You are on page 1of 10

Diamond and Related Materials 12 (2003) 47–56

Stress and structural properties of diamond-like carbon films deposited by


electron beam excited plasma CVD
Masahito Bana,*, Takeshi Hasegawab, Sadao Fujiia, Junzo Fujiokaa
a
Technical Institute, Kawasaki Heavy Industries, Ltd., 118 Futatsuzuka, Noda, Chiba 278-8585, Japan
b
FCT Research Department, Japan Fine Ceramics Center (JFCC), FCT Laboratory, AIST Tsukuba Central 5,
National Institute of Advanced Industrial Science and Technology (AIST), 1-1 Higashi 1-Chome, Tsukuba, Ibaraki 305-8565, Japan

Received 17 October 2001; received in revised form 15 November 2002; accepted 18 November 2002

Abstract

Diamond-like carbon (DLC) films prepared using CH4 or C6H6 with varying deposition parameters by an electron beam
excited plasma CVD system were investigated for the internal stress, dynamic hardness and structural properties such as the film
density, total, bonded and unbound hydrogen contents, sp3 ratio and graphite crystallite. From the correlations between internal
stress and structural properties, the following conclusions were derived. The fraction of unbound hydrogen to total hydrogen
content was the most influential factor for the compressive stress of the DLC films deposited from CH4 . It is suggested that
unbound hydrogen may be trapped into the disordered microstructure of graphite crystallites embedded in the network of film.
For the DLC films deposited from C6H6, it was shown that the compressive stress was correlated with not only the fraction of
unbound hydrogen content but also the degree of cross-linking between graphite crystallites in the film.
䊚 2002 Elsevier Science B.V. All rights reserved.

Keywords: CVD; Diamond-like carbon; Stress; Hydrogen; Microstructure

1. Introduction In the case of hydrogen-free amorphous carbon films,


the compressive stress increases almost linearly with
Diamond-like carbon (DLC) films, consisting of car- increasing the fraction of sp3 bonded carbon w6,7x. The
bon and hydrogen, demonstrate excellent tribological mechanism of generating the compressive stress can be
properties such as high hardness, self-lubricating and explained by a subplantation model proposed by Rob-
anti-galling ability, but it is difficult to use DLC films ertson w8x and Davis w9x, where medium-energy ions
under severe lubricating conditions unless the durability cause a quenched-in strain and density increment by
is improved. That is, it is essential for DLC films to directly entering into a interlayer region or displacing a
perform the high toughness and high adhesion to steel- surface atom into an interstitial site in the interlayer
based materials, and to increase the film thickness. The region. The increased stress and density result in the
principal cause of deteriorating the durability of DLC transformation of the stable sp2 bonding into the meta-
films is its high compressive stress reported to reach stable sp3 bonding. On the other hand, an internal stress
values of several gigapascals w1,2x. Techniques to depos- of (hydrogenated) DLC films must be discussed consid-
it DLC films with low compressive stresses have been ering the effects of not only C–C bonds but also C–H
developed by means of containing additional elements, bonds and unbound hydrogen in the films. Moreover,
like nitrogen and silicon, into the film w3–5x. However, Raman spectra of DLC films predict the existence of
in spite of such successes in the deposition methods, graphite crystallites embedded into the network of car-
little is known about the origins of compressive stress bon and hydrogen w10x, which can be taken into consid-
in DLC. eration for discussing the internal stress. The previous
*Corresponding author. Tel.: q81-4-7124-0302; fax: q81-4-7124- studies have pointed out that DLC films with lower
5917. total hydrogen content and higher density were apt to
E-mail address: ban_m@khi.co.jp (M. Ban). indicate higher compressive stress and hardness w11,12x.

0925-9635/03/$ - see front matter 䊚 2002 Elsevier Science B.V. All rights reserved.
doi:10.1016/S0925-9635(02)00265-0
48 M. Ban et al. / Diamond and Related Materials 12 (2003) 47–56

While Grill and Patel investigated the stress of DLC Table 1


films deposited using various source gases and reported Deposition conditions
that the compressive stresses were directly correlated to No. Gas Negative bias voltage Distance Pressure
the relative fraction of unbound hydrogen and increased (V) (mm) (mTorr)
with increasing the fraction of unbound hydrogen inde-
1 CH4 0 395 6
pendently of the source gases w13x. In these previous 2 150 395 6
studies, the internal stress and hardness were investigat- 3 200 395 6
ed for the correlations with few structural properties, 4 300 395 6
and it is not clear how other structural properties can 5* 400 395 6
affect the internal stress. 6 400 195 6
7 400 295 6
In the present study, internal stress and dynamic 8 400 345 6
hardness are measured in the DLC films deposited by 9* 400 395 6
an electron beam excited plasma (EBEP) CVD system 10 400 395 4.5
w14,15x from CH4 or C6H6 under the several deposition 11* 400 395 6
conditions. The deposited films are evaluated for the 12 400 395 8
13 400 395 10
structural properties such as the film density, total,
bonded and unbound hydrogen contents, sp3 ratio and 14 C 6H 6 150 295 4.5
15 300 295 4.5
graphite crystallite. The origins of internal stress gener- 16 400 295 4.5
ated in the DLC films are discussed based on the 17** 700 295 4.5
correlations between internal stress and structural 18 700 245 4.5
properties. 19** 700 295 4.5
20 700 345 4.5
21 700 395 4.5
2. Experimental details 22 700 295 3
23** 700 295 4.5
The DLC films discussed in the present work were 24 700 295 6
deposited by the EBEP-CVD system w14x. A Si (1 0 0) 25 700 295 8
substrate with 2 inch in diameter and 300–425 mm The symbol * or ** represents the same sample.
thickness was fixed on the holder without water-cooling
and placed parallel to the direction of an electron beam
as shown in a previous paper w14x. A negative bias was enough to be measured accurately, were obtained. In the
applied to the substrate using a 2-MHz radio frequency present study, the uncertainty of measurement for the
generator. stress values was negligibly small. The measured cur-
Table 1 summarizes the deposition conditions. CH4 vature of the substrate arises from both the intrinsic
or C6H6 was used as the source gas. The negative bias stress of DLC film and the stress generated by a
voltage to the substrate, distance of the substrate from mismatch in thermal expansion between the DLC and
an electron beam and pressure during deposition were the substrate material. However, the thermal expansion
taken as the deposition parameters, and varied for each coefficient of DLC was reported to be 2.3 ppm 8Cy1,
gas as shown in Table 1. The DLC films with the which was the same range as that of Si (1 0 0) substrate
thicknesses of 340–810 nm were deposited on the w17x. This can be confirmed from the experimental
substrates. The film thickness was determined from results that when the stresses of DLC films on Si
profilometric measurements of a step formed during substrates were measured at the range of room temper-
deposition. Before deposition, the substrate was cleaned ature to 200 8C, little change in the measured stresses
by accelerated ions in an argon plasma for 10 min. was observed w17,18x. Therefore, the stress by the
The internal stresses of the films were calculated from difference in thermal expansion can be ignored, and the
the curvature of Si substrate and the DLC film thickness measured stresses are assumed to be the intrinsic stresses
using the well-known Stoney’s formula defined by of DLC films.
Dynamic hardness measurements were carried out
SsdEsds2y3r2(1yn)df, (1) using a nano-indentation tester (Elionix ENT-1100) with
a load of 40 mgf. Loading and unloading curves vs. the
where S is the internal stress, d the bending of the depth of indentation were obtained. Five separate meas-
substrate, n the Poisson’s ratio of the substrate, r the urements were performed on different locations, and the
radius of the substrate, Es the Young’s modulus of the results represent an average of these measurements.
substrate, and ds and df the thicknesses of the substrate Hydrogen forward scattering spectrometryyelastic
and the film, respectively. The values of 1.13=105 MPa recoil detection (HFSyERDA) was performed using a
and 0.42 were adopted for Es and n, respectively w16x. Heq ion beam with an energy of 2.275 MeV to estimate
The bending of the substrate was measured by a profi- the total hydrogen content in the DLC film, wHxtotal, and
lometer, and the values of 8.2–37.6 mm, being large the film density. For the calculation of the DLC film
M. Ban et al. / Diamond and Related Materials 12 (2003) 47–56 49

density, the bulk densities of 1.12=1023 and


1.30=1023 atoms cmy3 were used for carbon and hydro-
gen, respectively.
Fourier transform infra red (FT-IR) analysis was
carried out using a Perkin Elmer Spectrum GX system
in the 400–4000 cmy1 range to investigate the chemical
bonding. An average of 100 scans was taken for each
sample. The content of hydrogen bonded to carbon
atom, wHxbond, was estimated from the total areas of the
infra red absorption in C–H bonds centered at 2900
cmy1 in the FT-IR spectra w19x using the following
formula w11x,
wHxbondsAyŽŽ1.68=105.sq0.916A. , (2)
where A is the area of the absorption band in square
centimeter and s is the film density in grams per cubic
centimeter, which is the value obtained from the HFS
analysis. The infra red absorption band at 2900 cmy1
corresponding to the C–H stretching vibration was also
used to estimate the sp3 ratio in C–H bonds. The de-
convolution of the spectra was performed using the
individual C–H vibrations, CH2 sp2 (olefinic) at 3020
cmy1, CH sp2 (olefinic) at 3000 cmy1, CH3 sp3 (asym-
metrical) at 2965 cmy1, CH sp3 at 2920 cmy1, CH2
sp3 (asymmetrical) at 2920 cmy1, CH3 sp3 (symmetri-
cal) at 2870 cmy1 and CH2 sp3 (symmetrical) at 2850
cmy1 w11,19x. The areas of absorption coefficients were
calculated for both C–H bonds with sp3 carbon and C–H
bonds with sp2 carbon. The sp3 ratio in C–H bonds was Fig. 1. The variations in the compressive stress as functions of (a)
defined as the ratio of the areas of C–H bonds with negative bias voltage; (b) distance from an electron beam and (c)
pressure during deposition. The open and solid circles represent the
sp3 carbon to the total areas of C–H bonds.
films from CH4 and C6H6, respectively.
Raman spectroscopy was carried out using an argon
ion laser with a wavelength of 514.5 nm in the 900–
1800 cmy1 range. The intensity ratio of D-peak to G- distance from an electron beam and pressure during
peak, I(D)yI(G), in broad bands observed from 1200 to deposition, respectively. The results are plotted for the
1600 cmy1 in Raman spectra was calculated w19x. It is DLC films deposited using CH4 or C6H6. All films
well known that G-peak and D-peak correspond to indicated the compressive stresses and the obtained
spectra of sp2 bonded carbon in graphite crystals and values were in the same range as the previous studies
defect-induced disorders in the microcrystalline graphite, w1–4,11,18x. Regarding the DLC films from CH4, the
respectively. compressive stress had a peak value of 3.9 GPa at the
By electron energy loss spectroscopy (EELS), the negative bias voltage of 200 V (Fig. 1a), while it
fraction of sp3 bonded carbon in the DLC film was increased with both the distance (Fig. 1b) and the
calculated from the 280–360 eV region of carbon K pressure (Fig. 1c). On the other hand, the DLC films
edge electron-energy-loss spectra for four samples of from C6H6 showed a monotonous increase in compres-
No. 2, 5, 14 and 17 shown in Table 1. The carbon K sive stress with increasing the negative bias voltage
edge contains information on both the s (sp3) and the (Fig. 1a) and decreasing both the distance (Fig. 1b) and
p (sp2) bonding in the films. The EELS analysis was the pressure (Fig. 1c). In this matter, it is noticed that
performed on a H-9000UHR (Hitachi) scanning trans- the dependences of compressive stress on deposition
mission electron microscope with an EELS spectrometer parameters differ strongly by the precursor. In an EBEP-
(Gatan, PEELS Model 676). CVD system, a higher degree of gas decomposition can
be provided at a smaller distance from an electron beam,
3. Results and discussion a lower pressure and a higher negative bias voltage.
3.1. Correlations between deposition parameters and Taking this into consideration, as an overall trend, the
film properties DLC films from CH4 indicated lower compressive stress
when deposited under a higher degree of decomposition
Fig. 1a–c shows the variations in the measured of the precursor in the plasma and on the substrate
internal stresses as functions of negative bias voltage, surface. In contrast, when C6H6 was used, the higher
50 M. Ban et al. / Diamond and Related Materials 12 (2003) 47–56

sure during deposition are shown in Fig. 4a–e; Fig. 5a–


e and; Fig. 6a–e, respectively.
From Figs. 4a, 5a and 6a, it is found that the film
density was especially affected by both the negative bias
voltage and the difference in source gas. The film
deposited with the bias voltage of 0 V indicated the
density of 0.9 g cmy3 (Fig. 4a), while other films
prepared under applying a negative bias voltage had the
density of more than 1.3 g cmy3. The increase in film
density with applying a negative bias voltage can be
explained from the replacement of hydrogen atoms and
the condensation of carbon atoms due to ion bombard-
ment. The films obtained from C6H6 had the density of
1.4–1.8 g cmy3, which were larger than those from
CH4; 0.9–1.5 g cmy3. Concerning wHxtotal and wHxbond,
as well as the film density, the negative bias voltage
and the source gas were the most influential parameters
as shown in Figs. 4b, 5b, 6b, 4c, 5c and 6c. For the
films from CH4, increasing the negative bias voltage
from 0 to 150 V, wHxtotal and wHxbond rapidly decreased
from 69 and 51% to less than 50 and 20%, respectively.
The films from C6H6 showed reductions by 36 and 65%
in wHxtotal and wHxbond while the negative bias voltage
increased from 150 to 700 V. In the case of depositions
with applying the negative bias voltage, the DLC films
from CH4 showed wHxtotal of 36–48% and wHxbond of
15–20%, and when C6H6 was used, wHxtotal and wHxbond
were in the range of 25–40% and 4–18%, respectively.
Fig. 2. The variations in the dynamic hardness as functions of (a) A value subtracted bonded hydrogen content from total
negative bias voltage; (b) distance from an electron beam and (c) hydrogen content, wHxtotalywHxbond , means the content
pressure during deposition. The open and solid circles represent the
films from CH4 and C6H6, respectively.
of unbound hydrogen, wHxunbound, which is considered to
be not directly bonded to carbon atom. This unbound
hydrogen is not infrared active so that it cannot be
degree of decomposition of the precursor made the detected by FT-IR analysis. The comparisons between
compressive stress higher. It seems to arise from the Figs. 4b, 5b, 6b, 4c, 5c and 6c imply that the deposited
difference in the bonding structure of two precursors, DLC films contained a large amount of the unbound
that is, CH4 molecule consists of sp3 bonded carbon and hydrogen. The DLC films from CH4 or C6H6 contained
hydrogen atoms while C6H6 molecule sp2 bonded carbon wHxunbound of 17–30% or 21–27%, respectively. These
and hydrogen atoms.
Fig. 2a–c shows the dynamic hardnesses of the DLC
films from CH4 or C6H6 as functions of the deposition
parameters similar to Fig. 1. Comparing Fig. 2 with Fig.
1, the variations of the dynamic hardness demonstrated
the same tendency as those of the compressive stress.
The correlations between dynamic hardness and com-
pressive stress are shown in Fig. 3 for the DLC films
deposited from CH4 or C6H6. It is observed from the
figure that both of the compressive stresses increased
linearly when the dynamic hardnesses increased, but
each slope considerably differed by the source gas. This
result implies that the deposited DLC films in the present
work had the different microstructure in accordance with
the difference in precursor.
The variations in the film density, wHxtotal, wHxbond, Fig. 3. The correlations between dynamic hardness and compressive
sp3 ratio in C–H bonds and I(D)yI(G) with the negative stress. The open and solid circles represent the films from CH4 and
bias voltage, distance from an electron beam and pres- C6H6, respectively.
M. Ban et al. / Diamond and Related Materials 12 (2003) 47–56 51

to those of wHxbond as functions of the deposition


parameters. That is, an increase in the sp3 ratio in C–H
bonds was accompanied with an increase in the content
of hydrogen connected to carbon atom. The result can
be explained that when an sp2 bonded carbon is trans-
formed into an sp3 bonded carbon, the remained dan-
gling bond is terminated with hydrogen atom. In the
case of the film with the sp3 ratio in C–H bonds of
100% (Fig. 4d), it is considered that the majority of
carbon atom is connected with hydrogen atom. Taking

Fig. 4. The variations in (a) film density; (b) wHxtotal ; (c) wHxbond; (d)
sp3 ratio in C–H bonds and (e) I(D)yI(G) as a function of the neg-
ative bias voltage. The open and solid circles represent the films from
CH4 and C6H6, respectively. The sample * indicated relatively high
value, but the cause has been unclear.

results are in agreement on the previous studies w13x.


The film prepared from CH4 with 0 V bias included a
considerable amount of the total hydrogen, 68%, and its
majority was the bonded hydrogen.
It is seen in Figs. 4d, 5d and 6d that the obtained
values of sp3 ratio in C–H bonds were more than 70%,
and carbon atoms connected to hydrogen atoms in the Fig. 5. The variations in (a) film density; (b) wHxtotal ; (c) wHxbond; (d)
sp3 ratio in C–H bonds and (e) I(D)yI(G) as a function of the distance
films were mainly hybridized in sp3 bonding. Comparing from an electron beam. The open and solid circles represent the films
Figs. 4d, 5d, 6d with Figs. 4c, 5c and 6c, the variations from CH4 and C6H6, respectively. The sample * indicated relatively
of sp3 ratio in C–H bonds revealed similar tendencies high value, but the cause has been unclear.
52 M. Ban et al. / Diamond and Related Materials 12 (2003) 47–56

Therefore, it can be noticed that the sp3 ratio of carbon


without protonation is much smaller than that of carbon
with protonation (sp3 ratio in C–H bonds), which
implies that a large number of sp2 bonded carbon sites
exist in the DLC film. Robertson proposed the cluster
model and its modification, and predicted that sp2 sites
form clusters embedded in an sp3 bonded matrix and
the sp2 clusters are limited to single sixfold rings and
short chains of sp2 sites w20,21x.
The presence of sp2 clusters may be supported by the
observation of D-peak corresponding to spectra of dis-
ordered microcrystalline graphite. Tuinstra and Koenig
reported that I(D)yI(G) calculated from Raman spectra
was inversely proportional to the size of graphite crys-
tallite w22x. Figs. 4e, 5e and 6e show that the DLC films
from CH4 or C6H6 had almost the same range of I(D)y
I(G) and the similar tendency as functions of the
investigated deposition parameters. Namely, the size of
graphite crystallites decreased with increasing the neg-
ative bias voltage, and decreasing the distance from an
electron beam. These trends imply that graphite crystal-
lites become finer, when ions with higher energy are
bombarded at the surface of growing film, and also a
film is deposited under a higher degree of gas
decomposition.
The ratios of sp3 to sp2 bonded carbon, sp3 ysp2,
obtained from EELS analysis are plotted as a function
of wHxtotal for the samples of No. 2, 5, 14 and 17, as
given in Fig. 7. The solid line in the figure presents the
following equation,

sp3ysp2sŽ6wHxtotaly1.yŽ8y13wHxtotal., (3)

and represents the ideal relationship between sp3 ysp2


and hydrogen content predicted by the fully constrained
network (FCN) model proposed by Angus and Jansen
w23x. It was reported that this equation was obtained
based on the theories of random covalent networks,
assuming that DLC had a completely constrained ran-
dom amorphous network. It is notable from Fig. 7 that
Fig. 6. The variations in (a) film density; (b) wHxtotal ; (c) wHxbond; (d)
sp3 ratio in C–H bonds and (e) I(D)yI(G) as a function of the pressure
during deposition. The open and solid circles represent the films from
CH4 and C6H6, respectively. The sample * indicated relatively high
value, but the cause has been unclear.

the value of density, 0.92 g cmy3, into account, the film


seems to have a structure near polyethylene. The values
of sp3 ratio obtained from FT-IR analysis do not include
any useful information of carbon without protonation.
On the other hand, EELS analysis can provide some
information of not only carbon with protonation but also
carbon without protonation. From EELS, the sp3 ratio
of 67, 71, 61 and 54% were obtained for No. 2, 5, 14
and 17 as shown in Table 1, respectively, and those Fig. 7. The variations in the sp3ysp2 ratio as a function of wHxtotal. The
were less than the values obtained from FT-IR analysis. line is Eq. (3).
M. Ban et al. / Diamond and Related Materials 12 (2003) 47–56 53

Fig. 8. The variations in the compressive stress as a function of the Fig. 9. The variations in the compressive stress as a function of
film density. The open and solid circles represent the films from wHxtotal. The open and solid circles represent the films from CH4 and
CH4 and C6H6, respectively. C6H6, respectively.

the values of sp3 ysp2 obtained from the films were


stress (and hardness, see Fig. 3), which is consistent
higher than those predicted by the theory. This means
with the results reported in the previous studies w11,12x.
that the DLC films in the present work have the
However, in regard to the DLC films from CH4, the
overconstrained structure, which can be commonly char-
compressive stress varied largely in spite of little varia-
acterized in most DLC films reported w24,25x. It is
tions of both the film density and total hydrogen content,
explained in Ref. w23x that one of the reasons why the
and it seems to be attributed to other structural
random covalent network of DLC is overconstrained is
properties.
that the medium range order such as graphite crystallites
The compressive stresses are plotted as a function of
is formed in the network. As mentioned above, the idea
the fraction of wHxunbound to wHxtotal in Fig. 10. The figure
is considered to be consistent with the results obtained
shows the obvious relationship and the compressive
from our DLC films, and will be adopted for the
stresses increased almost linearly with increasing the
following discussion.
fraction of wHxunbound to wHxtotal for the DLC films from
CH4 or C6H6. Grill and Patel reported that the values of
3.2. Correlations between internal stresses and structur-
compressive stress were 0.84 and 1.4 GPa in the case
al properties
of the fractions of unbound hydrogen of 0.44 and 0.58
All data obtained in the present work are used for the for DLC films deposited from CH4, respectively w13x,
following discussion to investigate the correlations which is in good agreement with our results. Also, Ref.
w13x described that the compressive stresses of DLC
between internal stresses and structural properties.
Fig. 8 shows the variations in compressive stresses as films prepared from various precursors such as CH4,
a function of the film density. It is observed that the C2H2, C6H12 and C6H12qH2 increased almost linearly
compressive stress of DLC films from C6H6 increased as a function of the fraction of unbound hydrogen.
linearly from 0.8 to 2.5 GPa while the film density
increased from 1.4 to 1.8 g cmy3. On the other hand,
the DLC films from CH4 had almost the same range of
density, but demonstrated a wide range of compressive
stresses from 1.6 to 3.9 GPa (not mentioned for the film
with the density of 0.9 g cmy3 because of not having
the structure of DLC).
Fig. 9 shows the variations in compressive stresses as
a function of wHxtotal. As described above, an increase
in wHxtotal was accompanied with a decrease in the film
density. Accordingly, the compressive stress of DLC
films from C6H6 decreased monotonously with increas-
ing wHxtotal, while concerning the DLC films from CH4,
the compressive stress was independent of wHxtotal. From
Figs. 8 and 9, it is found that for the DLC films Fig. 10. The variations in the compressive stress as a function of the
deposited using C6H6, the films with higher density and fraction of wHxunbound to wHxtotal. The open and solid circles represent
lower total hydrogen content had higher compressive the films from CH4 and C6H6, respectively.
54 M. Ban et al. / Diamond and Related Materials 12 (2003) 47–56

Taking into consideration that I(D)yI(G) is inversely


proportional to the size of graphite crystallite in the
network of film, when CH4 was used for the deposition
of DLC films, the films including the larger graphite
crystallites showed higher compressive stresses. On the
contrary, in the case of the DLC films deposited from
C6H6, when the finer graphite crystallites were included,
the compressive stresses were higher. The previous
studies have suggested both cases that the compressive
stress (or hardness) decreased or increased with increas-
ing I(D)yI(G) w10,26x. Fig. 12b shows the variations of
the fraction of wHxunbound to wHxtotal as a function of
I(D)yI(G). From the figure, as for the films from CH4,
Fig. 11. The variations in the compressive stress as functions of the
the fraction of wHxunbound to wHxtotal increased with
sp3 ratio in C–H bonds by FT-IR. The open and solid circles represent decreasing I(D)yI(G). In other words, an increase in
the films from CH4 and C6H6, respectively. the size of graphite crystallites was accompanied with
an increase in the fraction of unbound hydrogen content,
which means the compressive stress increment as already
However, the results in the present work displayed that mentioned in Fig. 10. In the case of the DLC films
the absolute values of compressive stress at the same from C6H6, it can be explained in the same fashion as
fraction of wHxunbound to wHxtotal indicated the discrepan- the films from CH4 that the compressive stress increased
cies by the difference in precursor. Hence, it means that
a fraction of wHxunbound to wHxtotal is not the only factor
to determine the internal stress of any DLC films
deposited from several precursors.
The compressive stresses in the films prepared from
CH4 or C6H6 are presented in Fig. 11 as a function of
the sp3 ratio in C–H bonds obtained from FT-IR. As can
be seen in Fig. 11, the compressive stresses were apt to
decrease with increasing the sp3 ratio in C–H bonds.
The present tendency is contrary to that for hydrogen-
free amorphous carbon, where the compressive stress
increased with increasing the sp3 ratio w6,7x. This fact
is related to the existence of C–H bonds. That is, for
the DLC films (hydrogenated amorphous carbon films),
an increase in the sp3 ratio in C–H bonds was accom-
panied with an increase in hydrogen atom connected to
carbon atom (Figs. 4c and d, 5c and d, 6c and d), and
it is considered that the network in film is segmented
by the termination with hydrogen atoms so that the
compressive stress is released. This can be supported by
FCN model. On the other hand, a hydrogen-free amor-
phous carbon film has a more rigid structure by forming
a larger number of C–C single bonds than C–C double
bonds. It should be noticed that in the case of the films
from CH4, the scatter was considerably large and for
example, the obtained compressive stresses were in the
range from 2 to 4 GPa at the sp3 ratio in C–H bonds of
approximately 80%. Therefore, it can not be considered
that the sp3 ratio in C–H bonds has a direct effect on
the compressive stress of the films from CH4.
Fig. 12a displays the correlations of the compressive
stresses with I(D)yI(G) calculated by Raman analysis.
Fig. 12. The variations in (a) the compressive stress; (b) the fraction
It is observed from the figure that the compressive stress of wHxunbound to wHxtotal and (c) wHxunbound as a function of I(D)yI(G).
decreased for the DLC films from CH4 and increased The open and solid circles represent the films from CH4 and C6H6,
for the DLC films from C6H6 as I(D)yI(G) increased. respectively.
M. Ban et al. / Diamond and Related Materials 12 (2003) 47–56 55

with the fraction of wHxunbound to wHxtotal (see both Figs. Summing up the correlations between compressive
12a and b or Fig. 10), but it is notable from Fig. 12b stress and structural properties from the above discus-
that the fraction of wHxunbound to wHxtotal decreased with sions, the following can be derived. First, as for the
decreasing I(D)yI(G), that is, increasing the size of DLC films deposited from CH4, the fraction of unbound
graphite crystallites. This indicates that an increase in hydrogen to total hydrogen content was the most influ-
the size of graphite crystallites coincides with an ential factor for the compressive stress. The film density
increase in the fraction of bonded hydrogen content. and total hydrogen content had little effect on the
That is to say, it is suggested that when the size of compressive stress. The DLC films including a larger
graphite crystallites increases, the termination with size of graphite crystallites, involving a decrease in the
hydrogen may proceed in the boundaries of graphite sp3 ratio in C–H bonds, demonstrated higher compres-
crystallites and the bonding between graphite crystallites sive stresses. It is highly probable that the unbound
and surrounding networks reduces. The presence of hydrogen is trapped into the disordered microstructure
disconnected graphite crystallites results in a decrease of graphite crystallites, and the insertion of unbound
in the compressive stress. When the segmentation of hydrogen can contribute to an increase in volume and
network occurs, the unbound hydrogen can give little the strains by the expansion induce a compressive stress.
effect on an increase in compressive stress. Second, concerning the DLC films deposited from
Finally, we will discuss where and how the unbound C6H6, the several factors were correlated with the
hydrogen exists in the DLC film. From I(D)yI(G) compressive stress. As well as the DLC films deposited
˚
values, the effective sizes of graphite crystallite, La (A), from CH4, the insertion of unbound hydrogen is sug-
can be estimated using the equation, gested to give a contribution to the compressive stress.
Moreover, it is found that the compressive stress
I(D)yI(G)s44yLa, (4) increased with increasing the film density, and decreas-
ing the total and bonded hydrogen contents, accompa-
obtained from the experimental results of Tuinstra and nied with a decrease in the size of graphite crystallites
Koenig w22x. La was found to be in the range of 30– and sp3 ratio in C–H bonds. This means that the
140 A ˚ for the present DLC films. These La values are compressive stress increment arises from the promotion
much larger than the cluster size predicted by the of cross-linking between graphite crystallites embedded
modified cluster model w21x. This may be comprehended into the three-dimensional network. Namely, it is con-
that the internal microstructure of effective cluster is in sidered that the change of sp2 sites into sp3 C–C bonds
disorder and the minimum size of sp2 sites is much by the cross-linking induces the expansion in film in
smaller than La. The presence of a broad D-peak in the accordance with an increase in bond length, which
Raman spectra of DLC films can support the existence results in an increase in the compressive stress.
of disordered microstructure in graphite crystallites.
Here, the unbound hydrogen contents are plotted as a 4. Conclusions
function of I(D)yI(G) in Fig. 12c. The figure obviously
shows that an increase in the unbound hydrogen content The DLC films prepared using CH4 or C6H6 with
was directly correlated with an increase in the size of varying the negative bias voltage, distance from the
graphite crystallites independently of the investigated electron beam and pressure during deposition by the
precursor. These facts imply that the location where the EBEP-CVD system are investigated for the internal
unbound hydrogen is incorporated is related to the stress, dynamic hardness and various structural proper-
disordered microstructure in graphite crystallites. More- ties. The major results are summarized as follows.
over, in Fig. 7, note that if the sp3 ysp2 values are plotted (1) As an overall trend, the DLC films from CH4
using, not the total hydrogen content including the indicated lower compressive stresses, when they were
unbound and bonded hydrogen content but the bonded deposited under a higher degree of decomposition of
hydrogen content, the experimental and theoretical val- the precursor in the plasma and on the substrate surface.
ues display larger discrepancies. This might indicate that In contrast, in the case of C6H6, a higher degree of
the unbound hydrogen is not completely independent of decomposition of the precursor made the compressive
a skeleton formed in the DLC film. It follows from the stress higher.
above discussion that the unbound hydrogen may be (2) The compressive stresses in DLC films are direct-
trapped into the disordered microstructure of graphite ly proportional to the dynamic hardness, but these slopes
crystallites, and ‘chemisorbed’ to carbon (not enough to considerably differed by the precursor.
absorb infrared light) rather than ‘perfectly unbound’. (3) The fraction of unbound hydrogen to total hydro-
To make sure of the location and the condition of gen content mostly affected the compressive stress of
unbound hydrogen, the further studies such as the direct the DLC films deposited from CH4. The DLC films
measurements of unbound hydrogen need to be carried including a larger size of graphite crystallites, accom-
out. panied with a decrease in the sp3 ratio in C–H bonds,
56 M. Ban et al. / Diamond and Related Materials 12 (2003) 47–56

demonstrated higher compressive stresses. These results w3x D.F. Franceschini, C.A. Achete, F.L. Freire Jr., Appl. Phys.
imply that the insertion of unbound hydrogen, suggested Lett. 60 (1992) 3229.
w4x W.C. Vassell, A.K. Gangopadhyay, T.J. Potter, M.A. Tamor,
to be into the disordered microstructure of graphite M.J. Rokosz, JMEPEG 6 (1997) 426.
crystallites, can contribute to an increase in volume, and w5x M. Ban, T. Hasegawa, Surf. Coat. Technol. 162 (2003) 1.
the strains by the expansion induce the compressive w6x S. Xu, D. Flynn, B.K. Tay, S. Prawer, K.W. Nugent, S.R.P.
stress. Silva, Y. Lifshitz, W.I. Milne, Philos. Mag. B 76 (1997) 351.
(4) Concerning the DLC films deposited from C6H6, w7x P.J. Fallon, V.S. Veerasamy, C.A. Davis, J. Robertson, G.A.J.
the insertion of unbound hydrogen is also suggested to Amaratunga, W.I. Milne, J. Koskinen, Phys. Rev. B 48 (1993)
4777.
give a contribution to the compressive stress. In addition, w8x J. Robertson, Diamond Relat. Mater. 2 (1993) 984.
an increase in compressive stress was correlated with an w9x C.A. Davis, Thin Solid Films 226 (1993) 30.
increase in the film density, and a decrease in the total w10x B. Marchon, N. Heiman, M.R. Khan, A. Lautie, J.W. Ager III,
and bonded hydrogen contents. It is considered that the D.K. Veirs, J. Appl. Phys. 69 (1991) 5748.
compressive stress increment is attributed to the pro- w11x P. Couderc, Y. Catherine, Thin Solid Films 146 (1987) 93.
w12x P. Koidl, C. Wild, B. Dischler, S. Wagner, M. Ramsteiner,
motion of cross-linking between graphite crystallites in
Mater. Sci. Forum 52 and 53 (1989) 41.
the three-dimensional network. This can be explained w13x A. Grill, V.V. Patel, Diamond Films Technol. 1 (1992) 219.
that the change of sp2 sites into sp3 C–C bonds by the w14x M. Ban, M. Ryoji, S. Fujii, J. Fujioka, New Diamond Front.
cross-linking induces the expansion in film in accor- Carbon Technol. 9 (1999) 425.
dance with an increase in bond length, which leads to w15x T. Hara, M. Hamagaki, A. Sanda, Y. Aoyagi, S. Namba, J.
an increase in the compressive stress. Vac. Sci. Technol. Part B 5 (1987) 366.
w16x J.W. Zou, K. Schmidt, K. Reichelt, B. Stritzker, J. Vac. Sci.
Technol. A 6 (1988) 3103.
Acknowledgments w17x I.A. Blech, P. Wood, J. Vac. Sci. Technol. A 11 (1993) 728.
w18x H. Yamada, O. Tsuji, P. Wood, Thin Solid Films 270 (1995)
The authors would like to express many thanks to Dr 220.
S. Nishijima for valuable advice on the procedure for w19x M. Ban, M. Ryoji, T. Hasegawa, Y. Mori, S. Fujii, J. Fujioka,
the present study and Mr N. Takahashi of Kawaju Diamond Relat. Mater. 11 (2002) 1353.
w20x J. Robertson, Phys. Rev. Lett. 68 (1992) 220.
Techno Service Corp. for discussions about the experi-
w21x J. Robertson, J. Non-Cryst. Solids 198–200 (1996) 614.
mental results. w22x F. Tuinstra, J.L. Koenig, J. Chem. Phys. 53 (1970) 1126.
w23x J.C. Angus, F. Jansen, J. Vac. Sci. Technol. A 6 (1988) 1778.
References w24x S. Kaplan, F. Jansen, M. Machonkin, Appl. Phys. Lett. 47
(1985) 750.
w1x K.J. Clay, S.P. Speakman, N.A. Morrison, N. Tomozeiu, W.I. w25x Y. Wang, H. Chen, R.W. Hoffman, J.C. Angus, J. Mater. Res.
Milne, A. Kapoor, Diamond Relat. Mater. 7 (1998) 1100. 5 (1990) 2378.
w2x R.G. Lacerda, F.C. Marques, Appl. Phys. Lett. 73 (1998) 617. w26x R.G. Lacerda, F.C. Marques, Appl. Phys. Lett. 73 (1998) 617.

You might also like