Full text of DOI:10.1557/JMR.2000.0017

Journal of
MATERIALS RESEARCH
Welcome
Comments
Help
Pyrocarbon anisotropy as measured by electron diffraction
and polarized light
a)
Xavier Bourrat, B e´ atrice Trouvat, Guillaume Limousin, and G e´ rard Vignoles
Laboratoire des Composites Thermosructuraux (UMR 5801 CNRS-SNECMA-CEA-UB1),
Universit e´ Bordeaux 1, 3 All e´ e La Bo e´ tie, F-33 600 Pessac, France
Fran c¸ ois Doux
SNECMA, BP 37, F-33 165 Saint M e´ dard-en-Jalles cedex, France
(Received 3 June 1999; accepted 14 October 1999)
This work deals with the measurement of pyrocarbon anisotropy on very thin fiber
coatings used to control the interfacial behavior in carbon/carbon composites.
Differentiation of the various pyrocarbons was performed through computerized image
analysis of the electron diffraction patterns by measuring the azimuth opening of the
carbon 002 diffraction arcs. This orientation angle decreases when the texture switches
from rough to smooth laminar. The relationship with the polarized light measurement
technique at a lower resolution is discussed.
I. INTRODUCTION
was well related to the x-ray “Bacon’s factor of aniso-
tropy.” Technical aspects have been further developed.
7
In the field of carbon/carbon composites, the relation-
ship between structure and mechanical properties is of
prime interest. This is mainly due to the high anisotropy
of the hexagonal form of carbon involved in those ma-
terials, especially in the different forms used in carbon/
carbon composites. A good example is the carbon fiber
itself Young’s modulus of which has been shown to be
directly related to the orientation angle of the basal
For example, Stevens has used a synchronous micropo-
9
larimeter to approach the bireflectance (R –R ).
e
o
However, the reflectance anisotropy level (R /R ) can
o
e
be easily studied on a regular polarizing microscope, ac-
8
cording to Diefendorf et al. The technique has not a
good spatial resolution (>2 ␮m) but is very accurate to
distinguish low-pressure and low-temperature chemical
vapor deposition/chemical vapor infiltration (CVD/CVI)
1
planes along the fiber axis.
12
pyrocarbons. It will be considered in this work as a
benchmark. For that reason, it will be described precisely
in Sec. III. A.
In the case of pyrocarbon, it is already known that
most of the physical properties change with their texture.
In the peculiar case of carbon composites, it has been
2
The optical texture of pyrocarbons is indeed well re-
shown that changing the texture of pyrocarbon between
1
3
lated in many situations, but a previous work has
shown that the scale of the relevant phenomena at the
interface of a C/C was far below the optical resolution, at
the scale of a few hundred carbon layers. A procedure
based on electron diffraction image analysis was devel-
oped to extract some quantitative structural data with a
high spatial resolution.
fiber and matrix by means of an interphase makes it
possible to change the tensile behavior from brittle to
nonbrittle. It is therefore especially important to measure
and control the anisotropy in order to monitor the inter-
2
,3
face,
e.g., fiber/matrix chemical bonding strength,
coefficient of thermal expansion (CTE), etc.
Different methods have been developed to measure the
anisotropy. Most are based on x-ray diffraction
1,4–7
8–11
II. EXPERIMENTAL METHODS
(XRD).
The others involved optical methods,
by
measuring the reflecting power anisotropy. For example,
an “optical anisotropy factor” has been introduced which
is defined as the ratio of maximum-to-minimum intensity
of the reflected light by rotating the stage on a 15-␮m-
large field. This was performed with a microphotometer
A series of model carbon/carbon composites, based on
single filaments or bundles of the same fibers, was pre-
pared by a CVD process in a hot-wall furnace described
elsewhere. Propane and also methane were used to man-
age the different pyrocarbon coatings. First a thin layer
3
10
and was later developed by Koizlik and Gr u¨ bmeier on
a smaller target: from 10 ␮m down to 5 ␮m. This factor
(
(
0.5 ␮m) was deposited with 3 different nanotextures
Table I). The aim of this work is to control the nano-
texture of these coatings because of their important me-
chanical role. The matrix was then deposited on the 3
materials under the same CVD conditions.
14
a)
Address all correspondence to this author.
e-mail: bourrat@lcts.u-bordeau.fr
92
J. Mater. Res., Vol. 15, No. 1, Jan 2000
© 2000 Materials Research Society
http://journals.cambridge.org
Downloaded: 04 Feb 2015
IP address: 193.255.248.150

X. Bourrat et al.: Pyrocarbon anisotropy as measured by electron diffraction and polarized light
TABLE I. Microstructure of pyrocarbons (matrix is designated by M
and interphases by Ia, Ib, and Ic).
CM30ST. The selected-area diffraction technique (SAD)
was preferred for its high angular resolution, with an
aperture of 110 nm in diameter (5-␮m selection aper-
ture). The illumination was set up with a high-emission,
Extinction angle and optical texture
Ae
C -lens moderately excited (spot size “2”), a 100 ␮m
1
Pyrocarbon
(deg)
Microtexture
condenser aperture was used, and a C -lens was defo-
2
M
Mint
Ia
Ib
Ic
17
18
15
13
11
Smooth laminar
Rough laminar
Smooth laminar
Smooth laminar
Smooth laminar
cused in order to keep a high signal intensity with a
negligible ␣ angle. The camera length on the plate was
L ס 1.35 m in order to obtain a high-resolution diffrac-
tion pattern. Three exposure times were used to obtain
the right optical density range, on the 002 ring (5 to 10 s).
The film was a Scientia™ film from Agfa.
A. Optical microscopy
C. Image analysis
For optical measurements, the same deposit conditions
were used, but only one type of pyrocarbon was depos-
ited as a thick coating to perform the measurement. Pol-
ishing was performed according to a classical protocol.
Finishing was obtained with a diamond paste (0.5 ␮m).
All the measurements were made in this study on a
MeF3 Reichert–Jung apparatus. It is an inverse micro-
scope equipped with a polarizer and a rotatable analyzer.
Samples are polished sections of the deposit with the
basal plane edge-on (maximum bireflectance). The ori-
entation of the anisotropy plane on the stage of the op-
tical microscope is set either at 45° or, for cylindrical
deposits, measurement has to be performed in the same
diagonal position as shown in Fig. 1 (empty arrow). In
this situation the incident light is plane polarized and has
its vibration direction parallel to the polarizer. Originally,
the analyzer is crossed: a Maltese cross appears with the
extinguished branches parallel to the Nicols.
The transmission electron microscopy (TEM) micro-
graphs were digitized with an Arcus II™ scanner from
Agfa equipped with a transmission back and with a reso-
lution of 2400 dpi. The optical density (OD) accessible
with that scanner is much better than that of the emul-
sion, 0.08 < DO < 3.30. The pattern was recorded with
the absolute density scale, by means of the automatic
control. A squared size image (1600 × 1600 pixels) was
obtained.
The image processing algorithms were developed on a
PowerMac 7100/66 (RAM, 72Mo, and a CPU, 66 MHz)
with the LabView™ software from National Instrument
(Austin, TX), provided with the Concept Vi™ image
analysis virtual instruments library from Graftek
(France). This software enables the development of dedi-
cated image processing applications in the LabView™
context.
B. Transmission electron microscopy
III. RESULTS
A. Optical measurement of the anisotropy
Samples were obtained by mechanical thinning and
ion-milling with a Gatan DuoMill™ at room temperature.
The details of the procedures have been described pre-
1. Fundamentals of the technique
15
The measurement of the optical properties of opaque
viously. Electron diffraction was conducted in a Philips
materials as pyrocarbon can be performed by means of
reflected polarized light. Optical properties depend on
three basic phenomena: reflection, refraction, and ab-
sorption of light. The plane which contains the electric
field to which the human eye is sensitive is considered as
the light vibration plane. As far as wave propagation in
anisotropic crystals is concerned, directions, speeds, and
refractive indices are always related to the normal to the
wave, as a convention. For anisotropic media such as
graphite, propagation in a given direction occurs follow-
ing two orthogonal polarized waves with different speeds
depending on their refractive index and reflectances as
represented in Fig. 2. For a plane light wave parallel to
the crystal face [Figs. 3(a) and 3(b)] the ordinary rays
(spherical) are transmitted without refraction whereas the
extraordinary rays (ellipsoidal) are propagating with re-
spect to the optical axis and do not follow Descart’s law.
FIG. 1. Cross section of coated fiber (F, single fiber coated with PyC,
pyrocarbon) in optical microscopy under cross-polars.
Two very different refractive indices (n ס 2.15 ± 0.3
o
J. Mater. Res., Vol. 15, No. 1, Jan 2000
Downloaded: 04 Feb 2015
93
http://journals.cambridge.org
IP address: 193.255.248.150

X. Bourrat et al.: Pyrocarbon anisotropy as measured by electron diffraction and polarized light
1
8
and n ס 1.81) as well as two absorbing coefficients
(
For the amplitude, absorption has to be taken into ac-
count. It is convenient to define the reflectance of the
graphite crystal for the different directions. Reflectance
is given by Fresnel’s equation:
e
18
␹ ס 0.66 ± 0.02 and ␹ ≈ 0.01) are related to these
o e
two vibrations inside the graphite crystal inducing a
strong birefringence and dichroism, respectively.
The situation of interest in this study is represented in
Fig. 3(b): the polished surface of the graphite crystal
through which the plane wave penetrates contains the
optical axis. This surface is perpendicular to the basal
planes. The ordinary and extraordinary rays are inter-
mingled. Waves propagate in the crystal following the
same path but still polarized at 90° and with the maxi-
mum speed difference ␷ − ␷ . Out of the crystal (of
2
2
͑
͑
n − N ͒ + k
R =
,
(2)
2
2
n + N ͒ + k
where N is the index of the embedding medium (in air
N ס 1), n the refractive index, and k the extinction
coefficient (k ס 1.42 and k ≈ zero). Experimental
measurements of the graphite reflectance in air were re-
ported by Ramdhor or Capdecome and more recently
o
e
o
e
16
17
thickness, l), vibrations (wavelength, ␭) are out of phase:
18
by Ergun:
⌽e − ⌽ ס 2␲(n − n )l/␭ (1)
,
o
e
o
Ramdhor: R ס 0.225 Re ס 0.050
,
(3)
(4)
o
where n − n is the index difference.
e
o
Ergun: R ס 0.2775 Re ס 0.083
.
o
The situation described in Fig. 3(b) is the case where the
highest bireflectance can be measured: R /R ס 3.34
o
e
following the more recent values of Eq. (4).
In-between crossed polars, in the optical microscope,
this situation is schematically represented in Fig. 4. The
graphite planes are oriented in the diagonal position at
␣
ס 45° from the normal position. Entering the crystal,
the polarized wave is then resolved into two equal com-
ponents projected with the same magnitude into the or-
dinary (OA) and the extraordinary direction (OC). Then,
the wave amplitudes reflected in the same directions are
proportional to the square root of the ordinary and
extraordinary reflectance: R and R , respectively. If
FIG. 2. Reflectance indicatrix of graphite: negative uniaxis.
o
e
the intensity transmitted by the polarizer is considered
FIG. 3. Wave propagation in graphite: (a) optical axis in any orien-
tation, (b) parallel to the optical axis (A , B , C , the ordinary wave
FIG. 4. Measurement of the extinction angle, A , in a graphite crystal
o
o
o
e
surface; A , B , C , the extraordinary wave surface).
with polarized light.
e
e
e
9
4
J. Mater. Res., Vol. 15, No. 1, Jan 2000
Downloaded: 04 Feb 2015
http://journals.cambridge.org
IP address: 193.255.248.150

X. Bourrat et al.: Pyrocarbon anisotropy as measured by electron diffraction and polarized light
to be 1, the amplitude of the two refracted beams emerg-
ing from the crystal can be vectorially written as the
following:
As far as the anisotropy is concerned, only the laminar
pyrocarbons can be analyzed in this way. Granular py-
rocarbons with more or less gas-phase nucleation cannot
be related to the same scale, whereas isotropic pyrocar-
bons can be easily distinguished by their faint reflec-
tance. Rough laminar (RL) and smooth laminar (SL)
have been regularly distinguished up until the work of
1/2
OA ס (R ) cos ␣ cos(␻t − ⌽)
,
1
o
1
/2
OC ס (R ) sin ␣ cos(␻t)
,
(5)
(6)
1
e
21
and using Ergun’s values from Eq. (4):
OA ס 0.29 cos 45 cos(␻t − ⌽)
Bokros. If the two polars are of good quality, the value
of A ranges from 10° to 22–24° for highly oriented
e
1
laminar pyrocarbons as reported in Table II. It should be
ס 0.20 cos(␻t − ⌽)
,
noted that pyrocarbons with values lower than A ס 10°,
e
12
which are sometimes called dark laminar, were never
clearly related to a laminar but more likely to a granular
pyrocarbon.
OC ס 0.53 sin 45 cos(␻t) ס 0.37 cos(␻t)
1
The resultant in the case of graphite is the vector OR. The
polarization plane of this vector has been rotated by an
angle A , regarding OP. If the analyzer is uncrossed by
the same angle, the crystal is completely extinguished
e
2
. Measurements
An example of an optical measurement is provided in
(Ae ≈ 16.3°). Physically in the analyzer, both rays are
Fig. 5. A fiber with a thick coating is selected in order to
obtain an accurate value. When the two Nicols are
crossed a Maltese cross appears. By rotation of the ana-
lyzer anticlockwise, the first quadrant (arrow) is extin-
guished and then becomes bright again. The angle value
for the maximum extinction is called the extinction
deflected into the transmission direction of the analyzer.
The resultant amplitudes are
1/2
OA ס (R ) cos ␣ cos(␻t − ⌽) cos(45 + Ae)
,
2
o
1/2
OC ס (R ) sin ␣ cos(␻t) cos(45 − Ae)
,
(7)
2
e
as |OC | ס |OA | in the extinction conditions, and after
angle, A , and is expressed in deg.
e
2
2
simplification
1/2
1/2
(
R ) /(R ) ס tan(45 + Ae)
,
(8)
o
e
TABLE II. Optical texture of pyrocarbons and extinction angle (Ae).
where A is expressed in deg. This analysis is giving
e
Optical texture
Domain of extinction angle
a good approximation of the reflectance anisotropy
Rough laminar
Smooth laminar
Isotropic
Ae > 18°
ratio R /R :
o
e
12° < A < 18°
e
2
A < 4°
e
R /R ס tan (45 + A )
.
(9)
o
e
e
This should be completely true if the two refracted
beams, OC and OA were still in phase. This difference,
1
1
however, is small. It corresponds to the first-order “iron
gray” of the Michel–L e´ vy chart, that is to say a dark gray
extinction which is sufficient to obtain the R /R ratio.
o
e
For pyrocarbons, bireflectance is produced by a “pseu4-
docrystal,” the turbostratic stacking of carbon layers.
First, the intrinsic bireflectance of that turbostratic pile
19,20
increases with increasing order of the stacking
and
also varies with heteroatoms content. Second, because
the measurement is obtained by means of an optical mi-
croscope, the response is averaged on a large volume
depending on the resolving power of the objective lens
(about 0.5 ␮m). Because of textural features (cones, cur-
vatures, etc.) only the average contribution is accessible.
As a result, this technique integrates structural and tex-
tural variations. Fortunately, both vary in the same way.
For that reason it was preferred to talk of the “apparent
reflectance anisotropy ratio Rmax /Rmin.” For the same
reason, it is best to consider the A angle value and not
e
FIG. 5. Measurement of the extinction angle (A ) in pyrocarbon with
e
the Rmax/Rmin ratio.
polarized light (example of a coated fiber).
J. Mater. Res., Vol. 15, No. 1, Jan 2000
Downloaded: 04 Feb 2015
95
http://journals.cambridge.org
IP address: 193.255.248.150

X. Bourrat et al.: Pyrocarbon anisotropy as measured by electron diffraction and polarized light
In this work, different CVD/CVI conditions were se-
lected in order to process a broad range of different py-
anisotropy plane (1) can be either tilted by an angle ␣
regarding the z axis (2) or twisted by an angle ␤ as
represented in (3). In the reciprocal space, this disorien-
tation gives rise to the distribution of the intensity on the
00l sphere (of 1/d00l radius). The electron diffraction
pattern is a plane section of the reciprocal space E*, xOy
in Fig. 7, where two 002 arcs appear perpendicularly to
the anisotropy plane. The electron diffraction pattern
gives direct access to ␤-angle distribution, because the
wavelength of the electron beam is very short (␭ ס
1.969 pm); thus, the Ewald sphere radius (1/␭) is large
and can be considered as a plane in the region analyzed.
Measurement of the azimuth opening of these arcs
along the Debye–Scherrer ring is proposed in this work
as the quantification of the anisotropy. The value of the
interlayer spacing (d002) can be obtained as well by a
previous standardization with a graphite pattern, obtained
under similar conditions.
rocarbons from rough to smooth laminar: 18° > A > 11°.
e
Optical measurement are reported on Table I.
First, under the same CVD/CVI conditions, the matrix
has a higher A value when measured inside the tow
e
(
M ) as compared to the external tow surface or to the
int
single fiber composite (M). The increase is minute but
reproducible (18° versus 17°). This effect can be related
to the higher residence time of gaseous precursors in-
volved in the infiltration process with respect to deposi-
tion process, i.e., the progress of the homogeneous
12
reactions within the tow as shown in a previous work.
Second, the conditions selected to obtain a poorly
oriented pyrocarbon for the interphase Ic gave a value of
1°. This is the lowest value found for a laminar pyro-
carbon. This microtexture is obtained by decreasing the
partial pressure of propane.
1
Finally, the intermediate nanotextures were reached by
changing the mother molecule for specific pressure/
2. Measurement algorithms
temperature conditions. Interphase Ia gave A ס 15°
e
The first step is to correct the optical density of the
film in order to retrieve the real diffracted intensity on the
whole peak which is to be measured. The response of a
charge coupled device (CCD) camera or that of an im-
aging plate might be the best solution. With a film, the
with methane and A ס 13° for Ib with propane at the
e
same temperature but at a lower pressure.
B. Pyrocarbon anisotropy as measured by
electron diffraction
1. Pyrocarbon diffraction and anisotropy
Pyrocarbons grow with the hexagonal form of car-
bon. Unlike graphite, their structure is limited to two
dimensions, i.e., the hexagonal lattice of the aromatic
4
layer. This has been known as the turbostratic struc-
22
ture since its discovery by Warren in 1941. The layers
are stacked one on the other with a rotational disorder.
These small stacks form domains. As these domains are
very small, some coherent scattering occurs in between
the layers even if not perfectly parallel. The diffracted
spots observed in the reciprocal space are not the 00.l
graphite spots, but the so-called 00l turbostratic inter-
layer interference.
FIG. 6. Model of turbostratic piles of carbon layers (according to
Bourrat ).
4
4
The different pyrocarbons can be described by the
size of the structural unit, the carbon layer, L , the num-
2
ber N of layers coherently stacked together, and the in-
terlayer distance in between the layers: d002 spacing. The
coherent lengths l and l can be measured by dark-field
a
c
techniques. They describe the coherent domain in paral-
lel and perpendicularly to the layer, respectively (Fig. 6).
In addition to the turbostratic pile, the different pyro-
carbons are differentiated by their anisotropy. When they
are highly anisotropic, all the layers are parallel to the
anisotropy plane. As this anisotropy decreases, more and
more domains are misaligned with respect to the aniso-
tropy plane. In the reciprocal space E*, this gives rise to
the polar distribution of the 002 spots as represented in
FIG. 7. Schematic of anisotropy in the real and reciprocal spaces
(according to Oberlin ).
23
23
Fig. 7 following Oberlin. Layers aligned parallel to the
9
6
J. Mater. Res., Vol. 15, No. 1, Jan 2000
Downloaded: 04 Feb 2015
http://journals.cambridge.org
IP address: 193.255.248.150

X. Bourrat et al.: Pyrocarbon anisotropy as measured by electron diffraction and polarized light
pattern must first be strictly recorded in the linear domain
automatically, provided that the central spot is system-
atically saturated by the linear expansion. The program
finds the two 002 maxima, so the Debye–Scherrer ring
can be drawn and superimposed on the pattern. If the
operator agrees with the result, the intensity is then ex-
tracted versus the azimuth angle (Fig. 10). The program
fits these peaks to two Gaussian curves, and the aniso-
tropy OA is given as the mean value of the two widths at
half-height.
of the emulsion. Most of the central peak is out of this
domain. The emulsion must then be standardized. For
that purpose, a series of increasing exposure times are
performed with increasing beam intensities in order to
plot the curve of Fig. 8, the optical density as a function
of the exposure time. In this way, every digitized pattern
is corrected by means of this transfer function in order to
recover the real dynamics. Finally, a linear expansion is
performed in order to find the dynamics of the 255 gray
levels from the foot to the top of the 002 peaks.
3. Anisotropy of pyrocarbon interphase in
At this stage, a tool was developed to scan the inten-
sity across the diagram (Fig. 9) passing through the cen-
ter. The center of the diagram was previously found
C/C composites
The opening angle OA and the interlayer spacing d002
have been measured for the different pyrocarbons. Un-
fortunately, it was not possible to obtain both optical and
diffraction measurements on exactly the same sample.
Figure 11 shows a 002-dark-field image of an interphase
Ia. The inset is a low magnification of the whole cross
section of the microcomposite. The three different car-
bons can be easily identified, fiber (F), Ia interphase, and
matrix (M).
The fiber is characterized by small coherent domains
with an isotropic nanotexture. A high-resolution micro-
graph (Fig. 12) shows this isotropic nanotexture. The
layers are defective with a fairly low coherent stacking
lc ס 1.1 nm, which gives an average lc/d002 ≈ 3 layers
per domain.
FIG. 8. Experimental darkening curve of Scientia™ emulsion.
Except on rare places, the interphase is disbonded
from the fiber surface. This disbonding already exists in
3
the original composite but amplified during thinning.
Electron diffraction shows that this pyrocarbon is well
organized (OA ס 61°) with a d002 spacing lower than
that of the matrix. The dark-field technique reveals that
FIG. 10. Front panel of the software. Experimental pattern and azi-
muth plot (white) to be analyzed. Gaussian fit (down) of the experi-
mental intensity. OA is the mean width at half-height of the two
Gaussian curves.
FIG. 9. Image analysis: (a) electron diffraction pattern (3D), (b) radial
scan and plot of the intensity along the Debye–Scherrer 002 ring.
J. Mater. Res., Vol. 15, No. 1, Jan 2000
Downloaded: 04 Feb 2015
97
http://journals.cambridge.org
IP address: 193.255.248.150

X. Bourrat et al.: Pyrocarbon anisotropy as measured by electron diffraction and polarized light
the first 300 layers (≈50 nm) on the fiber surface are
more organized. Domains are more contrasted exhibiting
a columnar shape. However, this was not quantified.
The matrix was deposited using a different procedure.
Optical measurements were found to be close to, but
different from, those of the interphase: A ס 17° versus
e
1
5° in the Ia interphase. OA values obtained in the matrix
and the interphase belong to the same range: 59° < OA <
1°. Only d002 spacing was seen to be systematically
6
lower in the interphase than in the matrix: e.g., 3.4 ver-
7
sus 3.5 Å for the matrix in Fig. 11.
5
The same analysis was conducted for the other two
materials processed with the same fiber but different in-
terphases Ib and Ic (Figs. 13 and 14). The aim was to
lower the interphase anisotropy: A ס 13° and 11°,
e
respectively.
Interphase Ib has the same poor adhesive behavior as
interphase Ia. Nearly all the interphase is debonded from
the fiber surface. Scanning electron microscopy (SEM)
on the bulk material verified that this disbonding is not an
artifact but existed previously and was amplified by sam-
3
pling. Otherwise it appears very different from inter-
phase Ia. The structure and nanotexture of Ib give values
very close to those of the matrix. Only two measurements
were possible: interlayer spacing were 3.5 and 3.6 Å
6
2
FIG. 12. Lattice fringes image of the fiber [high-tenacity polyacrylo-
nitrile (HT PAN)-based carbon fiber] (inset: electron diffraction
pattern).
FIG. 11. Microcomposite with interface Ia (cross section, TEM 002
dark field). The inset shows a low magnification of the microcomposite.
FIG. 13. Microcomposite with interface Ib (cross section, TEM 002
dark field). The inset shows a low magnification of the microcomposite.
9
8
J. Mater. Res., Vol. 15, No. 1, Jan 2000
Downloaded: 04 Feb 2015
http://journals.cambridge.org
IP address: 193.255.248.150

X. Bourrat et al.: Pyrocarbon anisotropy as measured by electron diffraction and polarized light
and OA 60° and 61°. The matrix and Ib were obtained
just by changing the pressure of the propane. It should be
noted that propane does not produce the surface effect as
noticed with methane.
With interphase Ic the composite interface is not dis-
bonded (Fig. 14, inset). This is very important from the
viewpoint of the interface mechanics. That was not the
case with interphase Ia and Ib or without interphase. It is
assumed that the chemical bonding with the fiber is
stronger and/or that transversal residual stresses are bet-
ter matched. The nanotexture of Ic is clearly poorer:
65° < OA < 74°. The interlayer spacing is measured in a
broad range from 3.6 to 3.7 Å, higher than for the other
5
pyrocarbons. These values are distinguished from the
previous ones. They correspond to the less oriented lami-
nar pyrocarbon obtainable as verified by optical analysis.
The 002 dark-field technique shows that this deposit is
homogeneous.
FIG. 15. Plot of the electron diffraction data: d002 spacing versus
orientation angle OA.
4. Comparative variation of anisotropy OA and d002
interlayer spacing
linear decrease of A does not correspond to an equiva-
e
lent linear decrease of OA value. Only Ic or Mint can be
clearly distinguished with OA as a single variable. The
same phenomenon occurs with d002 spacing. The de-
crease of the reflectance ratio can occur without any
decrease of d002 spacing; that is the case of pyrocarbon Ia
with the same spacing as the rough laminar pyrocarbon
Figure 15 shows the plots of the interlayer spacing
d002) versus the orientation angle OA for the different
pyrocarbons. Despite the spread of the data, the various
pyrocarbons can be distinguished on this plot. Generally,
(
when the reflectance ratio or A decreases, the orientation
e
angle OA and the d002 spacing increase. But in detail, the
Mint (d002 ≈ 3.4 Å), but with a lower anisotropy (OA ס
9
59°). It can thus be seen that the optical properties of
pyrocarbons (such as reflectance ratio, A ) are probably
e
not linearly related to structural parameters as the long-
range anisotropy (OA) or d002 spacing of the turbostratic
pile. It should be noted that the d002 spacings measured
locally by SAD vary in a broader range than those usu-
ally obtained by XRD values measured on bulk samples.
This is attributed to stress relaxation during sampling.
Figure 15 shows interesting results on the matrix.
First, when the matrix is infiltrated (M ) in the core of
int
the tow (3000 filaments), the optical measurement shows
a very small increase from 17° (for M) to 18° (for Mint,
empty triangles). Diffraction measurement OA is much
more sensitive: OA is seen to be 15 deg lower inside the
bundle than at the surface.
When the matrix is deposited on single fibers (solid
triangles), Fig. 15 shows that the data are highly spread
along the ordinate. For a mean OA value of 59°, d002 can
vary considerably: 3.5 < d002 < 3.7 Å. A direct exami-
5
nation of the pyrocarbon by high-resolution TEM is not
easy to perform in the area where the diffraction patterns
is realized. However, from a statistical point of view,
high resolution has revealed some heterogeneity at the
scale of 100 nm, the scale of the diffraction technique.
Figure 16 shows the ordering of the smooth laminar py-
rocarbon (A ס 17°) in the best areas. Layer size could
FIG. 14. Microcomposite with interface Ic (cross section, TEM 002
dark field). The inset shows a low magnification of the microcomposite.
e
be measured: L ≈ 6 nm and L ≈ 1 nm for that pyrocar-
2
1
J. Mater. Res., Vol. 15, No. 1, Jan 2000
Downloaded: 04 Feb 2015
99
http://journals.cambridge.org
IP address: 193.255.248.150

X. Bourrat et al.: Pyrocarbon anisotropy as measured by electron diffraction and polarized light
time, the quantity of lattice defects increases, mainly lat-
3
tice disclinations, vacancies, sp hybridization, hydrogen
atoms, etc. In other words, disturbances of the electronic
structure occur which are likely to control the absolute
value of the reflectance. For that reason, the physical
meaning of A is somewhat complex, since this measure-
e
ment is only a function of the apparent reflectance ratio,
A ס f(R /R ). In addition to the disorientation of the
e
o
e
carbon layers, A thus sums up different effects.
e
As far as OA measurement is concerned, the volume
selected by a very small aperture (110 nm) was chosen
13
for being optimum: large enough to be representative
and small enough to ignore the curvature produced by the
fiber for example. Finally, among diffraction data d002
spacing was seen to be widely scattered. This is partly
due to the occurrence of disturbed areas at the same scale
and very likely to stress relaxation which occurs (or not)
during thinning.
V. CONCLUDING REMARKS
The optical method is based on the measurement of the
apparent reflectance ratio of the pyrocarbon. This ratio is
decreasing with the structure (intrinsic) and with the an-
isotropy. Instead of using a photomultiplier (PM) this
method involves the rotation of the analyzer of a regular
polarization device. It has the advantage of being easy to
perform. However, it requires a deposit at least 2 ␮m
thick and high-quality optical polishing. This measure-
ment is only semiquantitative, but it enables the separa-
tion of the different low-temperature pyrocarbons. This
was demonstrated in this work on the different smooth
laminar pyrocarbons obtained by isobaric isothermal (I)-
CVD/CVI.
The electron diffraction technique is irreplaceable to
analyze coatings as thin as 0.5 ␮m and less. Coupled to
an image analysis software, it gives the following struc-
ture parameters: (i) the orientation angle, OA, of the
scattering domains which is a measurement of the “geo-
metric” anisotropy; (ii) the interlayer spacing d002 which
is a basic measurement of the short distance ordering of
the carbon turbostratic structure. Other information can
be extracted from the diffraction pattern. This requires
more work to extract them with accuracy, e.g., (iii) Lc,
FIG. 16. Matrix nanotexture with a disturbed area. Inset: schematic of
the ⍀-like shape defect (TEM lattice fringes technique).
bon. In other places, some levels are disturbed by some
gas-phase nuclei, Fig. 16. They induce the rotation of the
layers which are subsequently deposited, inducing ⍀-like
features. Contrary to the mechanism controlling the
decreasing anisotropy of laminar pyrocarbons, a com-
plete rotation of the layers occurs here (perpendicu-
larly to the anisotropy plane). Also, as the carbon is
poorly organized in those regions, from place to place,
d002 spacing can thus vary with the abundance of these
defects.
IV. DISCUSSION
CVD conditions (T, P, flow rate, or mother molecule)
were optimized in this work, in relation with the optical
scale, from A ס 17° for the matrix to A ס 11° for the
e
e
Ic interphase. The diffraction data have clearly evidenced
that the optical scale might not be linearly related to the
anisotropy as defined by the orientation angle of the co-
herent domains. The question is to better understand
what relationships exist between the two scales.
the turbostratic pile thickness; (iv) L , the coherent extent
a
of the carbon layer, or (v) P , the ratio of graphite stack-
1
ing in the turbostratic pile.
In comparison of the two techniques, optical measure-
ment, on one hand, and the electron diffraction OA and
d002 spacing, on the other hand, have shown a good cor-
relation, on the whole, and a useful complementarity.
Both gave the same variations among the first samples
reported in this work (I-CVD/CVI smooth laminar pyro-
carbon). Moreover, the two scales appear to be linearly
related.
The scales at which these two techniques are applied
are different; optical measurement integrates volumes at
the micrometer scale, whereas electron diffraction ap-
plies to volumes 100 times smaller. When the anisotropy
decreases for the different pyrocarbon types, L and L ,
a
c
the coherent lengths are known to decrease. In the mean
1
00
J. Mater. Res., Vol. 15, No. 1, Jan 2000
Downloaded: 04 Feb 2015
http://journals.cambridge.org
IP address: 193.255.248.150

X. Bourrat et al.: Pyrocarbon anisotropy as measured by electron diffraction and polarized light
Depending on the spatial resolution required and on
7. G.E. Bacon, J. Appl. Chem. 6, 477 (1956).
. R.J. Diefendorf and E.W. Tokarsky, Air Force Report AF
3(615)-70-C-1530 (1971).
8
the nature of the parameters required, one or the other of
the techniques will be equivalent and at the same time
complementary. Indeed, optical measurements integrate
all scales up to the micrometer, while electron diffraction
gives information from the nanometer to the micrometer
scale depending on the area selected.
3
9
. D.W. Stevens, Proc. 6th Int. Materials Symposium, Univ. of Cali-
fornia at Berkeley, 1976, pp. 129–141.
10. K. Koizlik and H.B. Gr u¨ bmeier, Leitz-Mitt. Wiss. U. Techn. Bd.
VI, No. 4 (1974), pp. 121–126.
1. T. Hirai and S. Yajima, J. Mater. Sci. 2, 18 (1967).
2. P. Dupel, X. Bourrat, and R. Pailler, Carbon 33, 1193 (1995).
3. X. Bourrat, R. Paillet, and P. Hanusse, Proc. of the 21st Biennial
Conf. on Carbon, Buffalo, June 13–18, 1993, State Univ. of NY at
Buffalo (Am. Carbon Society, Washington, DC, 1993), pp. 94–95.
4. R. Naslain, in High Temperature Ceramic-Matrix-Composites,
Ceramic Transactions. Vol. 58, edited by A.G. Evans and
R. Naslain (1992), pp. 23–39.
1
1
1
ACKNOWLEDGMENT
1
The authors wish to thank the SEP division of
SNECMA for a grant given to B.T.
1
1
5. S. Jacques, A. Guette, F. Langlais, and X. Bourrat, J. Mater. Sci.
3
2, 2969 (1997).
6. P. Ramdhor, Arch. Eisenh u¨ ttenwes Dtsch 25A (1928);
H. Schneiderhohn and P. Ramdhor, in Lehrbuch der Erzmik-
roskopie (Berlin, 1941), T11.
REFERENCES
1
1
7. L. Capdecome, Bull. Soc. Fr. Mineral. 78, 181 (1943).
8. McCartney and S. Ergun, Fuel 37, 272 (1960); S. Ergun,
Nature 14, 135 (1967).
1
2
. W. Ruland, Appl. Polymer Symp. 9, 293 (1969).
. B. Trouvat, X. Bourrat, and R. Naslain, Proc. 23rd Biennial Conf.
on Carbon, Pennstate 18–23 July 1997 (American Carbon Soci-
ety, Washington, DC, 1997).
1
9. R.A. Forest, H. Forest, H. Marsh, C. Cornford and B. T. Kelly, in
Chemistry and Physics of Carbon, edited by P.L. Thrower (Marcel
Dekker, New York, 1984), Vol. 19.
3
4
. B. Trouvat, Thesis, University of Bordeaux 1, OA, 1635 (16 De-
cember 1996).
. X. Bourrat, in Sciences of Carbon Materials, edited by H. Marsh
and F. Rodriguez-Reinosso (Univ. Alicantes Publisher, Alicante,
Spain 1999), Chap. 1.
2
2
0. J.N. Rouzaud and A. Oberlin, Thin Solids Films 105, 75 (1983).
1. J.C. Bokros, in Chemistry and Physics of Carbon, edited by P.L.
Walker, Jr. (Marcel Dekker, New York, 1969), Vol. 5, p. 1.
2. B.E. Warren, Phys. Rev. 59, 693 (1941).
2
2
5
6
. D.B. Fiscbach, J. Appl. Phys. 37, 2202 (1966).
. W. Ruland, J. Appl. Phys. 38, 3585 (1967).
3. A. Oberlin, Carbon 17, 77 (1979).
J. Mater. Res., Vol. 15, No. 1, Jan 2000
101
http://journals.cambridge.org
Downloaded: 04 Feb 2015
IP address: 193.255.248.150