in dimensionless variables for pure nitrogen
atmosphere and for the nitrogen/argon gaseous mixture case (inset)
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
300 Enza Fazio et al.
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.
Laser Ablation-Deposited CNx Thin Films 301
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
302 Enza Fazio et al.
Index
sp3 /sp2 bonding ratio, 293, 295 G band, 293, 294, 298
G. Messina, S. Santangelo (Eds.): Carbon, The Future Material for Advanced Technology Ap-
plications, Topics Appl. Phys. 100, 303–327 (2006)
© Springer-Verlag Berlin Heidelberg 2006
304 Stefano Lagomarsino and Silvio Sciortino
1.1 Band A
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.
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 :
Modeling of the Transport Properties of Diamond Radiation Sensors 307
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
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
Modeling of the Transport Properties of Diamond Radiation Sensors 309
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]
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.
Modeling of the Transport Properties of Diamond Radiation Sensors 311
Fig. 5. Schematic plot of the trap parameters of the samples under study, deduced
from the analysis of the radiation-induced currents
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:
Modeling of the Transport Properties of Diamond Radiation Sensors 315
(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.
316 Stefano Lagomarsino and Silvio Sciortino
s = Nv σvp . (10)
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:
Modeling of the Transport Properties of Diamond Radiation Sensors 317
-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
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
318 Stefano Lagomarsino and Silvio Sciortino
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:
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
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.
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
320 Stefano Lagomarsino and Silvio Sciortino
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
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
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
324 Stefano Lagomarsino and Silvio Sciortino
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
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
326 Stefano Lagomarsino and Silvio Sciortino
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
Modeling of the Transport Properties of Diamond Radiation Sensors 327
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
G. Messina, S. Santangelo (Eds.): Carbon, The Future Material for Advanced Technology Ap-
plications, Topics Appl. Phys. 100, 329–343 (2006)
© Springer-Verlag Berlin Heidelberg 2006
330 Giuliana Faggio et al.
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
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.
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
334 Giuliana Faggio et al.
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.
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
Fig. 5. Micro-PL spectra measured at the growth surface of two continuous sam-
ples, grown at [CH4 ] = 2% and [CH4 ] = 3%
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)
338 Giuliana Faggio et al.
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:
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
Nucleation of CVD Diamond on Mo Substrates 339
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.
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
Nucleation of CVD Diamond on Mo Substrates 341
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
342 Giuliana Faggio et al.
[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
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-
G. Messina, S. Santangelo (Eds.): Carbon, The Future Material for Advanced Technology Ap-
plications, Topics Appl. Phys. 100, 345–358 (2006)
© Springer-Verlag Berlin Heidelberg 2006
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 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.
3 Results
3.1 Optical and SEM Characterisation
Fig. 1. Optical micrographs (×10 objective) of the samples studied in this work
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 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.
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
352 Maria G. Donato et al.
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
sample SCD5 using two different excitation wavelengths: 514.5 nm (black ) and
457.9 nm (light gray)
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
Optical Characterisation of High-Quality Homoepitaxial Diamond 355
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
sample SCD7 using two different excitation wavelengths: 514.5 nm (black ) and
457.9 nm (light gray)
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.
1
1st order x10
Raman peaks
Normalised PL intensity (a. u.)
2nd order
Raman bands
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
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
[1] M. G. Donato, G. Faggio, M. Marinelli, G. Messina, E. Milani, A. Paoletti,
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
358 Maria G. Donato et al.
[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
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
G. Messina, S. Santangelo (Eds.): Carbon, The Future Material for Advanced Technology Ap-
plications, Topics Appl. Phys. 100, 359–380 (2006)
© Springer-Verlag Berlin Heidelberg 2006
360 Paolo M. Ossi and Antonio Miotello
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.
2 Experimental Details
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
362 Paolo M. Ossi and Antonio Miotello
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.
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
Pulsed Laser Deposition of Carbon Films 363
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
Pulsed Laser Deposition of Carbon Films 365
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
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
Pulsed Laser Deposition of Carbon Films 367
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.
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.
Pulsed Laser Deposition of Carbon Films 369
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
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
4 Discussion
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
Pulsed Laser Deposition of Carbon Films 375
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
376 Paolo M. Ossi and Antonio Miotello
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.
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
378 Paolo M. Ossi and Antonio Miotello
Acknowledgements
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
[10] C. Mapelli, C. Castiglioni, G. Zerbi, K. Müllen: Phys. Rev. B 60, 12710 (1999)
363
Pulsed Laser Deposition of Carbon Films 379
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
380 Paolo M. Ossi and Antonio Miotello
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,
G. Messina, S. Santangelo (Eds.): Carbon, The Future Material for Advanced Technology Ap-
plications, Topics Appl. Phys. 100, 381–403 (2006)
© Springer-Verlag Berlin Heidelberg 2006
382 Chiara Castiglioni et al.
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.
Raman Spectra and Structure of sp2 Carbon-Based Materials 385
(a)
(b)
ν = 1476 cm−1 I = 1.0 106 A4 /amu
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
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)
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.
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:
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
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
392 Chiara Castiglioni et al.
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]
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)
Raman Spectra and Structure of sp2 Carbon-Based Materials 395
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
396 Chiara Castiglioni et al.
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
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
[2] C. Mapelli, C. Castiglioni, G. Zerbi, K. Müllen: Phys. Rev. B 60, 12710 (1999)
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)
Raman Spectra and Structure of sp2 Carbon-Based Materials 401
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
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
G. Messina, S. Santangelo (Eds.): Carbon, The Future Material for Advanced Technology Ap-
plications, Topics Appl. Phys. 100, 403–421 (2006)
© Springer-Verlag Berlin Heidelberg 2006
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].
2 Experimental Details
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
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
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
406 Leonid Khriachtchev
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
Raman Spectroscopy and Optical Properties of DLC Films 407
(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
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
= 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.
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
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
412 Leonid Khriachtchev
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
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
Raman Spectroscopy and Optical Properties of DLC Films 413
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-
414 Leonid Khriachtchev
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
Raman Spectroscopy and Optical Properties of DLC Films 415
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
416 Leonid Khriachtchev
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
Raman Spectroscopy and Optical Properties of DLC Films 417
2
1580
(a)
1560 1
1540
G-band position (cm )
-1
(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-
418 Leonid Khriachtchev
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.
Raman Spectroscopy and Optical Properties of DLC Films 419
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
420 Leonid Khriachtchev
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
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
1 Introduction
Restricting (2) to first-order terms, combining (1) with (2) and applying
trigonometric identities, we get:
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
K2f,10 = eL · R · eS .
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
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
428 Margit Koós, Miklós Veres, Sára Tóth, and Miklós Füle
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
Raman Spectroscopy of CVD Carbon Thin Films 429
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.
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
430 Margit Koós, Miklós Veres, Sára Tóth, and Miklós Füle
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
Raman Spectroscopy of CVD Carbon Thin Films 431
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
432 Margit Koós, Miklós Veres, Sára Tóth, and Miklós Füle
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
Raman Spectroscopy of CVD Carbon Thin Films 433
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
434 Margit Koós, Miklós Veres, Sára Tóth, and Miklós Füle
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
Raman Spectroscopy of CVD Carbon Thin Films 435
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-
436 Margit Koós, Miklós Veres, Sára Tóth, and Miklós Füle
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
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.
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
Raman Spectroscopy of CVD Carbon Thin Films 439
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.
440 Margit Koós, Miklós Veres, Sára Tóth, and Miklós Füle
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
Raman Spectroscopy of CVD Carbon Thin Films 441
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
442 Margit Koós, Miklós Veres, Sára Tóth, and Miklós Füle
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
Raman Spectroscopy of CVD Carbon Thin Films 443
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
444 Margit Koós, Miklós Veres, Sára Tóth, and Miklós Füle
[14] A. C. Ferrari, J. Robertson: Phys. Rev. B 61, 14095 (2000) 427, 428, 429
[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
[28] C. Mapelli, C. Castiglioni, G. Zerbi, K. Müllen: Phys. Rev. B 60, 12710 (2000)
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
[37] A. C. Ferrari, J. Robertson: Phys. Rev. B 63, 121405 (2001) 437
[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
Raman Spectroscopy of CVD Carbon Thin Films 445
1 Introduction
2 Experimental
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.
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
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
452 Lucia Calliari et al.
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].
454 Lucia Calliari et al.
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.
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
The Role of Hydrogen in Amorphous Carbon 457
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
458 Lucia Calliari et al.
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
The Role of Hydrogen in Amorphous Carbon 459
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-
460 Lucia Calliari et al.
Fig. 7. π plasmon peak (EEL spectrum) at four selected temperatures. The zero-
loss peaks are normalized to a common height
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
462 Lucia Calliari et al.
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
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
G. Messina, S. Santangelo (Eds.): Carbon, The Future Material for Advanced Technology Ap-
plications, Topics Appl. Phys. 100, 463–504 (2006)
© Springer-Verlag Berlin Heidelberg 2006
464 Emanuele Pace et al.
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
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
UV-Induced Photoconduction in Diamond 467
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
468 Emanuele Pace et al.
4 Synthesis of Diamond
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
470 Emanuele Pace et al.
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
UV-Induced Photoconduction in Diamond 471
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].
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
UV-Induced Photoconduction in Diamond 473
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
E
Qc = Qph µτ , (2)
d
and, correspondingly
Qc E
Id = = Iph µτ . (3)
t d
ηp E
η = EQE = Iph µτ = η0 G. (4)
Ip d
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.
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
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
transverse
co-planar
Photocurrent (nA)
10 E = 2.5x104 V/cm
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)
@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
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
-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
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
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
482 Emanuele Pace et al.
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
UV-Induced Photoconduction in Diamond 483
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].
UV-Induced Photoconduction in Diamond 485
7 Single-Pixel Detectors
7.1 Photodetectors
7.2 Photoconductor
dN N
= G(t) .
dt τ
UV-Induced Photoconduction in Diamond 487
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)
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.
488 Emanuele Pace et al.
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
UV-Induced Photoconduction in Diamond 489
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.
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.
UV-Induced Photoconduction in Diamond 491
Fig. 13. The absolute quantum efficiency of diamond detectors compared with
CCD and MCP detectors
9 Applications
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
9.1 UV Lasers
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-Induced Photoconduction in Diamond 495
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.
498 Emanuele Pace et al.
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
UV-Induced Photoconduction in Diamond 499
[47] E. Pace: Recent Res. Devel. Optical Eng. 2, 317 (1999) 466
[48] M. Marinelli, E. Milani, E. Pace, A. Paoletti, M. Santoro, S. Sciortino, A. Tuc-
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
[63] M. D. Whitfield, S. P. Lansley, O. Gaudin, R. D. McKeag, N. Rizvi,
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
UV-Induced Photoconduction in Diamond 501
[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
[91] M. G. Donato, G. Faggio, M. Marinelli, G. Messina, E. Milani, A. Paoletti,
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
[95] M. Marinelli, E. Milani, E. Pace, A. Paoletti, M. Santoro, S. Sciortino, A. Tuc-
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
[99] M. D. Whitfield, S. P. Lansley, O. Gaudin, R. D. McKeag, N. Rizvi,
R. B. Jackman: Diam. Relat. Mater. 10, 650 (2001) 478, 493
[100] M. D. Whitfield, S. P. Lansley, O. Gaudin, R. D. McKeag, N. Rizvi,
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
502 Emanuele Pace et al.
[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
UV-Induced Photoconduction in Diamond 503
[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
504 Emanuele Pace et al.
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
G. Messina, S. Santangelo (Eds.): Carbon, The Future Material for Advanced Technology Ap-
plications, Topics Appl. Phys. 100, 505–520 (2006)
© Springer-Verlag Berlin Heidelberg 2006
506 Giuseppe Compagnini et al.
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
Vibrational Spectroscopy in Ion-Irradiated Carbon-Based Thin Films 509
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-
510 Giuseppe Compagnini et al.
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
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
1
1 ≡ C⎯H sp
2 1000 eV/C-atom
C⎯H sp
G
0 D
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
516 Giuseppe Compagnini et al.
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
Vibrational Spectroscopy in Ion-Irradiated Carbon-Based Thin Films 517
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
518 Giuseppe Compagnini et al.
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
Vibrational Spectroscopy in Ion-Irradiated Carbon-Based Thin Films 519
Index
a-C, 506 carbyne, 515
a-Si1−x Cx , 508 clustering, 510
sp1 hybridisation, 514 cross-links, 510, 511, 514
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
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