Accepted Manuscript

Comparison of different surface treatments of carbon fibers used as reinforcements

in epoxy composites: Interfacial strength measurements by in-situ scanning electron
microscope tensile tests

Yu Liu, Delong He, Ann-Lenaig Hamon, Benhui Fan, Paul Haghi-Ashtiani, Thomas
Reiss, Jinbo Bai
PII: S0266-3538(18)30090-3
DOI: 10.1016/j.compscitech.2018.08.018
Reference: CSTE 7351

To appear in: Composites Science and Technology

Received Date: 11 January 2018

Revised Date: 15 August 2018
Accepted Date: 16 August 2018

Please cite this article as: Liu Y, He D, Hamon A-L, Fan B, Haghi-Ashtiani P, Reiss T, Bai J, Comparison
of different surface treatments of carbon fibers used as reinforcements in epoxy composites: Interfacial
strength measurements by in-situ scanning electron microscope tensile tests, Composites Science and
Technology (2018), doi: 10.1016/j.compscitech.2018.08.018.

This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to
our customers we are providing this early version of the manuscript. The manuscript will undergo
copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please
note that during the production process errors may be discovered which could affect the content, and all
legal disclaimers that apply to the journal pertain.
 ACCEPTED MANUSCRIPT
Comparison of different surface treatments of carbon fibers used as reinforcements in
epoxy composites: Interfacial strength measurements by in-situ scanning electron
microscope tensile tests

Yu LIU, Delong HE*, Ann-Lenaig HAMON, Benhui FAN, Paul HAGHI-ASHTIANI, Thomas
REISS and Jinbo BAI*

Laboratoire Mécanique des Sols, Structures et Matériaux (MSSMat), CNRS UMR 8579,

PT
CentraleSupélec, Université Paris-Saclay, 3 rue Joliot-Curie, 91192 Gif-sur-Yvette Cedex,
France

RI
Corresponding email: [email protected], [email protected]

SC
KEYWORDS: Interface strength, Interphase, FIB, STEM, EDX, fibers/Epoxy composites

ABSTRACT

U
In-situ characterization of the fiber/matrix interfacial failure behavior at microscopic scale is
AN
important to optimize the fiber surface treatment and to design high performance composites.
In this study, in-situ tensile tests in scanning electron microscope (SEM) were used to
M

investigate the interfacial adhesion strength of epoxy composites reinforced by four kinds of
carbon fibers (CF)—raw CF, desized CF, carbon nanotube-grafted CF (CNT-CF) and
D

oxidized CNT-CF. The crack initiation position and fracture failure mode were well recorded.
The strains and the interfacial adhesion strength were obtained for these four kinds of
TE

composites. It was found that the interfacial strength decreased from 53 MPa to 48 MPa after
removing the sizing on carbon fiber surface. However, by grafting CNTs on the CF surface,
EP

the interfacial strength reached 55 MPa and was further increased to 58 MPa after a simple
thermal oxidation treatment. Moreover, energy dispersion X-ray analysis (EDX) was carried
C

out using scanning transmission electron microscopy (STEM). The EDX mapping
AC

demonstrated that oxygen aggregated at the interfaces of raw CF/epoxy and oxidized CNT-
CF/epoxy. Thus, a combination of CNT grafting with chemical functionalization should be
necessary to achieve high performance carbon fiber reinforced polymer composites.

*Corresponding author. E-mail: [email protected] (J. BAI); [email protected]

(D. He)
 ACCEPTED MANUSCRIPT
1. INTRODUCTION

The interface between fibers and matrix is considered to have a major influence on the
mechanical properties of composites [1-4]. The good bonding or interaction at the interface is
essential for achieving high shear and off-axis strength, since the stress exerted on the
polymer matrix is transferred to fibers through the interface [5-8]. Increasing the surface
roughness and depositing a transition layer are two major methods to improve the interfacial

PT
properties between a polymer matrix and fibers. For instance, covering a sizing layer on fiber
surface has proved to be an efficient method to enhance the interaction between CF and the

RI
polymer matrix [9, 10]. In recent years, the introduction of carbon nanotubes (CNTs) into
conventional continuous fiber-reinforced polymer to create hierarchical reinforced structures

SC
is a current topic of interest [11-19]. Due to their unique structure, outstanding strength and
modulus, excellent electrical and thermal properties, CNTs have been exploited to improve

U
mechanical and functional properties for the fiber-reinforced composites [20-24]. To seek an
optimum interaction at the interface for hierarchical composites, it is important to evaluate the
AN
interfacial strength in different states of surface modification.

Over the last four decades, numerous works have been published to reveal the
M

interfacial properties between CF and epoxy matrix. Single fiber filament composite tests [25],
three-point bending test [26], single fiber pull-out tests [27] and droplet methods [28] have
D

been applied to various systems. A method has also been developed in which fibers are
TE

pushed out of the matrix rather than being pulled out [29]. Each of these methods attempts to
consider the interfacial shear strength between fibers and matrix. Also, another procedure [30]
EP

determined the tensile strength of the fiber-matrix interface by embedding fibers

perpendicularly to the loading axis through the thickness of bond-shaped specimens and
observed using scanning electron microscopy (SEM). The fibers with high interfacial
C

adhesion were reported to reinforce the composite’s longitudinal shear. Meanwhile, the
AC

transverse tensile strength and fracture surfaces with resin-coated fibers indicated the good
fiber–matrix adhesion.

Here, we develop an in-situ measurement of the interfacial strength for CF/epoxy

composites using a SEM and a nano-manipulator which applies the tensile loads on the
samples. Four kinds of CF of different surface state–raw CF, desized CF, carbon nanotube-
grafted CF (CNT-CF) and oxidized CNT-CF–have been used to study the various interfaces
in the composites. A lamella containing half-fiber and half-epoxy is prepared by a focused ion
 ACCEPTED MANUSCRIPT
beam (FIB) where the fiber is perpendicular to the tension axis, along the width of the
specimen. This allows both an observation of the fracture behavior of the interface and a
measurement of its interfacial strength. Furthermore, an elemental analysis has been carried
out in a scanning transmission electron microscopy (STEM) by combining an energy
dispersive X-ray analysis (EDX) mapping which offers some insight into the effectiveness of
the surface modification of the CF.

PT
2. EXPERIMENTS

2.1 Materials preparation

RI
The growth of CNTs on CFs was carried out in a chemical vapor deposition (CVD)

SC
quartz tube with a diameter of 100 mm, which was heated by an electrical resistance furnace
(CARBOLITE®). One kind of T700 grade CF fabric was used in this study. Its diameter was
around 7 µm. The furnace was heated up to 600 ºC in an argon and hydrogen flow with a rate

U
of 1.2 and 0.3 L/min, respectively. After 10 min stabilization, a xylene solution containing
AN
0.1 g/ml ferrocene was injected in form of spray into the reactor by a syringe injector, with an
injection speed of 24 ml/min. The growth lasted 3 min for CNTs grafted on the CFs. The
M

furnace was finally cooled down to room temperature under argon protective atmosphere (1
L/min) [31].
D

The as-prepared CNT-CF were subsequently oxidized: they were heated up to a

TE

temperature of 400 °C and stabilized for 4 min in air. Since the commercial CFs are coated by
a sizing layer and since the influence of this layer is best studied by comparing with the rough
CFs (without this layer), a “desizing process” has been conducted by heating the as-received
EP

CFs up to a temperature of 650°C for several minutes in an argon atmosphere.

A 1080S epoxy resin (Resoltech, France), accompanied with the curing agent 1084
C

(Resoltech, France) with a mass ratio of 1/3, was used. The resin and curing agent were mixed
AC

homogeneously and then a small bundle of CF fabric was slowly put into the epoxy matrix for
a thorough immersion. Afterwards, the mixture was degassed for 60 min inside a vacuum
oven at room temperature. The curing process was conducted at 60 ºC for 15 hours. At last,
the post-curing process was conducted at 100 ºC for 1.5 h. The as-prepared composites were
cut into small cuboids for further microstructure characterizations.

2.2 Samples preparation for in-situ SEM tensile test

 ACCEPTED MANUSCRIPT
CF/Epoxy composites used for the FIB-SEM was previously polished to obtain an
ultra-smooth surface. The FIB sample preparation was performed in a Helios 660 (FEI) Dual
Beam FIB-SEM system with a Ga+ ion source at 30 kV. The sample preparation procedures
are showed in Fig.1. First, a cross-section with vertically aligned CF was selected (Fig. 1A)
and a platinum (Pt) protective layer with a thickness of 1.0 µm was deposited on the surface
of the target milling area (Fig.1B). To reduce the testing error, the Pt layer should cover the

PT
center of the target fiber, as shown in the insert scheme. Then, a 21 nA and a 2.5 nA beam
current were applied respectively to obtain the lamella with a thickness of ≃1.5 µm (Fig.1C).

RI
Afterwards, the section was detached from the surrounding material and transferred to a TEM
half-grid for fine thinning (Fig.1D). Then, fine thinning was carried out only at the “stretching”

SC
areas to remove the surface deposited materials. To avoid stress concentration during
stretching, two quarter-circles were made in the CF region (Fig.1E). Finally, the micro-
manipulator was glued with the stretching sample by depositing a Pt layer. To avoid the point

U
force loaded on the sample, the deposited Pt covered the whole width of the stretched sample
AN
(Fig.1F). The width and the thickness for the tested samples was 5.0 µm and 1.5 µm,
respectively and the epoxy part had a length ranging from 6.5 µm to 7.0 µm.
M

Once the manipulator was sturdily bonded to the tensile sample by Pt deposition, the
magnification of electron beam images and ion beam images were set properly, normally at
D

6500×. The current of the ion beam was set to 80 pA to decrease the damage to the tensile
TE

sample. A stepwise displacement of 200 nm was applied to the manipulator. Both electron
beam image and ion beam image were acquired to measure the deformation during each
tensile step.
EP

All the tensile experiments were conducted following the same procedure, eight for
raw CF/epoxy, five for desized CF/epoxy, five for CNT-CF/epoxy and seven for oxidized
C

CNT-CF/epoxy, respectively.
AC
 ACCEPTED MANUSCRIPT

PT
RI
SC
Fig. 1 Preparation of the sample for micro-tensile test: A) the polished surface containing
several CFs which are marked out by red circles; B) a Pt protection is deposed on the

U
CF/Epoxy interface with a dimension of 20µm×2µm×1µm, and the red circle points the
complete CF section (The insert shows the precise region where the tensile section was cut
AN
from); C) the tensile sample is milled down to a thickness of 1 µm with a width of 8 µm; D)
the lamella is lifted out by the manipulator and welded to the copper grid edge by Pt
M

deposition; E) two quarter-circles are conducted on the interface to avoid the stress
concentration; F) the manipulator is attached to the lamella by deposing a Pt layer, the final
D

stretching direction is vertical to the interface.

TE

2.3 Characterization

The microstructure of different CF surfaces was observed by SEM (ZEISS, LEO 1530
EP

Gemini) at 5 kV. The TEM sample preparation process is described in detail elsewhere [32].
TEM imaging was performed by a Titan3 G2 (FEI) with a field emission gun (XFEG)
C

operating at an accelerating voltage of 80 kV. Energy-dispersion X-ray microscopy (EDX)

was carried out on the same equipment and operated at 80 kV in a STEM mode. All the
AC

images were acquired by high-angle annular dark-field detector (HAADF). The EDX
mapping was acquired for 10min.

3. RESULTS and DISCUSSION

 ACCEPTED MANUSCRIPT

PT
RI
U SC
AN
Fig. 2 SEM images showing the CF morphology: (A) original CFs; (B) desized CFs; (C)
M

CNT-grafted CFs; and (D) CNT-CF to measure the thickness of the CNT layer.

The morphologies of the raw CFs, desized CFs and CNT-CF are shown in Fig. 2. The
D

raw fibers have a smooth and cylindrical surface with a diameter of around 7 µm (Fig. 2A).
TE

After desizing, the CF surface has no obvious change and few differences could be detected at
a SEM resolution scale (Fig. 2B). The in-situ growth of homogeneous CNTs on the CFs was
EP

achieved by an easy one-step CVD method using the metalorganic compound ferrocene as a
catalyst precursor. As shown in Fig. 2C, the CNTs grow homogeneously on the whole surface
of the fabric and on each CF. More importantly, each CF is covered by a continuous
C

cylindrical shell consisting of aligned CNTs. The thickness of the CNT layer on the CF was
AC

measured by SEM, which can be seen in Fig. 2D. A part of the CNT was removed manually
(i.e. using a macroscopic tool) before the SEM observation. The thickness of the CNT layer is
about 440 nm. The real thickness of the CNT layer may be larger than this value since the CF
cylinder may not be tilted to the right position. To create some oxygen-containing groups on
the CF and CNT surfaces, the CNT-CF were heated up to a temperature of 400 °C in
atmospheric condition for 3 min. No obvious change could be observed by SEM (Fig. S1B).
 ACCEPTED MANUSCRIPT
To evaluate the axial interfacial strength between CFs and the epoxy matrix, an in-situ
method has been developed in a SEM column. There are several advantages to testing fibers
using in-situ SEM and a significant one among them is a precise measurement of the strain
along the entire gage length. Besides, the stretching process can be clearly observed along
with the fracture behavior.

PT
RI
SC
Fig. 3 Typical tensile specimen, showing a reduced gage section and enlarged shoulders.

U
Considering the typical tensile specimen in Fig. 3, it has enlarged ends or shoulders for
griping. The important part of the specimen is the reduced gage section. The cross-sectional
AN
area of the gage section is reduced relatively to that of the remainder of the specimen so that
deformation and failure will be localized in this region. Another important area is the
M

connecting region between the gage section and the shoulders, which should be great enough
so that the stress concentration does not affect the gage section. Usually, the R/W ratio is used
D

to evaluate the stress concentration degree. In the traditional tensile test, this ratio is larger
TE

than 2 to avoid a stress concentration.

C EP
AC

Fig. 4 Sketch of a tensile specimen measuring the interfacial strength between the CF and
epoxy matrix; (B) SEM image of final tensile specimen prepared by FIB-SEM.

A specimen in micrometre-scale has been designed as shown in Fig. 4A. It keeps a

half shape of traditional tensile specimen, with a gage section formed by the epoxy matrix and
a shoulder section formed by the CF. To get rid of the stress concentration caused by the neck
 ACCEPTED MANUSCRIPT
region, care was taken to place the interface far enough from it. The final specimen prepared
by FIB is shown in Fig. 4(B). Theoretically, the stress transferred homogeneously in the
cross-section all along the tensile specimen. The stress acting on the epoxy matrix equals the
stress at the interface. Hence, the interfacial strength at the interface can be obtained by
measuring the stress loading on the epoxy matrix part. The following steps show the details to
estimate the adhesion strength: (1) the original length l0 of epoxy matrix can be precisely

PT
measured on the SEM image from the interface to the edge of Pt deposition. (2) the tensile
force is applied by shifting the manipulator discontinuously with the smallest step (200 nm)

RI
until the failure is reached. (3) the images are acquired after each shifting step to monitor
length change in epoxy matrix.

U SC
AN
M
D
TE
C EP
AC
 ACCEPTED MANUSCRIPT
Fig. 5 Representative failure behavior for 4 different composites: (A) raw CF/epoxy, (B)
desized CF/epoxy, (C) CNT-CF/epoxy, (D) oxidized CNT-CF/epoxy failure at the interface
and (E) oxidized CNT-CF/epoxy failure in the epoxy matrix.

Fig. 5 provides the failure behavior of four different composites during tensile tests
with rectangular-shaped samples. Fig. 5(A) shows the raw CF/epoxy before and after failure,

PT
in which the fracture happened in the interface. By measuring the deformation on the matrix
part, the interfacial stress could be estimated precisely. This situation happened in most part

RI
of tests of the desized CF (Fig. 5B), CNT-CF/epoxy (Fig. 5C) and oxidized CNT-CF/epoxy
composites (Fig. 5D). Fig. 5(E) shows a strong interaction in the interface, in which the

SC
fracture happens in the epoxy matrix. In this situation, the interfacial stress exceeds the
fracture stress of epoxy matrix. This situation happened only in two tested samples of the

U
oxidized CNT-CF/epoxy composites. Fig. S9(C) shows a failure tensile test due to the
AN
debonding of the manipulator. Usually, the epoxy matrix breaks in the plastic region. Hence,
if the failure happens in the Pt deposition part, the specimen could not be re-soldered since the
second tensile test in the fractured sample could not reflect the real interfacial stress based on
M

the strain.
D
TE
C EP
AC
 ACCEPTED MANUSCRIPT
Fig. 6 Stretching process of a raw CF/epoxy composite sample observed by SEM with two
series: (A) electron and (B) ion images; (C) strain-step curves during the stretching process,
data extracted from two series of images; (D) sketch of the global sample position, with
regard to the stretching direction and the two beam sources.

Fig. 6(A) presents SEM images where an increasing tensile displacement is applied to
an interface until failure. The green arrows on the top indicate the tensile direction, and the

PT
CF/epoxy interface is pointed out by a green line. It is obvious that the CF part is thicker than
the matrix, which is caused by the milling process: Since CFs are much harder than polymers,

RI
the CF cannot achieve the same thickness as the matrix with the same milling conditions.
However, it is found that the interface position remains at the same position during the

SC
stretching process, which indicates the traverse strain of the CF can be negligible compared
with the polymer domain due to its much higher modulus.

U
As a method of evaluation and confirmation of the tensile displacement, ion beam
AN
images were also acquired during testing along with electron images, as shown in Fig. 6(B).
For each tensile stepwise, the error in displacement can be effectively reduced by measuring
and comparing the two series of images. Fig. 6(D) shows the sample position: the angle
M

between the electron beam and the sample surface plane is 38° and the angle between the ion
beam and the sample surface plane is 90°. The stretching direction is perpendicular to the
D

plane generated by the two beam sources.

TE

The apparent strain is defined by the change in the length of the segment between the
interface and the deposited Pt. As shown in Fig. 6(C), the curves about the strain-stretching
EP

steps in the two series of images are extracted. During the stretching process, the manipulator
moves by steps of 200 nm. However, in the first stage, the strain is not increased due to the
C

tiny bending existing in the manipulator. The two curves almost overlap, showing little
difference. Hence, the strain measurement can be conducted with either type of image which
AC

exhibit a clear CF/epoxy interface.

 ACCEPTED MANUSCRIPT

PT
RI
Fig. 7 (A) Displacement-tensile step plot of stretching samples extracted from ionic images.

SC
The black, red, green and blue lines are the results from the raw CF/epoxy, desized CF/epoxy,
CNT-CFs/epoxy and oxidized CNT-CFs/epoxy, respectively. (B) Fracture strengths of the
raw CF/epoxy, desized CF/epoxy, CNT-CF/epoxy and oxidized CNT-CF/epoxy.

U
The displacement is measured by either ionic or electron images acquired during each
AN
tensile step, as shown in Fig. 7(A). The strain of all curves increases linearly during the
stretching. The curves have a similar slope since the major strain is localized on the epoxy
M

part with almost the same geometry. The small discrepancy in the curves is due to the flurry
boundary in some images (the green lines in Fig. 6(A) and (B)). According to the measured
D

results, the failure step happens between steps four and nine. The fracture of raw CF/epoxy
TE

interface mostly happens in step six, leading to an average strain of 5.14 %. The desized
CF/epoxy tend to break at a lower step, mostly at the fourth and fifth step, leading to an
average strain of about 4.15 %. After grafting CNTs, the interaction between CNT-CF and
EP

epoxy recover, and the fracture mostly happens in sixth, leading to an average strain of
5.59 %. After oxidation, the interaction between oxidized CNT-CF and epoxy matrix is
C

further increased. The breaking point tends to happen at a higher step. Even for two samples
AC

of oxidized CNT-CF/epoxy composite, the fracture happens at the connection point or in the
epoxy matrix, rather than at the interface, which indicates that the interfacial strength between
oxidized CNT-CF and epoxy is larger than the fracture strength of epoxy matrix.

Since the stress-strain values of pure epoxy can be easily acquired by tensile test (as
shown in Fig. S7, the interfacial strength between CFs and epoxy can be estimated by fitting
the corresponding strain in the stress-strain curves, as shown in Fig. 7(B). The fracture
strength of epoxy is 66.5 MPa. The average strain of oxidized CNT-CF/epoxy is 6.82 %
before failure, corresponding to 58 MPa in the stress-strain curves. However, since several
 ACCEPTED MANUSCRIPT
samples broke in the matrix part, the actual interfacial strength of oxidized CNT-CF/epoxy is
possibly larger than 58 MPa. By using the same method, the interfacial strength of raw
CF/epoxy, CF-CNT/epoxy and desized CF/epoxy are 53, 55 and 48 MPa, respectively. These
values are close to the values reported by other groups [33]. There have three aspects which
result in the limited improvement over desized CF/epoxy composites. First, as observed in
figure 7B, calculated tensile strength variation in desized CF/epoxy composites is much

PT
smaller than in the three other composites. In intrinsic CF, CNT-CF and oxidized CNT-
CF/epoxy composites, there always exist low values at the same level as the desized

RI
CF/epoxy, which was due to improper testing interfaces (unmodified CFs). However, in the
results, due to the limitation of the tested samples, these values were also included to calculate

SC
the average value for each composite, which could lead to the reduce of the improvement.
Secondly, according to the XPS results in Figure S4, there is a small O peak in the spectrum
of desized CF. Even though the sizing layer has been removed by heating the CFs in an argon

U
atmosphere, an inevitable oxidation of the CF surface has happened, which could enhance the
AN
interfacial adhesion due to the existence of oxygen-containing groups. Thirdly, about the
failure behavior, two samples failed in the epoxy part for the oxidized CNT-CF/epoxy
M

composite. However, we have used the ultimate stress of epoxy to calculate the bond strength
of the oxidized CNT-CF/epoxy composite. The real value should be higher than that. All three
D

aspects lead to reducing the difference between them.

To understand the function of the CNT layer on the CF surface, the STEM combined
TE

with EDX has been employed to study the CF/Epoxy composite interface. This technique was
suitable for acquiring high-resolution images with a detailed elemental information and
EP

quantitative analyses on the interphase composition [34-37]. Four TEM lamellas were
prepared by FIB (Fig. S8). The CF orientation is parallel to the lamella plane. In the four
C

images, the CF appears as dark since the sample thickness is locally larger than in the epoxy
AC

because of the relatively higher hardness of CF compared to the epoxy matrix. Furthermore,
since CF, CNTs and epoxy resin are carbon-based materials, the contrast difference is weak in
bright field. The HAADF detector is preferable in STEM mode, which is known as the Z-
contrast configuration. It has two major advantages: one is that small differences in the
average atomic number or thickness of the material can be observed with a higher contrast
difference in the dark field image; the other is that the interested region can be focused to a
very small point (#0,1 nm in diameter), which helps to acquire high resolution EDX mapping.
Fig. 8 shows two STEM HAADF images of raw CF/epoxy and CNT-CF/epoxy, respectively.
 ACCEPTED MANUSCRIPT
For the raw CF/epoxy (Fig. 8A), the bright region corresponds to the CF part with a larger
thickness than the epoxy part while the dark part is the pure epoxy. The interface in between
is the obvious boundary. CNT-CF/epoxy is presented in Fig. 8B. A tubular structure can be
clearly observed by STEM under high resolution. After grafting CNTs on a CF surface, the
epoxy matrix can immerse the CNT layer and wet the CF surface completely without creating
voids in the interface region. The CNTs are dispersed in the matrix homogeneously. The

PT
average thickness of the CNT layer is around 440 nm, which has already been confirmed by
SEM image. The bright spots in the matrix domain should be the catalyst used for the CNT

RI
growth.

U SC
AN
M

Fig. 8 STEM Z-contrast (HAADF) image showing the interface region of the CF and epoxy:
(A) raw CF/epoxy and (B) CNT-CF/epoxy interfaces.
D

Fig. 9(A) presents a STEM image of a raw CF/epoxy interface. The green rectangle
TE

corresponds to the region selected for EDX analysis where three elements–oxygen, carbon
and chlorine–were investigated. Chlorine exists in the sizing layer but has migrated to the
EP

epoxy domain and has aggregated at the interface. For oxygen, there also exists an
aggregation at the interface, indicating that the oxygen-containing groups have a higher
C

density than in the epoxy matrix. For carbon, the aggregation in the CF domain is obvious.
AC
 ACCEPTED MANUSCRIPT

PT
RI
SC
Fig. 9 High magnification STEM Z-contrast (HAADF) image showing a CF perpendicular to
the epoxy foil prepared by FIB: Interfaces of (A) raw CF/epoxy, (B) desized CF/epoxy, (C)

U
CNT-CF/epoxy and (D) oxidized CNT-CF/epoxy, respectively. The regions marked out by
AN
the green rectangle were analyzed by EDX, the elements were marked out by different colors,
green for oxygen, blue for carbon, pink for chlorine and red for iron, respectively.
M

Fig. 9(B) shows a STEM image of a desized CF/epoxy interface. After the removal of
the sizing layer, there is no oxygen aggregation in the interface. From the CF to the epoxy
D

domains, the element presence changes gradually both for carbon and oxygen. Compared with
the raw CF/epoxy composite, the interface is more blurred.
TE

Fig. 9(C) and (D) correspond to the observation of CNT-CF/epoxy and modified
CNT-CF/epoxy composites, respectively. The CNT layers are distributed homogeneously on
EP

the CF surfaces. The bright spots near the interface correspond to the heavier iron catalyst
particles, which is confirmed by EDX. The CNT layer morphology can be figured out in the
C

epoxy matrix. In the case of the CF-CNT/epoxy interface, the elements also change gradually.
AC

However, in the oxidized CF-CNT/epoxy interface, an oxygen aggregation is observed. As far

as carbon is concerned, there is no element aggregation layer.

Grafting a layer of CNTs on the surface of CF is a common method to modify the

interaction between the polymer matrix and CFs. The effect of reinforcement is obvious since
the CNT layer can greatly increase the contact area at the interface. Moreover, the CNTs, with
their radial arrangement, can insert in the matrix and further increase the interfacial strength.
The CNT layer can also increase the conductivity of the entire composites.
 ACCEPTED MANUSCRIPT
After removing the sizing layer, the interaction between the CFs and the epoxy matrix
is greatly weakened due to the decrease of chemical bonds. After grafting CNTs on the
surface, the increased interfacial strength is obvious compared to the case of desized
CF/epoxy composites, since CNTs can largely increase the contact area between matrix and
CFs. Furthermore, the CNT layer, with its mechanical properties, could enhance the
surrounded matrix. The oxidized CNT-CF/epoxy composite presents the best interfacial

PT
strength of the four composites. The oxidized CNT-CF are fabricated by heating the sample at
400 °C for 4 min to increase the oxygen-containing groups on the surface. The obvious

RI
aggregation of oxygen at the interface can largely increase the interfacial strength between the
CFs and the epoxy matrix. The same reason could be used to explain the interaction between

SC
raw CFs, which contains a sizing layer to increase the interaction between CF, and epoxy
matrix. As we have mentioned before, increasing the surface roughness and depositing a
transition layer are two major methods to improve the interfacial properties. Here, oxidized

U
CNT-CF combine both methods, i.e. grafting CNTs increases the surface roughness and
AN
introduces oxygen-containing groups to increase the chemical bonding, which leads to the
best interfacial strength of the four types of composites.
M

4. CONCLUSION

In this work, an in-situ tensile test was conducted in an SEM to measure the interfacial
D

strength between carbon fibers and the epoxy matrix. Four types of CF reinforced epoxy
TE

composites–raw CF, desized CF, CNT-CF and oxidized CNT-CF–have been investigated.
Based on the results, it was found that grafting a CNT layer on the CF surface can modify the
EP

interfacial properties and reinforce the epoxy matrix surrounding the CF. After further
oxidation of CNT-CF, the interfacial strength can be further increased. Through a STEM and
EDX analysis, it was found that the interaction between oxygen-containing groups and epoxy
C

was the crucial factor in determining the interfacial strength. This knowledge can be used to
AC

guide the modification of other types of reinforcement to fabricate composites with excellent
mechanical properties.

5. SUPPORTING INFORMATION

This material includes the tensile test results of pure epoxy, XPS analyses of four
types of CF, TGA/DTG curves of CNT-CF, simulation work, SEM images of desized CFs
and oxidized CNT-CF, and TEM images of lamellae.
 ACCEPTED MANUSCRIPT
6. ACKNOWLEDGMENTS

This work was carried out within the MATMECA consortium and supported by the
ANR under contract number ANR-10-EQPX-37. It has benefited from the facilities of the
Laboratory MSSMat (UMR CNRS 8579), CentraleSupélec, France. We thank Dr. L. WANG
and Dr. Y. CHEN for the SEM sample preparation. Yu LIU thanks China Scholarship Council
for the fellowship.

PT
7. REFERENCES

RI
[1] Y. Gan, Effect of Interface Structure on Mechanical Properties of Advanced Composite
Materials, International Journal of Molecular Sciences 10(12) (2009) 5115.

SC
[2] J. Schultz, L. Lavielle, Interfacial Properties of Carbon Fiber—Epoxy Matrix Composites,
Inverse Gas Chromatography, American Chemical Society1989, pp. 185-202.

U
[3] J. George, M. Sreekala, S. Thomas, A review on interface modification and
characterization of natural fiber reinforced plastic composites, Polymer Engineering &
AN
Science 41(9) (2001) 1471-1485.
[4] J. Schultz, L. Lavielle, C. Martin, The role of the interface in carbon fiber-epoxy
M

composites, The Journal of Adhesion 23(1) (1987) 45-60.

[5] D. He, B. Fan, H. Zhao, X. Lu, M. Yang, Y. Liu, J. Bai, Design of Electrically Conductive
D

Structural Composites by Modulating Aligned CVD-Grown Carbon Nanotube Length on

TE

Glass Fibers, ACS Applied Materials & Interfaces 9(3) (2017) 2948-2958.
[6] E. Totry, J.M. Molina-Aldareguía, C. González, J. Llorca, Effect of fiber, matrix and
interface properties on the in-plane shear deformation of carbon-fiber reinforced composites,
EP

Composites Science and Technology 70(6) (2010) 970-980.

[7] S.-J. Park, M.-H. Kim, J.-R. Lee, S. Choi, Effect of Fiber–Polymer Interactions on
C

Fracture Toughness Behavior of Carbon Fiber-Reinforced Epoxy Matrix Composites, Journal

AC

of Colloid and Interface Science 228(2) (2000) 287-291.

[8] I.K. Ismail, M.D. Vangsness, On the improvement of carbon fiber/matrix adhesion,
Carbon 26(5) (1988) 749-751.
[9] Z. Dai, F. Shi, B. Zhang, M. Li, Z. Zhang, Effect of sizing on carbon fiber surface
properties and fibers/epoxy interfacial adhesion, Applied Surface Science 257(15) (2011)
6980-6985.
[10] R.L. Zhang, Y.D. Huang, D. Su, L. Liu, Y.R. Tang, Influence of sizing molecular weight
on the properties of carbon fibers and its composites, Materials & Design 34 (2012) 649-654.
 ACCEPTED MANUSCRIPT
[11] R. Sager, P. Klein, D. Lagoudas, Q. Zhang, J. Liu, L. Dai, J. Baur, Effect of carbon
nanotubes on the interfacial shear strength of T650 carbon fiber in an epoxy matrix,
Composites Science and Technology 69(7) (2009) 898-904.
[12] H. Qian, A. Bismarck, E.S. Greenhalgh, G. Kalinka, M.S.P. Shaffer, Hierarchical
Composites Reinforced with Carbon Nanotube Grafted Fibers: The Potential Assessed at the
Single Fiber Level, Chemistry of Materials 20(5) (2008) 1862-1869.

PT
[13] X. He, F. Zhang, R. Wang, W. Liu, Preparation of a carbon nanotube/carbon fiber multi-
scale reinforcement by grafting multi-walled carbon nanotubes onto the fibers, Carbon 45(13)

RI
(2007) 2559-2563.
[14] A. Warrier, A. Godara, O. Rochez, L. Mezzo, F. Luizi, L. Gorbatikh, S.V. Lomov, A.W.

SC
VanVuure, I. Verpoest, The effect of adding carbon nanotubes to glass/epoxy composites in
the fiber sizing and/or the matrix, Composites Part A: Applied Science and Manufacturing
41(4) (2010) 532-538.

U
[15] A.H. Barber, Q. Zhao, H.D. Wagner, C.A. Baillie, Characterization of E-glass–
AN
polypropylene interfaces using carbon nanotubes as strain sensors, Composites Science and
Technology 64(13–14) (2004) 1915-1919.
M

[16] E. Bekyarova, E.T. Thostenson, A. Yu, M.E. Itkis, D. Fakhrutdinov, T.-W. Chou, R.C.
Haddon, Functionalized Single-Walled Carbon Nanotubes for Carbon Fiber−Epoxy
D

Composites, The Journal of Physical Chemistry C 111(48) (2007) 17865-17871.

[17] M. Li, Y. Gu, Y. Liu, Y. Li, Z. Zhang, Interfacial improvement of carbon fiber/epoxy
TE

composites using a simple process for depositing commercially functionalized carbon

nanotubes on the fibers, Carbon 52 (2013) 109-121.
EP

[18] z.g. zhao, L. CI, H. Cheng, J. Bai, Growth of multi-walled carbon nanotube of different
morphologies on carbon fibers by floating catalyst method, Carbon 43(3) (2004).
[19] L. CI, J. Bai, Direct growth of aligned carbon nanotubes on the surface of ceramic fibers,
C

Carbon 43(4) (2005).

AC

[20] J. Zhu, J. Kim, H. Peng, J.L. Margrave, V.N. Khabashesku, E.V. Barrera, Improving the
Dispersion and Integration of Single-Walled Carbon Nanotubes in Epoxy Composites through
Functionalization, Nano Letters 3(8) (2003) 1107-1113.
[21] E. Bekyarova, E.T. Thostenson, A. Yu, H. Kim, J. Gao, J. Tang, H.T. Hahn, T.W. Chou,
M.E. Itkis, R.C. Haddon, Multiscale Carbon Nanotube−Carbon Fiber Reinforcement for
Advanced Epoxy Composites, Langmuir 23(7) (2007) 3970-3974.
 ACCEPTED MANUSCRIPT
[22] V.P. Veedu, A. Cao, X. Li, K. Ma, C. Soldano, S. Kar, P.M. Ajayan, M.N. Ghasemi-
Nejhad, Multifunctional composites using reinforced laminae with carbon-nanotube forests,
Nature materials 5(6) (2006) 457-462.
[23] A. Godara, L. Mezzo, F. Luizi, A. Warrier, S.V. Lomov, A.W. van Vuure, L. Gorbatikh,
P. Moldenaers, I. Verpoest, Influence of carbon nanotube reinforcement on the processing and
the mechanical behaviour of carbon fiber/epoxy composites, Carbon 47(12) (2009) 2914-2923.

PT
[24] L. Gao, T.-W. Chou, E.T. Thostenson, Z. Zhang, A comparative study of damage sensing
in fiber composites using uniformly and non-uniformly dispersed carbon nanotubes, Carbon

RI
48(13) (2010) 3788-3794.
[25] A.N. Netravali, R.B. Henstenburg, S.L. Phoenix, P. Schwartz, Interfacial shear strength

SC
studies using the single-filament-composite test. I: Experiments on graphite fibers in epoxy,
Polymer Composites 10(4) (1989) 226-241.
[26] S.C. Lee, S.T. Jeong, J.N. Park, J.K. Sun, G.J. Cho, A study on mechanical properties of

U
carbon fiber reinforced plastics by three-point bending testing and transverse static response,
AN
Journal of Materials Processing Technology 201(1–3) (2008) 761-764.
[27] C. DiFrancia, T.C. Ward, R.O. Claus, The single-fiber pull-out test. 1: Review and
M

interpretation, Composites Part A: Applied Science and Manufacturing 27(8) (1996) 597-612.
[28] B. Miller, P. Muri, L. Rebenfeld, A microbond method for determination of the shear
D

strength of a fiber/resin interface, Composites Science and Technology 28(1) (1987) 17-32.
[29] A. Godara, L. Gorbatikh, G. Kalinka, A. Warrier, O. Rochez, L. Mezzo, F. Luizi, A. Van
TE

Vuure, S. Lomov, I. Verpoest, Interfacial shear strength of a glass fiber/epoxy bonding in

composites modified with carbon nanotubes, Composites Science and Technology 70(9)
EP

(2010) 1346-1352.
[30] K.R. Totten, B. Kutub, L.A. Carlsson, In situ determination of the fiber–matrix interface
tensile strength, Journal of Composite Materials 50(5) (2016) 589-599.
C

[31] J. Bai, Procede de synthese de nanotubes de carbone sur materiaux micrometriques longs
AC

et particulaires, Google Patents, 2010.

[32] Y. Liu, A.-L. Hamon, P. Haghi-Ashtiani, T. Reiss, B. Fan, D. He, J. Bai, Quantitative
Study of Interface/Interphase in Epoxy/Graphene-Based Nanocomposites by Combining
STEM and EELS, ACS Applied Materials & Interfaces 8(49) (2016) 34151-34158.
[33] E. Mäder, S. Melcher, J.-W. Liu, S.-L. Gao, A. Bianchi, S. Zherlitsyn, J. Wosnitza,
Adhesion of PBO fiber in epoxy composites, Journal of materials science 42(19) (2007) 8047-
8052.
 ACCEPTED MANUSCRIPT
[34] W. Zhou, J. Lee, J. Nanda, S.T. Pantelides, S.J. Pennycook, J.-C. Idrobo, Atomically
localized plasmon enhancement in monolayer graphene, Nature nanotechnology 7(3) (2012)
161-165.
[35] C.B. Carter, D.B. Williams, Transmission electron microscopy, Springer-Verlag US2009.
[36] Q. Wu, M. Li, Y. Gu, Y. Li, Z. Zhang, Nano-analysis on the structure and chemical
composition of the interphase region in carbon fiber composite, Composites Part A: Applied

PT
Science and Manufacturing 56 (2014) 143-149.
[37] S. Yang, X. Feng, L. Wang, K. Tang, J. Maier, K. Müllen, Graphene‐Based Nanosheets

RI
with a Sandwich Structure, Angewandte Chemie International Edition 49(28) (2010) 4795-
4799.

U SC
AN
M
D
TE
C EP
AC