CARBON FIBER: PROPERTIES, TESTING, the mid-1960s [4,8,9], but the yield of the process was only

AND ANALYSIS about 15 wt%. Alternatively the use of PAN as a precursor

allowed the stretching to be carried out at lower tem-
SITI-ROS SHAMSUDDIN1, perature and increased the process yield to 50 wt%. The
KINGSLEY KIN CHEE HO2, starting point of the manufacturing process of both rayon-
KOON-YANG LEE1, JOHN M. and PAN-based carbon fibers is textile fibers (Fig. 1a). For
HODGKINSON1, AND pitch-based carbon fibers, petroleum pitch is converted
ALEXANDER BISMARCK1 into a mesophase pitch by a refining operation before
1 ImperialCollege London, spinning (Fig. 1b) [10].
London, UK The mechanical properties of PAN-based carbon fibers
2 Cytec Engineered Materials,
depend to a large extent on the mechanical properties of
Wrexham, UK the PAN precursor fibers. PAN-based carbon fibers are
commonly used as reinforcement in composite materials
used for applications in sporting goods, aerospace, the
INTRODUCTION
automotive industry, and so on. PAN was the original
precursor for high strength, that is, high performance,
Carbon fibers were first developed by Thomas Edison in
carbon fibers [11], and PAN-based carbon fibers still hold
1879 for use as filaments in early light bulbs. He took
the largest market share for carbon fibers [12]. This is
cellulose-based materials, such as cotton and bamboo,
mainly due to the following [13]:
and carbonized them at high temperature in a controlled
atmosphere [1]. This process, known as pyrolysis, is still
• The structure of PAN permits a faster rate of pyrol-
used today to manufacture carbon fibers [2,3]. Carbon
ysis and allows molecular chains to be aligned in the
fibers were produced for high temperature missile appli-
preferred orientation along the fiber axis.
cations beginning in the late 1950s and finally became
commercially available in the 1960s [4], representing a • PAN decomposes before melting.
breakthrough in production of high performance, low den- • PAN precursors can be stretched to up to 800% strain
sity materials. Today, carbon fibers offer a practically and further orientation improvement is possible dur-
attainable reinforcement material, which can be produced ing thermal stabilization at approximately 180◦ C and
in volume at reasonable cost [5] for use in fiber-reinforced through various postspinning modifications.
composites. • PAN provides high carbon yield of up to 50 wt% when
Perfect graphite possesses a theoretical Young’s modu- pyrolyzed at 1000◦ C and above.
lus and tensile strength of ∼1 TPa and 100 GPa, respec-
tively. However, due to the presence of defects on and The development of special grade PAN fibers is very
in carbon fibers, actual tensile properties are always an important to the entire carbon fiber industry. The first step
order of magnitude lower. The most perfect form of fibrous in preparing a PAN fiber precursor is to polymerize acry-
carbon up until the mid-1980s were graphite whiskers, lonitrile and spin it. PAN has highly polar nitrile groups.
with a measured Young’s modulus of ∼680 GPa and ten- As a result, strong dipole–dipole forces exist between adja-
sile strength of ∼20 GPa [6,7]. Today, the characteristic cent chains. These polar forces hinder molecular alignment
properties of carbon fibers are directly related to their during spinning. To improve alignment, 2–5% comonomer
manufacturing methods and can be measured using a is often added to acrylonitrile to form a copolymer of
variety of techniques. PAN [14]. The two most commonly used comonomers are
methylacrylate and itaconic acid. Each has a specific role.
Methylacrylate reduces glass transition temperature so
CARBON FIBER PREPARATION
that the fiber can be stretched in a steam environment
as opposed to a glycol environment. Itaconic acid acts as
Carbon fibers are manufactured commercially from three
an initiator for the formation of ladder polymers. Other
different precursors: rayon (regenerated cellulosic fibers),
comonomers such as acrylamide are also used. Once the
polyacrylonitrile (PAN), and petroleum pitch. Carbon fiber
homo- and copolymer have been selected, they are then
synthesis technology is protected very strongly by carbon
polymerized in aqueous sodium thiocyanate solution and
fiber manufacturers. However, the basic synthesis chem-
carefully filtered. The resulting PAN is then spun into
istry is known and a brief review is included in this
fibers.
article.
There are four spinning techniques. The most widely
used technique for spinning acrylic fibers is wet spinning.
Precursors for Carbon Fiber Manufacturing
In wet spinning, the acrylic precursor fibers are spun into
Carbon fibers were initially made by hot stretching rayon a coagulation bath that has a high percentage of solvent
during graphitization. An early commercially available (such as dimethyl sulfoxide or dimethylformamide) at low
rayon-based carbon fiber (Thornel® 50) achieved strength temperature. Fiber diameter depends on the spinneret
and modulus values of 2.76 and 345 GPa, respectively, in hole diameter, the let-off rate and the take-up velocity.

Wiley Encyclopedia of Composites, Second Edition. Edited by Luigi Nicolais and Assunta Borzacchiello.
© 2012 John Wiley & Sons, Inc. Published 2012 by John Wiley & Sons, Inc.
1
2 CARBON FIBER: PROPERTIES, TESTING, AND ANALYSIS

Textile Surface
Stretch Thermoset Carbonize Graphitize Sizing
fiber treatment
(a)

Figure 1. Schematic illustrations of

(a) PAN and rayon and (b) pitch car- Melt-spun
Stretch Thermoset Carbonize Graphitize Sizing
bon fiber manufacturing processes. fiber
Source: Adapted from Ref. 10. (b)

O2
Figure 2. Cyclization of PAN, which takes N N N N
place at 200◦ C in the presence of oxygen. N N N N

OH OH

Figure 3. Aromatization during stabilization of + O2 + 2H2O

PAN fibers in the presence of oxygen. N N N N N N

As the fibers are stretched during spinning, the trapped oxygen atmosphere) are important parameters that affect
solvent decreases the cohesive force among the nitrile the final structure and hence the ultimate mechanical
groups of the polymer chain, reducing the dipole–dipole properties of the fiber. If no load is applied to the
interaction [15]. This allows for better chain orientation spin-stretched PAN fibers during stabilization shrinkage
and mechanical properties of the carbon fibers. Stretching occurs [11,24–26]. It is believed that there are two types
also results in uncoiling of the polymer chains and there- of shrinkage: physical and chemical. Physical shrinkage
fore enables formation of ladder polymers. In general, is attributed to the entropy recovery of a drawn and
the higher the spin-stretch factor (defined as the ratio quenched polymer [27]. This is due to the randomly
of windup speed to the filament speed at the spinneret coiled conformation of the PAN molecule and intermolec-
hole [16]), the higher the tensile modulus of the resulting ular repulsive forces between the neighboring nitrile
carbon fibers [17]. group dipoles. By stretching the fibers during precursor
Thermal Stabilization of PAN Fibers preparation, conformational change is developed, which
elongates the isotactic regions allowing ladder polymer
The second step in processing PAN fibers into carbon formation to occur. It is the length of the isotactic
fibers is thermal stabilization by oxidation in air at a tem- regions, which determines the graphitic nuclei formed.
perature of approximately 220◦ C. Stabilization involves Chemical shrinkage, on the other hand, occurs as a
cyclization, dehydrogenation, and oxidation. Cyclization result of chemical reactions during stabilization. This
(Fig. 2) is a critical step in processing carbon fibers and is due to formation of cyclized ladder polymers [28,29]
takes place at around 200◦ C [18] via a radical reaction
and may be considered a measure of cyclization reaction
for PAN homopolymer and an ionic mechanism for PAN
[30]. An optimum heating rate for acrylic precursors
copolymer [8]. The fibers are kept in tension, while cycliza-
of 1–3◦ C/min up to 260–280◦ C is reported to result in
tion takes place so that the molecular orientation induced
maximum tensile strength of the resulting carbon fibers
during stretching is maintained [9].
During thermal stabilization, the color of the precursor [31].
fibers changes from white to yellow, to brown and ulti- Stabilization of precursor fibers has been studied in
mately to black. Berlin et al. [19] and Fester [20] believe various media [32]. It was found that precursor fibers sta-
that the color change is due to the development of a polyene bilized in an oxidizing medium produce better carbon fibers
structure during heat treatment of PAN. Others [21–23] than those stabilized in an inert atmosphere [33,34]. Fitzer
suggested that it is because of the formation of a condensed and Muller [35] demonstrated that oxygen initiates the for-
ring structure containing carbon–nitrogen double bonds mation of active centers for cyclization and also retards
(Fig. 3). stabilization by increasing the activation energy. Never-
The application of tension to the fibers, the heating theless, stabilization in an oxidizing medium is preferred
temperature during stabilization, and the medium (i.e., because it forms oxygen-containing groups such as —OH,
 CARBON FIBER: PROPERTIES, TESTING, AND ANALYSIS 3

N N N

– H2
N N N
Figure 4. Dehydrogenation: first stage of car-
N N N bonization at 400–600◦ C.

C=O, and —COOH, which helps generate cross-links dur- a relatively low heating rate of less than 5◦ C/min to
ing carbonization [36]. Fitzer and Muller [35] also pointed remove water and volatiles (Fig. 4 for reaction scheme,
out that PAN fibers can be stabilized at higher temper- adapted from Ref. 39). This slows down the mass transfer.
atures of 300–400◦ C at a very fast heating rate, in the If the heating rate is too high, it causes surface irregu-
order of 200–300◦ C/min. This leads to formation of a larities in the form of pores because of the diffusion of
highly ordered and compact structural arrangement in evolved gases [14]. In the second heating zone between
the stabilized fibers. The mechanical properties of the 600 and 1500◦ C, a higher heating rate is used because the
resulting carbon fibers show significant improvement over exothermic reactions and evolution of by-products occurred
those stabilized at temperatures of 200–300◦ C. This has already in zone 1 (see Fig. 5 for reaction scheme, adapted
considerably improved the economics of carbon fiber pro- from Ref. 39). However, other gases such as nitrogen
duction. and hydrogen cyanide as well as hydrogen and water are
produced by intermolecular cross-linking of the polymer
Carbonization of PAN-Based Fibers chains (Fig. 6 for reaction scheme, adapted from Ref. 30).
Stabilized precursor fibers are converted into carbon fibers During cross-linking, the carbon atoms of one cyclized
in the process of carbonization. This involves heat treat- sequence fit in to the spaces left by the nitrogen of the
ment of PAN fibers in an inert atmosphere [13]. This adjacent sequence [40]. This helps the formation of the
process further eliminates impurities by removing all ele- graphite structure in the lateral direction. An inert atmo-
ments other than carbon and leads to the formation of sphere is needed to avoid oxidation of the fibers at such
a graphitelike structure. The carbonization temperature high temperature. Common gases such as nitrogen, argon,
directly affects the graphitic structure of the produced car- or other nonoxidizing media such as HCl and ZnO have
bon fibers. This finding dates back to work in the 1960s been used [41].
by Watt et al. [37] at the Royal Aircraft Establishment The strength of a carbon fiber is largely associated with
(RAE). They discovered that application of tension dur- defects in its structure. Many of the defects in the bulk fiber
ing the initial oxidation step at 220◦ C was important arise from catalyzed graphitization of impurities present
for maintaining or even enhancing molecular alignment in the original PAN precursor. These impurities may be
during carbon fiber processing to 2500◦ C [37,38]. Heat dust particles or polymer gels, which have not been filtered
treatments in excess of 2500◦ C were found to result in high out of the dope before spinning the precursor. A decrease
modulus fibers, which were subsequently designated type in the tensile strength of the carbon fibers is reported
I fibers (structural models of high modulus PAN-based car- for heat treatments above 1500◦ C [38,42]. However, the
bon fibers can be found in Ref. 7). The RAE workers also tensile modulus keeps increasing at a slow rate up to
established that an optimum treatment of 1000–1500◦ C temperature of 3000◦ C [43], which results in fibers having
yields high strength carbon fibers. These high strength crystallite orientation mainly parallel to the fiber axis over
fibers are designated as type II carbon fibers. a long distance. It is noted that nitrogen cannot be used as
In the carbonization process of Type II fibers, the heat- a medium above 2000◦ C because nitrogen becomes active
ing rates in the two temperature zones are crucial. The and forms cyanogen (CN2 ) by reacting with carbon fibers
first zone with a temperature of up to 600◦ C requires [30].

N N N

N N N

–N2

N N N

Figure 5. Further denitrogenation: second stage of

N N N carbonization at 600–1500◦ C.
4 CARBON FIBER: PROPERTIES, TESTING, AND ANALYSIS

N N N N
N N N N

C C –H2O
H2 H2
O O O O
H2 H2
C C
N N N N
N N N N

N N N N
N N N N

H H OH H H OH –H2O
OH H H OH H H

N N N N
Figure 6. Intermolecular cross-linking of stabilized
N N N N fibers through oxygen-containing groups.

Carbon Fiber Surface Treatment most carbon fibers are intended for epoxy resin-based
composites) [47,48]. Normally, around 0.5–2 wt% resin of
After carbonization in the furnace, the surface of the car-
an epoxy is added to the fibers, which corresponds to a
bon fibers is treated. Kaelble et al. [44], Larsen et al. [45],
coating thickness of approximately 0.03 μm.
and Daukeys [46] have shown that when carbon fibers
without surface treatment are used to produce carbon
fiber-epoxy composites a low interlaminar shear strength ANALYZING THE PHYSICAL PROPERTIES OF CARBON
(ILSS) is observed. They also observed that the ILSS FIBERS
is directly related to fiber–matrix adhesion. Therefore,
the adhesion between the untreated carbon fibers and an Morphology of Carbon Fibers
epoxy matrix is weak. It was also observed that as the tem-
Scanning electron microscopy (SEM) allows the surface
perature of graphitization increases, the ILSS decreases,
morphology of carbon fibers to be investigated [49]. SEM
although the Young modulus of the carbon fibers increases.
produces images by scanning a high energy beam of
This prompted the development of a number of sur-
primary electrons across a sample in a raster pattern.
face treatments such as anodization, solution, and high
The raster scanning pattern is similar to that used in
temperature oxidation in order to improve fiber–matrix
a cathode-ray tube (CRT) television sets in which the
interfacial bonding. The specific details of these treat-
electron beam is (i) swept across the surface linearly in
ments have been reviewed by Donnet et al. [9]. As produced
the x direction, (ii) returned to its starting position, and
PAN-based carbon fibers contain many exposed graphitic
(iii) shifted downward in the y direction by a standard
edge planes at their surface, which are more reactive
increment. This process is repeated until the desired area
than the basal planes of graphitic crystals. Surface reac-
of the surface has been scanned. Signals are received
tions take place preferentially at these less-ordered edge
above the surface, in the z direction, and stored digi-
plane regions. The surface exposed edge planes of a treated
tally. These signals are converted to three-dimensional
PAN-based carbon fiber-epoxy composites typically exhibit
depth profile images that are ideal for surface structural
much better adhesion, possibly as a result of higher chem-
studies. Several types of signals are produced from a sur-
ical reactivity of the graphitic edge planes. It should be
face in this process, including back scattered, secondary,
noted that the strong anodic treatments of fibers are able
and Auger electrons [50,51] allowing for image quality
to induce degradation and/or decrease fiber modulus by
enhancement. Gold or chromium coating is typically not
disrupting the graphitic structure of the fibers. Typical
required for analysis of carbon fibers, as they are already
anodic treatments of fibers can be performed in aque-
electrically conductive. A scanning electron micrograph
ous sodium hydroxide, potassium hydroxide, sulfuric acid,
of carbon fibers is shown in Fig. 7, where the surface
nitric acid, and solutions of amine salts.
morphology of the fibers can be seen clearly.

Sizing of Carbon Fibers Determination of Diameters of a Single Carbon Fiber

In the last stage of carbon fiber manufacturing, a thin layer An accurate fiber diameter measurement method is impor-
of a size is typically coated onto the fibers to (i) protect tant in order to determine the mechanical properties of
the fibers from possible damage during transportation carbon fibers and to demonstrate the interfacial prop-
and handling and (ii) to improve interfilament adhesion erties in a model composite. The most practical means
and aid in wetting out the fiber in resin matrices (as for measuring diameter is using laser diffraction [52,53].
 CARBON FIBER: PROPERTIES, TESTING, AND ANALYSIS 5

ANALYZING THE SURFACE PROPERTIES OF CARBON

FIBERS

Surface Energy of Carbon Fibers

The wetting properties of solid materials are of both funda-
mental and practical importance in material performance
and are governed by the solid’s surface energy. Although
numerous techniques have been employed to characterize
the surface tension of solids, the indirect wetting tech-
nique, or contact angle measurement, remains one of the
most commonly used. Other techniques such as inverse
gas chromatography (IGC) can also be used for determin-
ing surface energy properties [54]. As carbon fibers do
not fall into that category, it is possible to determine their
surface energy indirectly using the wetting technique. The
wetting behavior of fibers in different test liquids (typically
Figure 7. Scanning electron micrograph of Hexcel AS4 fibers.
water, formamide, and diiodomethane) can be quantified
Image courtesy of Henry Maples, Imperial College London. gravimetrically using the previously described method for
fiber diameter determination, and if the fiber perimeter
is known the contact angle can be calculated using the
following equation:
This method uses a low power helium–neon laser beam
(0.8 mW). Two lenses, one convex lens to focus the laser mg
beam and one concave lens to reparallelize the beam, are cos θ = . (3)
Pf γlv
used to reduce the beam diameter. A polarization filter and
a blind are used to restrain the reflection from the lenses. If the investigated carbon fiber is more hydrophilic
The carbon fiber sample is set perpendicular to the laser (θ  90◦ ), a positive mass change will be detected as
beam. A photodiode on a movable rail is used to record the fibers are immersed in water. A typical measure-
the interference pattern. The movable rail is driven by an ment for hydrophilic fibers is illustrated in Fig. 8a. On
analog y−t writer. The fiber diameter (d) is determined the other hand, a negative mass change will be detected
using the following equation: if hydrophobic (θ  90◦ ) carbon fibers are immersed in
 water (Fig. 8b). This technique is very useful in the rapid
 2 
2L quantification, for example, of whether a surface modifica-
d = nλ 1 +
2
, (1) tion of carbon fibers has been successful, since an increase
nZ
in hydrophobic/hydrophilic functionalities with a covalent
bond character results in a change in the fibers’ surface
where λ is the laser wavelength (in nm), L the fiber-diode
wettability [55].
distance (in mm), n the number of node, and z half the
The Young–Dupré equation links the thermodynamic
node-to-node distance. This has proven to be a quick and
work of adhesion (Wa ) [56], that is, the reversible work
easy method to measure fiber diameter with an accuracy
required to separate a unit area of interface of two different
of 0.1 μm [52].
phases at equilibrium, with the wettability, that is, the
In addition to laser diffraction, the diameter of a
contact angle (θ), and the surface tension of the wetting
single carbon fiber can be evaluated using wetting mea-
phase (γlv ) as follows:
surements; the modified Wilhelmy method is used to
determine the wetting force of filaments of fibers as they
Wa = γlv (1 + cos θ ). (4)
are fully wetted and dewetted. The liquid used in this
method must have a low evaporation rate and a low sur-
face tension allowing the fibers to be completely wetted A range of comprehensive solid surface energy models
(θ = 0◦ , cos θ = 1), for example, n-dodecane. The absence have been developed, all of which focus on the estimation of
of any contact angle hysteresis during the measurements the solid surface energy (γs ) via physically acceptable com-
with low surface tension liquids is a good indication of com- bining rules in order to express the solid/liquid interfacial
plete wetting. The wetted fiber perimeter (Pf ) is calculated tension (γsl ) defined as follows:
using the following equation:
γsl = γs + γlv − Wa . (5)
mg
Pf = , (2)
γlv Two approaches exist to determine surface tensions
from measured contact angles, namely, the surface energy
where m is the mass change due to wetting, g the acceler- component [57–60] and the equation of state approach,
ation due to gravity, and γlv the surface tension of the test developed by Neumann and coworkers [61], respectively.
liquid used. A typical curve for mass gain of fibers due to A more recent version of the surface tension component
wetting/dewetting is shown in Fig. 8. approach, the acid/base or AB approach, was developed by
6 CARBON FIBER: PROPERTIES, TESTING, AND ANALYSIS

1.2
1.0

0.8 1.0

m (mg)
θr

m (mg)
0.6 0.8
θr
0.4
0.6 θa
θa
0.2
0.4
Figure 8. Representative mass gain as a 0.0
function of time of a single hydrophilic 0 200 400 600 800 1000 0 200 400 600 800 1000 1200
(left) and hydrophobic (right) carbon fiber
t(s) t(s)
immersing into and withdrawing from
water. (a) (b)

van Oss et al. [62–66]. This approach separates nonspe- The EoS approach assumes that the relationship
cific physical, such as dispersive interaction from direct between cos θ and γlv is universal, that is, the con-
chemical interactions, namely, Lewis acid–base (electron tact angle (θ) is solely determined by the solid (γs )
donor/acceptor) interactions. It assumes that the overall and the liquid (γlv ) surface tensions and is independent of
surface tension of a compound is the sum of the physi- the type and strength of the molecular interaction across
cal interaction components of the surface tension (γ LW ) the liquid/solid interface. The surface energy of a solid
(Lifshitz/van der Waals) and the chemical interaction or (γs ) can be calculated by measuring the contact angle of
acid/base component (γ AB ) of the surface tension: any liquid with a known surface tension on a solid by
combining the Young–Dupré equation with Neumann’s
γs = γsLW + γsAB . (6) (semiempirical) EoS for the work of adhesion:
√ 2
Since acid/base interactions are intrinsically specific Wa = 2 γlv γs e−β(γlv −γs ) , (10)
and depend on the complementary availability of corre-
sponding sites at the interface, the AB term is subdivided where β is a constant; β = 0.000115 (m2 /mJ)2 .
into an acid or electron acceptor (γ + ) and a base or electron The Wu surface tension model [68] for estimating the
donor (γ − ) parameter of the surface tension: surface free energy of carbon fibers (γs ) is calculated using
 the harmonic mean equation:
γsAB = 2 γs+ γs− . (7) p p
4γsd γld 4γs γl
γlv (1 + cos θ) = + p p, (11)
The LW component of the solid surface energy can be γsd + γld γs + γl
determined by measuring the contact angle of completely
non-polar (or dispersive) liquids, such as diiodomethane, where γlv and γs are the surface tension of the liquid
using and the surface free energy of the solid, respectively. This
approach is widely used to characterize the surface free
γlv (1 + cos θ)2 energy of low energy solids. The sum of γ d and γ p gives a
γsLW = . (8)
4 good approximation of the total surface free energy (γs ):
The acid/base approach combines all symmetrical and γs = γsp + γsd . (12)
non-specific forces in the LW component and separates
specifically and complementarily acting forces in the AB
component. Therefore, the total surface energy can be
determined by measuring the contact angles of at least Inverse gas chromatography (IGC) as a Tool to Evaluate
three test liquids with known surface tension components Surface Energy of Carbon Fibers
and parameters using This section describes the governing principles of IGC.
    For more IGC results of carbon fibers, the readers are
γlv (1 + cos θ ) = 2 γsLW γlLW + γs+ γl− + γs− γl+ . (9) referred to the article Carbon Fiber: Surface Proper-
ties. IGC is a robust and sensitive technique that requires
There are limitations associated with each surface only a small amount of sample for the determination of
energy model. As an example, due to the presence of surface energy. It is also easy to use IGC to ascertain
covalent C—F bonds in fluorinated carbon fibers where both the dispersive components of the surface energy and
the water contact angle exceeds 90◦ [67], the AB approach acid–base properties of the material under investigation
does not provide valid results; negative acid γ + parame- [69]. Only a small amount of probe gas is injected into
ters are calculated. Therefore, other methods such as Wu’s an inert carrier gas and its adsorption on the sample
and Neumann’s equation of state (EoS) approaches can be normally occurs in a linear relation to the concentration
used to determine the surface energy (γs ) of hydrophobic of the injected gas (Henry’s law region). The retention
carbon fibers. behavior of these injection pulses is recorded in the form
 CARBON FIBER: PROPERTIES, TESTING, AND ANALYSIS 7

of a net retention volume (Vn ), which is a fundamental Following Schultz and Lavielle [70], it is assumed that
surface thermodynamic property of the solid–vapor inter-
action process. When IGC measurements are carried out HAAB = KA DN + KD AN, (18)
at infinite dilution, the free energy of adsorption per mole
of probe (GA ) is given by the following equation [70]: which may be rearranged to

GA = −RT ln(Vn ) + C, (13) HAAB DN

= KA + KD . (19)
AN AN
where Vn is the net retention volume of the probe, C is
a constant, and R and T are the gas constant and tem- From Equation (18), the values of KA and KD may be
perature, respectively. The surface energy of a material AB
HA
can be described by the sum of a dispersive component, calculated by plotting against DN . The AN values,
AN AN
accounting for the capacity of the surface to exchange Lon- however, were later corrected by Riddle and Fowkes [77]
don or dispersive interactions, and a specific component, to include the dispersive force contributions (ANd ). The
referring to all other possibilities of interaction. The dis- modified acceptor number (AN∗ ) is then given by
persive component is called the non-specific component of
the surface energy, since London interactions always occur AN∗ = 0.288(AN − ANd ). (20)
irrespective of the partners brought into contact. The spe-
cific component cannot be simply measured. Among the To measure the surface energy of fibrous samples, the
specific interactions, acid–base interactions seem to play fibers are typically packed into presilanated glass columns
a key role in the interfacial interactions between the com- with silanated glass wool end frits. A series of alkane
ponents of a composite [59,71]. Dorris and Gray [72] have vapors are then used as probes to determine the dispersive
suggested that the key link between the change in free surface free energy of the fibers. Methane is used as a
energy of adsorption and the work of adhesion (WA ) is non-interacting probe to determine the dead-time of the
given by packed column. The injected probes are usually at infinite
dilution (4%) for a peak maximum analysis. The dispersive
−GA = aNA WA , (14) component of the surface energy can be calculated with
Equation (15) following the method proposed by Schultz
where a is the area covered by an adsorbed probe molecule and Lavielle [70]. The donor and acceptor numbers for the
and NA is the Avogadro’s number. When non-polar probes polar probes (such as ethanol, acetonitrile, ethyl acetate,
(e.g., n-alkanes) are used, only dispersive interactions and acetone) are taken from Gutmann [75] and Riddle and
occur with the solid surface. The resulting work of adhe- Fowkes [77] to evaluate the proton donating/accepting
sion is defined as ability of the fibers.
 1/2
WA = 2 γsd γld , (15) Brønsted Acid–Base Properties of Carbon Fibers
Interfacial phenomena are important in almost every
where γsd and γld are the dispersive components of the
industrial process, from heterogeneous catalysis to the
surface energy of the substrate and the probe, respectively
manufacturing of composite materials and from environ-
[73]. Combining Equations (12)–(14) leads to
mental protection to medical technology. Therefore, knowl-
1/2 1/2 edge about the presence of reactive functional surface
RT ln Vn = 2NA γsd a γld + C. (16)
groups is of general interest. Measuring the electrokinetic
or zeta-potential (ζ -potential) allows such information to
Therefore, for alkanes, a plot of RT ln(Vn ) against
1/2 be obtained.
a γld results in a linear function (Eq. 15). The dispersive The method of measuring the ζ -potential depends on
component of the substrate (γsd ) can thus be determined the nature of the material. For solid substrates, such as
from the slope of the plot. Several semiquantitative meth- carbon fibers, the most suitable method for measuring
ods can be used to estimate the acid–base characteristics ζ -potential is the streaming potential technique. If a solid
of the substrate [73,74]. A widely used method is calcu- is brought into contact with an aqueous electrolyte, the
lating the acceptor and donor constants (KA and KD ) of a formation of an interfacial charge causes a rearrangement
substrate using its adsorption enthalpy due to acid–base of the local free ions in the solution producing a thin
interactions with reference polar probes and the donor region of non-zero net charge density near the interface
and acceptor number (DN and AN) concept proposed by (Fig. 9) [78,79]. The arrangement of the charges at the
Gutmann [75]. The use of this method has been described solid–liquid interface and the balancing counter ions in the
in detail by Saintflour and Papirer [76] and Schultz and liquid is referred to as the formation of an electrical double
Lavielle [70]. The adsorption enthalpy due to acid–base layer (EDL). The EDL is generally viewed as consisting of
interactions can be calculated using
two layers: an immobile layer of partially hydrated ions
  opposite in sign to that of the surface (Helmholtz layer)
−GAB
A
d T and a diffuse layer stretching out into solution (Gouy
d(ln Vn )
−HAAB = =R . (17) layer) [80,81]. If a relative movement between the solid
d T1 d T1 and electrolyte is included, parts of the diffusive double
8 CARBON FIBER: PROPERTIES, TESTING, AND ANALYSIS

0 Distance 20

ζ plateau Basic
ψ
–
10
ζ Nonpolar

ζ -Potential (mV)
ψOHP
–
0
Shear plane

– –10
Diffuse layer
ψ IHP ζplateau
Acidic
OHP
–20
IHP Immobile layer 2 4 6 8 10
pH
Figure 9. Schematic illustration of an electrochemical double
layer—Gouy–Chapman–Stern–Graham (GCSG) model. Source: Figure 10. Schematic illustration of pH dependence of
Obtained from Ref. 79 with permission from Elsevier. ζ -potential.

layer are removed, which causes a potential difference a ζ = f (pH) curve without a plateau due to the lack of
across the sample dissociable groups. For these solids, the increase in nega-
  tive/positive ζ -potential with increasing/decreasing pH is
Es εr ε0 ζ 1 solely caused by the increasing adsorption of H3 O+ /OH−
= , (21)
P μ λb + 2λs /h ions. [84]. The isoelectric point (iep) is the pH at which
ζ = 0 mV. Determination of the iep allows the acidity or
where ζ is the ζ -potential, which is the electrical poten- basicity of a solid surface to be measured [56]. The iep
tial at the hydrodynamic plane of shear between the solid is determined by the average of all pKa and pKb of all
and the aqueous electrolyte, P is the pressure difference functional groups and the interface. The position of an iep
across the cell, Es is the streaming potential, ε0 , εr , λb , λs , at low pH indicates an acidic character of the solid surface
and h are the relative dielectric constant, electrical permit- and the position of an iep at high pH signifies that the
tivity of a vacuum, bulk conductivity, surface conductivity surface contains dissociable basic surface groups.
of the sample, and height of the channel, respectively. ζ -Potential measurements allow the state, type, and
If a relative movement between the solid and the elec- amount of dissociable surface functional groups on solid
trolyte solution is induced (in the streaming potential surfaces of interest to be estimated. Moreover, the disper-
method this is achieved by pumping a liquid through the sion forces proceed from solid surfaces to be characterized
packed fiber bed), the ions in the diffusive part of the by measuring ζ = f ([C]) [83]. It has been shown that the
EDL are removed at a shear plane at the border between difference in the adsorption free energies of electrolyte ions
the Helmholtz immobile layer and the diffusive layer, at the solid surface corresponds to the dispersive forces
resulting in an electrical potential difference (i.e., stream- occurring at the solid/electrolyte interface [85]. ζ -Potential
ing potential) across the sample, which can be measured measurements are a straightforward and reliable tool to
[82]. For a detailed theoretical description of the theory predict adhesive properties [86]. Bismarck et al. [87] have
of EDL and ζ -potential, please refer to discussion of the shown that an increasing difference in the ζ -potential
Helmholtz–Smoluchowski equation [83]. plateau values obtained from ζ = f (pH) measurements
The acidic and basic surface character of a material between composite components, that is, between the rein-
can be identified based on the shape of ζ = f (pH) curve forcement and matrix, also correlates with an increase in
as shown in Fig. 10. For solids containing acidic and/or the measured practical adhesion strength [87,88].
basic functional groups, these are dissociated according
to their pKa, b in the aqueous electrolyte. At low pH, the
Surface Area of Carbon Fibers
dissociation of acidic surface groups is suppressed; when
pH increases, the increased dissociation of acidic surface Brunauer–Emmett–Teller (BET) N2 adsorption is an
groups causes an increase in the negative ζ -potential. In analytical technique commonly used for determining the
the case of basic groups, at high pH, the dissociation of specific surface area of solids [89]. It has been used for
basic groups is suppressed. When pH decreases, the num- the characterization of carbon fibers [90,91]. During the
ber of positively charged groups increases. A complete measurement, N2 molecules are weakly adsorbed onto the
dissociation of acidic or basic functional groups results in fiber surface until saturation. The overall surface area (As )
a plateau in the ζ − f (pH) curve. Non-polar solids have is determined from the total number of gas molecules that
 CARBON FIBER: PROPERTIES, TESTING, AND ANALYSIS 9

Detector
X-ray

hv “Free electrons”
e– KE
hv Vacuum level
Electron Valence orbitals
energy BE
Substrate 2s Core orbitals Figure 11. Schematic illustration of the interaction
between an X-ray beam and a substrate in an XPS
(a) (b) system.

are adsorbed to the surface at a given partial pressure the matrix constituent atoms (Fig. 11b) [101–103]. There-
using the following equation: fore, very few electrons from a depth greater than two or
three times the IEMP will escape the bulk material.
P/P0 1 (C − 1)(P/P0 ) Complications due to charging of insulating samples
= + , (22) during XPS are not an issue with virgin carbon fibers
na (1 − P/P0 ) nm C nm C
because they are conductive so their surface does not
where P/P0 is the relative pressure, na the number of become positively charged. Charging effect is only common
gas molecules adsorbed, and nm the specific monolayer when the sample is an insulator because in that case
capacity [92]. more electrons are leaving the surface region than are
replaced by those from the bulk. Therefore, in the case
of virgin carbon fibers, it is uncommon to observe any of
ANALYZING THE SURFACE CHEMICAL COMPOSITION OF the usual effects due to charging such as the peak shifting
CARBON FIBERS to lower kinetic energies in XPS spectra [104]. During
a measurement, all of the electrons are collected by the
X-Ray Photoelectron Spectroscopy detector and transferred from the KE scale to the BE scale
using Equation (24) (more detailed information on XPS
X-ray photoelectron spectroscopy (XPS) is a technique that can be found in Refs [105–107]). A plot of the intensity
allows the composition of the outermost few atomic layers in counts per second (CPS) of emitted photoelectrons of a
of solid surfaces to be analyzed. XPS has been used to char- given KE versus the associated BE gives a XP spectrum.
acterize the surface composition of carbon fibers [93,94]. A typical XP spectrum and high resolution XP spectra of
XPS was judged by Briggs and Seah [95] as the most use- plasma fluorinated carbon fibers are shown in Figs. 12
ful technique for modern surface analysis with a detection and 13, respectively.
limit of about 0.5%. An extensive discussion of the basic The elemental composition of carbon fiber surfaces can
principles involving XPS is beyond the scope of this article. be determined from the peak positions corresponding to
For a detailed explanation, the readers are referred to the the ionization energies of the elements present, while
literature [95–100]. In general, XPS is a surface-specific, the area under the peak (normalized using appropriate
nondestructive, and surface-sensitive technique to deter- sensitivity factors) gives the concentration of the ele-
mine the chemical composition of solids by measuring the ments’ atmospheric percentage. This allows for quanti-
kinetic energy (KE) distribution of emitted photoelectrons: tative analysis of the surface compositions of solids [108].

A + hν → A+ + e− , (23)
× 103
KE = hν − BE, (24)
35

where A is an atom within the material, hv is the energy 30

of the X-ray beam, A+ is the ionized atom, and e− is
25 C 1s
the emitted photoelectron. In Equation (24), KE is the
kinetic energy and BE is the binding energy, also known 20
CPS

as the ionization energy. Electrons may be ionized from any F 1s O 1s

orbital with a binding energy less than the photoenergy of 15
the X-ray beam.
10
During XPS measurement, a surface is exposed to
Cl 2p
a monochromatic X-ray beam under vacuum (Fig. 11a), 5
which results in ejection of excited valence or core elec-
trons from an atom. The electrons are ejected from a
1200 1000 800 600 400 200 0
sample volume near the surface via an inelastic mean free
Binding energy (eV)
path (IEMP) of about 0.5–3 nm. The IEMP is defined as
the average distance that an electron will have to travel Figure 12. Survey XPS spectra (exemplary shown) of fluorinated
before it undergoes an inelastic collision with another of carbon fibers.
10 CARBON FIBER: PROPERTIES, TESTING, AND ANALYSIS

Cgraphitic/C-C

C-Ox
C-Fx
C 1s XP intensity 8 min
8 min

F 1s XP intensity
C-Hx
4 min
4 min

1 min
1 min

0 min
0 min

294 292 290 288 286 284 282 280 692 690 688 686 684 682
Binding energy /eV Binding energy /eV
(a) (b)

O = C-R
HO-RC = O
8 min
O 1s XP Intensity

4 min

1 min
OH/R-O-R

0 min

538 536 534 532 530 528 526

Binding energy /ev
(c)

Figure 13. High resolution XP spectra of plasma fluorinated T700 fibers. (a) C1s XP spectra,
(b) F 1s XP spectra, and (c) O 1s XP spectra. Source: Obtained from Ref. 67 with permission from
Elsevier.

Furthermore, high resolution XP spectra of the individual photon absorption requires the energy of the incident pho-
elements of interest in survey XP spectra allow the chem- ton to be equal to the energy difference between the two
ical state, that is, the bond environment, of the species to states of the molecule. The two states are accompanied by
be investigated. a change in the dipole moment of the molecule. This leads
to the molecule being excited from the vibrational state
to a virtual state. The virtual state system is very short
ANALYSIS OF THE BULK PROPERTIES OF CARBON FIBERS lived. The molecule relaxes from this virtual state by the
emission of photons. The Raman effect is the exchange in
D to G Ratio of Carbon Fibers energy between the incident and emitted radiation caused
Raman spectroscopy is a vibrational spectroscopic tech- by the inelastic scattering of photons [109].
nique; that is, it is defined by a non-zero change in Two prominent Raman-active modes can be observed in
polarizability during a vibration. When photons shine on a Raman spectra of carbon fibers. The Raman band at 1580
molecule, they can be absorbed or scattered. The process of cm−1 corresponds to E2g mode [110]. This band has been
 CARBON FIBER: PROPERTIES, TESTING, AND ANALYSIS 11

Power
supply

Polycarbonate Amplifier
backing rail

Four-point contact
device made out of Single carbon fiber on
brass PTFE substrate
0.00 Amps Figure 14. Schematic of the four-point method
for determining the conductivity of carbon fibers.
Source: Adapted from Ref. 118.

identified as the deformation (D) mode of the hexagonal six-member carbon ring. These electrons are free to move
C6 carbon ring structure with E2g symmetry, commonly and so are able to conduct electricity through the interlayer
observed in graphite crystal layers [111]. Polycrystalline sheets of the graphite structure. The four-point method
graphitic mode (G) exhibits an additional peak at 1355 can be used to determine the electrical conductivity of
cm−1 , corresponding to the A1g mode and is related to single carbon fibers, which eliminates problems of contact
the graphitic mode. The ratio of the intensities of the resistance (Fig. 14) [117]. For the measurement, a single
two observed bands at around 1355 cm−1 (G) and 1580 carbon fiber is selected and placed onto an insulating
cm−1 (D) is directly related to the amount of defective surface such as a Teflon sheet. The four-point contact
and crystalline modes in the carbon fiber sample. The device is then carefully placed onto the single carbon
ratio between the defective (D) and the graphitic band fiber in order not to break it. A constant amplified direct
(G) also measures quantities, including edge dislocations, current (DC) voltage is passed through the carbon fiber.
vacancies, and crystal edges. More simply stated, it is The current across the two-point sector is recorded using
proportional to the number of atoms of a graphitic plane a digital multimeter, and the electrical resistance (R) of
having one nongraphitic bond [112]. carbon fibers is calculated using Ohm’s law. The resistivity
(ρ) of a single filament is calculated using the fiber length
Crystal Structure of Carbon Fibers and the previously measured fiber diameter using the
following equation:
X-ray diffraction (XRD) is a versatile, nondestructive
technique that allows the crystallographic structure of Rπ d2 Uπ d2 1
materials to be determined [113]. A crystal lattice is a ρ= = = , (26)
4l 4Il κ
regular three-dimensional distribution of atoms in space.
These are arranged so that they form a series of parallel where κ is the specific conductivity, U the applied voltage,
planes separated from one another by a distance (d), which and I the measured current.
varies according to the nature of the material. In a crystal,
planes exist in a number of different orientations. X-rays Thermal Conductivity of Carbon Fibers
are diffracted by crystals because the spacing between
atoms is similar to the wavelength of X-rays. Bragg’s law Thermal conductivity is defined as the amount of heat
(Eq. 24) explains the diffraction of X-rays as reflection on flow through a material. It generally depends on struc-
the lattice planes in a crystal [114,115]: tural imperfection, composition, phase, temperature, and
magnetic field of a material [119]. It is expressed as
nλ = 2d sin θ. (25)
λ(T) = αρCp , (27)
When a monochromatic X-ray beam with wavelength
where α is thermal diffusivity, ρ the material density,
λ is projected onto a crystalline material at an angle
and Cp the specific heat capacity [119,120]. There are
θ, diffraction occurs. This diffraction is specific to the
various approaches taken by researchers to determine
distance traveled by the rays reflected from crystal planes
the thermal conductivity of carbon fibers, and many of
by a complete integer n of the light wavelength [116].
them report that the thermal conductivity of carbon fibers
can be measured accurately up to 1000 W/m/K for fibers
Electrical Conductivity of Carbon Fibers
with diameter ranging from 10 to 30 μm. A steady-state
Carbon fibers are electrically conductive because of the short-hot-wire method was developed by Zhang et al. [121].
delocalization of π -bond electrons above and below the By attaching a single carbon fiber onto one end of a pin fin
12 CARBON FIBER: PROPERTIES, TESTING, AND ANALYSIS

and attaching the other end of the pin fin to a short hot
wire, they managed to run a direct constant current, thus
Adhesive
maintaining the heat transfer at a steady state. However,
this method was found to have a high uncertainty, as the
error induced during the test is significantly higher as
compared to theoretical calculations. On the other hand,
Single-filament
Gallego et al. [119] measured thermal conductivity using test specimen
two approaches, the Angstrom apparatus (room tempera-
ture measurement) and steady-state thermal potentiome-
ter (temperature-dependent measurement). Both methods
were used to measure the thermal diffusivity calculated
using the following equation:
Adhesive
L2
α=  , (28)
T1
2 × t × ln T2

where L is the distance between the temperature sen-

sors, t is the time delay (time taken for the thermal Figure 15. A schematic diagram of a single carbon fiber glued
wave to travel from one sensor to another), and T1 onto a testing card.
and T2 are the thermal wave amplitudes measured
by the sensors. Both methods show better accuracy of
thermal conductivity measurement with ribbonlike carbon cross-sectional area (in mm2 ) determined using the pre-
fibers. viously described laser diffraction or modified Wilhelmy
method:
F
ANALYSIS OF MECHANICAL PROPERTIES OF CARBON σf = . (29)
FIBERS Af

The statistical distribution of fiber strengths can be

Tensile Properties of a Single Carbon Fiber analyzed using the two-parameter Weibull cumulative
From a practical point of view, the mechanical perfor- distribution [124]:
mance of an unidirectional fiber-reinforced composite is w
dominated by the properties of the fiber since the carbon −L σσ
Pf (L) = 1 − e 0 , (30)
fibers bear the load in a composite. When a single fiber
in an unimpregnated bundle of fibers breaks, the broken where Pf (L) is the probability of failure of a fiber of length
fiber sheds its stress so the load is equally shared among L at a stress less than or equal to σ , σ0 is the Weibull scale
the remaining fibers in the bundle. This is commonly parameter or characteristic stress, and w is the shape
referred to as equal load sharing (ELS) [122]. However, parameter or Weibull modulus, which describes the vari-
when a single-fiber breaks in a fiber-reinforced composite, ability of the failure strength. Weibull statistics assume
load sharing among the rest of the fibers is restrained by that a single carbon fiber of length L has independent
the matrix. In this case, stress is redirected back into the links of arbitrary unit length. Each link or segment has
broken fiber by plain shear interaction with the matrix at equal probability of failing or withstanding a given stress
the interface. The matrix localizes and redistributes the level. Rearranging the two-parameter Weibull cumulative
load. This overload is then carried by the fibers nearest distribution leads to
to the broken fiber. This phenomenon is known as local  
load sharing (LLS) [123]. In high loading structural appli- 1
ln ln = w ln σ − w ln σ0 + ln L. (31)
cations, a composite may exceed its loading limit, which 1 − Pf (L)
fractures fibers and eventually results in the failure of the
This equation allows determination of the Weibull scale
composite.
and shape parameters by plotting ln[ln(1/1 − Pf (L) )] as a
The tensile strength of a monofilament carbon fiber
function of ln σ . The gradient is w and the intercept is
can be determined using a tensile stress tester. Standards σ0 . Once the Weibull parameters have been established
such as the BS ISO Standard 11 566 have been developed from tests on shorter fibers, the tensile strength as well
for tensile stress testing guidance. To perform the test, a as the probability of fiber failure at a given length can be
single carbon fiber is glued onto a paper frame (Fig. 15). predicted using the following equation:
The single fiber is loaded at a standardized rate (typically
1 mm/min) until the fiber fails, while the loading force  1/ w
L1
versus displacement being logged using a computer. The σ0(2) = σ0(1) , (32)
L2
corresponding force at break is used to determine the ten-
sile strength (σf ) of a single carbon fiber (Eq. 27). F is the where σ0(2) and σ0(1) are strengths for the lengths of L2 and
measured tensile force (in N) at break and Af is the fiber L1 , respectively.
 CARBON FIBER: PROPERTIES, TESTING, AND ANALYSIS 13

QUANTIFICATION OF THE FIBER/MATRIX ADHESION

The most complex but important feature in the develop-

ment and manufacturing of composites is the fiber–matrix
interface. The quality of interaction between the fiber
reinforcement and polymer matrix is known to influence
the mechanical performance of the resulting composite.
Interfacial interaction can be divided into three common
categories: (i) mechanical interlocking, (ii) intermolecular
interactions, and (iii) chemical bonds between adhesive
(matrix) and adherent (fiber). When load is applied to a
composite, it is transferred from the matrix to the rein-
forcement via the interface. Therefore, a strong interface
is vital to guarantee good load transfer.
Figure 16. An image showing a polyvinylidene fluoride (PVDF)
Direct Wetting
droplet on a single carbon fiber at 180◦ C during direct wetting
An understanding of the wetting phenomena of polymers measurement.
on reinforcing fibers allows prediction of the interfacial
behavior [125,126]. Although the wettability of carbon
fibers can be measured gravimetrically using the modified Fmax
0.025
Wilhelmy method (as described previously), the Wilhelmy
method is not really that suited to determine the wetta-
bility of very thin fibers by polymer. This is due to the 0.020
viscosity and temperatures involved. Contact angle mea- Fiber–matrix
debonding
surements are the simplest method to quantify adhesion Force (N) 0.015
since the contact angle is directly linked to the thermo-
dynamic work of adhesion. However, for a drop-on-fiber 0.010
system, the better the wettability of the polymer on the Friction level
fiber, the more difficult is the determination of the droplet 0.005
length using direct observation of microscopic images
[127]. The contact angle determined from droplets on the 0.000
fibers is very sensitive to errors made in determining the
drop dimensions. A solution to maximize the accuracy of 0.00 0.02 0.04 0.06 0.08 0.10 0.12 0.14
the conventional maximum drop length–height method Distance (mm)
was proposed by Song et al. [127]. The authors developed
Figure 17. Force–displacement curve of a single-fiber pull-out
the generalized drop length–height method based on a
test. Source: Obtained from Ref. 131 with permission from Else-
drop-on-fiber system to measure the contact angle of liq- vier.
uids on single carbon fibers (Fig. 16). The advantage of
the generalized drop length–height method (based on the
principle of the Laplace–Young equation, neglecting the
pull-out test is performed using a piezomotor fixed on a
effect of gravity) is that contact angle is calculated from
very stiff frame in order to avoid energy storage in the free
the extracted drop profile of the entire droplet without
fiber length between the matrix surface and the clamping
the requirement to determine the maximum drop length.
device. The force required to debond and remove the fiber
As a result, the accuracy of the method to determine the
is measured with a load cell. A typical load–displacement
contact angle of liquid on fibers is increased, reducing the
curve is shown in Fig. 17 and the fiber and matrix after
statistical error and improving reliability and accuracy of
a pull-out test is shown in Fig. 18 [131]. The maximum
the measurement.
pull-out force (Fmax ) is correlated to the debonding event
along the interface between the embedded fiber length and
Single-Fiber Pullout
the matrix. The shape of the curve itself reflects the failure
The quality of adhesion at the fiber/matrix interface is event occurring at the interface [132]. The apparent inter-
often assessed by performing a mechanical test such as facial shear strength (τIFSS ) can be calculated from the
single-fiber pullout. This method has received much atten- Fmax required to trigger the debonding using the following
tion over the past few years [128–130], as it provides a equation:
direct measure of interfacial adhesion and can be used with
both brittle and ductile matrices. It is used to determine Fmax
τIFSS = , (33)
the apparent interfacial shear strength (τIFSS ), which is a π df L
measure of the practical adhesion between a carbon fiber
and polymer matrices. In the single-fiber pull-out test, where df is the diameter (or π df = Pf , the fiber perime-
a fiber is partially embedded to a predetermined length ter) of the fiber (determined using the modified Wilhelmy
between 50 and 150 μm into a polymer droplet. The fiber method) and L is the embedded fiber length.
14 CARBON FIBER: PROPERTIES, TESTING, AND ANALYSIS

Figure 18. Scanning electron micro-

graphs showing the wetting cone of the
fibers after pull-out test.

Chip camera calculated using the following equation:

Fmax
Microscope τIFSS = , (34)
Step driver Force transducer π df t
where df is the diameter (or π df = Pf , the fiber perime-
ter) of the fiber (determined using the modified Wilhelmy
Indenter needle
method) and t is the thickness of the specimen.
Joint Sample
Stage Single-Fiber Fragmentation
The single-fiber fragmentation test, developed from the
early work of Kelly and Tyson [134] is commonly used to
Piezo actuator evaluate the interfacial properties of fiber/polymer com-
Figure 19. A schematic diagram of a push-out test apparatus.
posites. Generally, the test is performed under a light
Source: Obtained from Ref. 133 with permission from Elsevier. microscope so that the fragmentation process can be
observed in situ. A single fiber is axially aligned in a
film of matrix material and loaded in tension. The tensile
stress is transferred to the fiber through the interfacial
Push-Out Test shear stress. As the shear stress reaches the fiber fracture
stress, the fiber breaks. Since the fractured fiber segments
The push-out test provides a means of characterizing the
continue to carry load further, increases in strain lead to
interfaces in realistic multifiber composites [133]. The test
additional fiber breaks. The process continues until sat-
specimen is prepared by laying the fibers into a mould
uration is reached. At saturation, the fiber segments are
and the matrix (typically epoxy) is introduced to embed
below the critical length at which stress transfer through
the fibers. For better handling, the composite is further
the fiber–matrix interface can reach the fiber fracture
embedded into a plastic cylinder using a cold mounting
stress. The average fragment length is then measured.
epoxy resin and polished until the desirable slice thickness
Given a constant fiber diameter and strength, shorter
is reached. A typical setup of the push-out test is shown
fragment lengths imply a stronger interface, that is, bet-
in Fig. 19. An indenter is fixed to the force sensor and
ter adhesion [135]. The apparent interfacial shear strength
mounted together with a reflecting light microscope on a
estimated from the Kelly–Tyson model [134] is given by
travel head. Once the fiber of interest is identified using
the following equations:
the optical microscope, the two-position specimen stage is
switched from the position under the microscope to the σf df
τIFSS = , (35)
one under the indenter. The fibers are pushed out from 2lc
the matrix through the movement of the indenter in the 4
fiber direction. The apparent interfacial shear strength is lc = l, (36)
3
 CARBON FIBER: PROPERTIES, TESTING, AND ANALYSIS 15

F REFERENCES
1. ACS. National historic chemical landmarks—Bacon’s break-
through. Washington (DC); 2007.
2. Edison TA. Electric lamp, U.P. Office. US patent 0223898;
1880.
3. Israel P. Edison: a life of invention. New York: Wiley; 1998.
4. Otani S. Carbon 1965;3(1):31–34. DOI: 10.1016/008-
6223(65)90024-2.
5. Lewis R. Carbon fibre: transforming F1. 2003.
6. Dorey G. Phys Technol 1980;11(2):56–61. DOI: 10.1088/
F F 0305-4624/11/2/I01.
2 L L 2 7. Bennett SC, Johnson DJ, Johnson W. J Mater Sci
2 2 1983;18(11):3337–3347. DOI: 10.1007/BF00544159.
Figure 20. Typical specimen configuration for short-beam shear 8. Gupta AK, Paliwal DK, Bajaj P. J Macromol Sci Polymer
test. Rev 1991;C31(1):1–89. DOI: 10.1080/15321799108021557.
9. Donnet JB, Wang TK, Rebouillat S, et al. Carbon fibers.
New York: Marcel Dekker; 1998.
where df is the fiber diameter, σf the fiber strength at the 10. Åström BTÅ. Manufacturing of polymer composites. London:
critical fragment length, which can be predicted from the Chapman & Hall; 1997.
Weibull distribution, and lc the critical fragment length 11. Johnson DJ. J Phys Appl Phys. 1987;20(3):286–291.
calculated from the mean fiber fragment length (l) at 12. Jones FR. Handbook of polymer-fibre composites. Longman:
saturation [136]. Harlow; 1994.
13. Donnet JB, Bahl OP. In: Meyers RA, editor. Volume 2,
Encyclopaedia of physical science and technology. New York:
Short-Beam Shear
Academic Press: 1987.
The short-beam strength method, based on classical beam 14. Bajaj P, Chavan RB, Manjeet B. J Macromol Sci Chem
theory, is the simplest and therefore most commonly 1986;A23(3):335–347. DOI: 10.1080/00222338608063397.
used test to characterize ‘‘apparent’’ interlaminar shear 15. Padhye MR, Karandikar AV. J Appl Polym Sci
behavior. During short-beam strength testing, interlam- 1987;33(5):1675–1682. DOI: 10.1002/app.1987.070330521.
inar shear is generated indirectly through a three-point 16. McIntyre JE, Kajiwara K, Nakajima T. Advanced fiber spin-
bending mode in order to characterize the apparent inter- ning technology. Cambridge: Woodhead Publishing; 1994.
laminar shear properties of unidirectional fiber-reinforced 17. Moreton R, Watt W. Carbon 1974;12(5):543–554. DOI:
composites. Short-beam shear testing is very similar to 10.1016/0008-6223(74)90056-6.
flexural testing except the span-to-thickness ratio is much 18. Grassie N, McGuchan R. Eur Polym J 1971;7(8):1091–1104.
lower. This test allows maximum shear stresses to be intro- DOI: 10.1016/0014-3057(71)90141-8.
duced throughout the thickness of the test specimen, while 19. Berlin AA, Dubinskaya AM, Moshkovskii YS. Vysokomol
reducing the tensile and compressive flexural stresses to Soedin 1964;6:1938.
a minimum. 20. Fester W. Textil Rundschau 1965;20:1.
A typical specimen configuration is shown in Fig. 20. 21. Burlant WJ, Parsons JL. J Appl Polym Sci
A composite laminate of 20 mm × 10 mm × 2 mm is sup- 1956;22:(101):249–256. DOI: 10.1002/pol.1956.1202210107.
ported by two cylindrical rollers. The specimen is loaded
22. Grassie N, Hay JN. J Appl Polym Sci 1962;56(163):189–202.
on the central roller located at mid-length of the specimen DOI: 10.1002/pol.1962.1205616316.
at displacement rate of 1 mm/min until failure according to
23. Friedlander HN, Peebles LH, Brandrup J, et al. Macro-
ASTM D2344 using standard mechanical test equipment. molecules 1968;1(1):79–86. DOI: 10.1021/ma60001a014.
The short-beam strength (F SBS ) is calculated using the
24. Johnson JW, Philips LN, Watt W. British patent I. 110.791;
following equation: 1968.
25. Layden GK. J Appl Polym Sci 1971;15(7):1709–1715. DOI:
Fmax
F SBS = 0.75 , (37) 10.1002/app.1971.070150714.
b×d
26. Rosenbaum S. J Appl Polym Sci 1965;9(6):2071–2084. DOI:
10.1002/app.1965.070090606.
where Fmax is the force at composite failure load, b the
27. Warner SB, Peebles LH, Uhlmann DR. J Mater Sci
specimen width, and d the specimen thickness.
1979;14(3):556–564. DOI: 10.1007/BF00772714.
28. Fitzer E, Heym M. Chem Ind 1976;16:663–676.
Acknowledgments
29. Fitzer E, Muller DJ. Makromol Chem 1971;144(1):117–133.
The support from the UK Engineering and Physical Sci- DOI: 10.1002/macp.1971.021440110.
ence Research Council (EPSRC) and Imperial College 30. Bahl OP, Manocha LM. Angew Makromol Chem
London’s Deputy Rector award for funding K.-Y.L. and 1975;48(1):145–159. DOI: 10.1002/apmc.1975.050480109.
Malaysian Ministry of Higher Education for S.-R.S. is 31. Fitzer E, Frohs W, Heine M. Carbon 1986;24(4):387–395.
greatly appreciated. DOI: 10.1016/0008-6223(86)90257-5.
16 CARBON FIBER: PROPERTIES, TESTING, AND ANALYSIS

32. Bracker P, Schurmans H, Verhoest J. Inorganic fibers and 60. Fowkes FM. J Adhes Sci Technol 1990;4(1):669–691. DOI:
composite materials. Volume 3, EPO applied technology 10.1163/156856190x00595.
series. New York: Pergamon Press; 1984. 61. Kwok DY, Li D, Neumann AW. Langmuir 1994;10(4):
33. Bahl OP, Mathur RB, Kundra KD. Fibre Sci Technol 1323–1328. DOI: 10.1021/la00016a057.
1980;13(2):155–162. DOI: 10.1016/0015-0568(80)90043-3. 62. van Oss CJ, Good RJ, Busscher RJ. J Disper Sci Technol
34. Raskovic V, Marinkovic S. Carbon 1978;16(5):351–357. DOI: 1990;11(1):75–81. DOI: 10.1080/01932699008943237.
10.1016/0008-6223(78)90073-8. 63. van Oss CJ, Chaudhury MK, Good RJ. Adv Colloid Interface
35. Fitzer E, Muller DJ. Carbon 1975;13(1):63–69. DOI: Sci 1987;28:35–64. DOI: 10.1016/0001-8686(87)80008-8.
10.1016/0008-6223(75)90259-6. 64. van Oss CJ, Giese RF, Good RJ. Langmuir 1990;6:(11):
36. Bahl OP, Manocha LM. Carbon 1974;12(4):417–423. DOI: 1711–1713. DOI: 10.1021/la00101a017.
10.1016?0008-6223(74)90007-4. 65. van Oss CJ, Good RJ, Chaudhury MK. J Colloid Interface Sci
37. Watt W, Philips LN, Johnson W. Engineer 1966;221:815. 1986;111(2):378–390. DOI: 10.1016/0021-9797(86)90041-x.
38. Johnson JW. In: Preston J, editor. Applied polymer sym- 66. van Oss CJ, Ju L, Chaudhury MK, et al. J Col-
posia, no 9 New York: Interscience Wiley; 1969. p. 229. loid Interface Sci 1989;128(2):313–319. DOI: 10.1016/
39. Goodhew PJ, Clarke AJ, Bailey JE. Mater Sci Eng 0021-9797(89)90345-7.
1975;17:3–30. 67. Ho KKC, Lee AF, Bismarck A. Carbon 2007;45(4):775–784.
40. Watt W. Nat Phys Sci 1972;236(62):10–11. DOI: 10.1038/ DOI: 10.1016/[Link].2006.11.015.
physci236010a0. 68. Wu S. Polymer interface and adhesion. New York: Marcel
41. Shindo A. In: Carbon fibers; their composites and applica- Dekker; 1982.
tions. London: The Plastic Institute London; 1971. p. 18. 69. Cantero G, Arbelaiz A, Llano-Ponte R, et al. Com-
42. Moreton R, Watt W, Johnson W. Nature 1967; pos Sci Technol 2003;63(9):1247–1254. DOI: 10.1016/
213(5077):690–691. DOI: 10.1038/213690a0. s0266-3538(03)00094-0.
43. Cooper GA, Mayer RM. J Mater Sci 1971;6(1):60–67. DOI: 70. Schultz J, Lavielle L. In: Lloyd DR, Ward TC, Schreiber
10.1007/BF00550292. HP, editors. Inverse gas chromatography: characterization
44. Kaelble DH, Dynes PJ, Crane LW, et al. J Adhes of polymers and other materials. ACS symposium series.
1975;7(1):25–54. DOI: 10.1080/00218467508078896. Washington (DC): American Chemical Society; 1989. pp.
185–202.
45. Larsen JV, Smith TG, Erickson PW. Carbon fiber surface
treatments. Maryland: NOLTR; 1971. p. 71. NOLTR-71-165. 71. Park SJ, Donnet JB. J Colloid Interface Sci
1998;206(1):29–32. DOI: 10.1006/jcis.1998.5072.
46. Daukeys RJ. J Adhes 1973;5(3):211–244. DOI: 10.1080/
00218467308075022. 72. Dorris GM, Gray DG. J Colloid Interface Sci
1980;77(2):353–362. DOI: 10.1016/0021.9797(80)90304-5.
47. Drzal LT, Rich MJ, Koenig MF, et al. J Adhes
1983;16(2):133–152. DOI: 10.1080/00218468308074911. 73. Shen W, Parker IH. Cellulose 1999;6(1):41–55. DOI:
10.1023/A:1009268102404.
48. Subramanian RV. In: Cantow HJ, editor. Electric phenom-
ena in polymer science. Volume 3, Advances in polymer 74. Riedl B, Kamdem PD. J Adhes Sci Technol 1992;
science. Berlin: Springer; 1979. 6(9):1053–1067. DOI: 10.1163/156856192X00935.
49. Thornton PR. Scanning electron microscopy: applications to 75. Gutmann V. Coordinat Chem Rev 1967;2(2):239–256. DOI:
materials and device science. London: Chapman and Hall; 10.1016/S0010-8545(00)80206-4.
1968. 76. Saintflour C, Papirer E. Ind Eng Chem Prod Res Dev
50. Flegler SL, Heckman JW, Klomparens KL. Scanning 1982;21(2):337–341. DOI: 10.1021/i300006a029.
and transmission electron microscopy: an introduction. 77. Riddle FL, Fowkes FM. J Am Chem Soc
New York: Oxford University Press; 1995. 1990;112(9):3259–3264. DOI: 10.1021/ja00165a001.
51. Wells OC. Scanning electron microscopy. New York: 78. Hunter RJ. Zeta potential in colloid science. New York:
McGraw-Hill; 1974. Academic Press; 1981.
52. Meretz S, Linke T, Schulz E, et al. J Mater Sci Lett 79. Werner C, Korber H, Zimmermann R, et al. J Colloid Inter-
1992;11(21):1471–1472. DOI: 10.1007/BF00729668. face Sci 1998;208(1):329–346. DOI: 10.1006/jcis.1998.5787.
53. Li CT, Tietz JV. J Mater Sci 1990;25(11):4694–4698. DOI: 80. Jacobasch HJ, Simon F, Werner C, et al. Tech Mess
10.1007/BF01129926. 1996;63:439.
54. Heng JYY, Pearse DF, Thielmann F, et al. Compos 81. Jacobasch HJ, Simon F, Werner C, et al. Tech Mess
Interf 2007;14(7–9):581–604. DOI: 10.1163/156855407 1996;63:447.
782106492. 82. Wagner CD, Davis LE, Zeller MV, et al. Surf Interface Anal
55. Ho KKC, Beamson G, Shia G, et al. J Fluor 1981;3(5):211–225. DOI: 10.1002/sia.740030506.
Chem 2007;128(11):1359–1368. DOI: 10.1016/[Link]. 83. Hubbard AT. Encyclopedia of surface and colloid science.
2007.06.005. New York: CRC Press; 2002.
56. Hunter RJ. Introduction to modern colloid science. 84. Aranberri-Askargorta I, Lampke T, Bismarck A. J Col-
New York: Oxford University Press; 1993. loid Interface Sci 2003;263(2):580–589. DOI: 10.1016/
57. Fowkes FM. J Phys Chem 1962;66(2):382. DOI: 10.1021/ S0021-9797(03)00294-7.
j100808a524. 85. Jacobasch HJ, Grundke K, Uhlmann P, et al. Compos Inter-
58. Fowkes FM. Ind Eng Chem 1964;56(12):40–52. DOI: faces 1995;3(4):293–320. DOI: 10.1163/156855495X00093.
10.1021/ie50660a008. 86. Campagne C, Devaux E, Perwuelz A, et al. Poly-
59. Fowkes FM, Mostafa MA. Ind Eng Chem Prod Res Dev mer 2002;43(25):6669–6676. DOI: 10.1016/S0032-
1978;17(1):3–7. DOI: 10.1021/i360065a002. 3861(02)00596-7.
 CARBON FIBER: PROPERTIES, TESTING, AND ANALYSIS 17

87. Bismarck A, Richter D, Wuertz C, et al. Colloid Surf 112. Ko TH. J Appl Polym Sci 1996;59(4):577–580. DOI:
A-Physicochem Eng Asp 1999;159(2–3):341–350. DOI: 10.1002/(SICI)1097-4628(19960124)59:4<577::AID-APP2>
10.1016/S0927-7757(99)00262-9. 3.0CO;2-Q.
88. Chiu HT, Wang JH. Polym Compos 1998;19(4):347–351. 113. Alexander LE. X ray diffraction methods in polymer science.
DOI: 10.1002/pc.10107. New York: Krieger Pub Co; 1979.
89. Ismail IMK. Carbon 1987;25(5):653–662. DOI: 10.1016/ 114. Fox TG, Garrett BS, Goode WE, et al. J Am Chem Soc
0008-6223(87)90219-3. 1958;80(7):1768–1769. DOI: 10.1021/ja01540a068.
90. Setoyama N, Li G, Kaneko K, et al. Adsorpt-J Int Adsorpt 115. Shriver DF, Atkins PW. Inorganic chemistry. 3rd edn.
Soc 1996;2(4):293–297. DOI: 10.1007/BF00879544. New York: Oxford University Press; 1999.
91. Lee YS. J Fluor Chem 2007;128(4):392. DOI: 10.1016/ 116. Tan KH. Principles of soil chemistry. New York: Marcel
[Link].2006.11.014. Dekker; 1998.
92. Gregg SJ, Sing KSW. Adsorption, surface area, and porosity. 117. Din GS. Methods of test for materials for electri-
London: Academic Press; 1991. cal purposes—measuring of the electrical resistance of
93. Hammer GE, Drzal LT. Appl Surf Sci 1980;4:340. non-metallic materials (DIN 53482). Germany: German
Institute of Standards; 1983.
94. Tressaud A, Moguet F, Flandrois S, et al. J Phys
Chem Solids 1996;57(6–8):745–751. DOI: 10.1016/ 118. Methods of test for materials for electrical
0022-3697(96)00343-5. purposes—measuring of the electrical resistance of
non-metallic materials (DIN 53482). Germany: German
95. Briggs D, Seah MP. Practical surface analysis: auger and
Institute of Standards; 1983.
X-ray photoelectron spectroscopy. Chichester: John Wiley &
Sons; 1990. 119. Gallego NC, Edie DD, Nysten B, et al. Carbon 2000;38(7):
1003–1010. DOI: 10.1016/S0008-6223(99)00203-1.
96. Rivière JC. Surface analytical techniques. New York:
Clarendon Press, Oxford University Press; 1990. 120. Pradere C, Batsale JC, Goyheneche JM, et al. Carbon
2009;47(3):737–743. DOI: 10.1016/[Link].2008.11.015.
97. Walls JM. Methods of surface analysis. Cambridge, New
York: Cambridge University Press; 1989. 121. Zhang X, Fujiwara S, Fujii M. Int J Thermophys 2000;
21(4):965–980. DOI: 10.1023/A:1006674510648.
98. Barr TL. Modern ESCA: the principles and practice of x-ray
photoelectron spectroscopy. Boca Raton (FL): CRC Press; 122. Phoenix SL, Taylor HM. Adv Appl Probab 1973;5:200–216.
1994. 123. Harlow DG, Phoenix SL. J Compos Mater 1978;
99. Attard G, Barnes C. Surfaces. New York: Oxford University 12(2):195–214. DOI: 10.1177/002199837801200207.
Press; 1998. 124. Weibull W. J Appl Mech 1951;18:293.
100. McCash EM. Surface chemistry. New York: Oxford Univer- 125. Connor M, Harding PH, Manson JAE, et al.
sity Press; 2001. J Adhes Sci Technol 1995;9(7):983–1004. DOI: 10.1163/
101. Powell CJ. Surf Sci 1994;299–300:34–48. DOI: 156856195X00833.
10.1016/0039-6028(94)90644-0. 126. Carroll BJ. Langmuir 1986;2(2):248–250. DOI: 10.1021/
102. Briggs D, Grant JT. Surface analysis by Auger and x-ray la00068a024.
photoelectron spectroscopy. Chichester, West Sussex: IM 127. Song BH, Bismarck A, Tahhan R, et al. J Colloid Interface
Publications; 2003. Sci 1998;197(1):68–77. DOI: 10.1006/jcis.1997.5218.
103. Beamson G, Briggs D. High resolution XPS of organic poly- 128. Li JX. J Appl Polym Sci 1994;53(2):225–237. DOI:
mers : the Scienta ESCA300 database. Chichester, New York 10.1002/app.1994.070530211.
John Wiley & Sons; 1992. 129. Chua PS, Piggott MR. Compos Sci Technol 1985;
104. Pijpers AP, Berresheim K, Wilmers M. Fresenius J Anal 22(2):107–119. DOI: 10.1016/0266-3538(85)90079-X.
Chem 1993;346:(1–3):104–109. DOI: 10.1007/BF00321392. 130. Chou CT, Penn LS. J Adhes 1991;36(2–3):125–137. DOI:
105. Egelhoff WF. Surf Sci Reports 1987;6:(6–8):253–415. DOI: 10.1080/00218469108027067.
10.1016/0167-5729(87)90007-0. 131. Ho KKC, Kalinka G, Tran MQ, et al. Compos Sci
106. Fadley CS. In: Brundle C, Baker A, editors. Electron spec- Technol 2007;67(13):2699–2706. DOI: 10.1016/jcomp-
troscopy: theory, techniques and applications. New York: scitech.2007.02.012.
Academic Press; 1977. 132. DiFrancia C, Ward TC, Claus RO. Compos A-Appl Sci Manuf
107. Turner NH. Investigations of surface and interfaces, Part-B. 1996;27(8):597–612. DOI: 10.1016/1359-835X(95)00069-E.
New York: John Wiley & Sons; 1993. 133. Kalinka G, Leistner A, Hampe A. Compos Sci Technol
108. Watts JF, Wolstenholme J. An introduction to surface anal- 1997;57(8):845–851. DOI: 10.1016/S0266-3538(96)00159-5.
ysis by XPS and AES. Chichester, West Sussex, New York: 134. Kelly A, Tyson WR. J Mech Phys Solids 1965;13(6):329–350.
John Wiley & Sons; 2003. DOI: 10.1016/0022-5096(65)90035-9.
109. Atkins PW. Physical chemistry: Science of Biology. 6th edn. 135. Feih S, Wonsyld K, Minzari D, et al. Testing procedure
New York: Freeman & Company; 1997. for the single fiber fragmentation test [Risø-R-1483(EN)].
110. Nemanich RJ, Lucovsky G, Solin SA. Mater Roskilde, Denmark: Risø National Laboratory; 2004.
Sci Eng 1977;31:157–160. DOI: 10.1016/0025-5416(77) 136. Ohsawa T, Nakayama A, Miwa M, et al. J Appl
90029-5. Polym Sci 1978;22(11):3203–3212. DOI: 10.1002/app.
111. Tuinstra F, Koenig JL. J Chem Phys 1970;53(3):1126–1130. 1978.070221115.
DOI: 10.1063/1.1674108.