Composites: Part A 39 (2008) 1141–1149

Contents lists available at ScienceDirect

Composites: Part A
journal homepage: www.elsevier.com/locate/compositesa

Effective properties of carbon black filled natural rubber: Experiments and modeling
B. Omnès a, S. Thuillier a,*, P. Pilvin a, Y. Grohens a, S. Gillet b
a
LIMATB, Université de Bretagne-Sud, BP 92116, 56321 Lorient, France
b
Société des polymères Barre Thomas SA, 194 route de Lorient, CS 74321, 35043 Rennes Cedex, France

a r t i c l e i n f o a b s t r a c t

Article history: Anti-vibrating parts in automotives are often made of natural rubber reinforced by carbon black. This
Received 9 October 2007 reinforcement, which comes from the filler–filler and filler–rubber interactions, leads to an increase of
Received in revised form 10 April 2008 the elastic modulus, the tensile strength and the hysteresis. The aim of this work is to develop a model
Accepted 10 April 2008
within a generalized self-consistent scheme for filled rubber to predict the elastic moduli. A morpholog-
ical pattern, which takes into account the occluded rubber, the bound rubber and a percolating network,
is proposed and the effective elastic properties are compared with experimental results obtained in both
Keywords:
uniaxial and oedometric compression. The filler volume fraction varies in-between 0% and 18% and the
Natural rubber
Carbon black
influence of the specific surface of the filler is investigated, by using N330 and N650 carbon blacks.
Reinforcement effect Ó 2008 Elsevier Ltd. All rights reserved.
Elastic moduli

1. Introduction when the distance between aggregates is low [6,12–16]. Carbon

black types depend mainly on their specific surface, which is the
For many industrial applications, like anti-vibrating structures, ratio of the total outer surface of the filler over its mass, their struc-
natural rubber is reinforced by carbon black. This leads to an in- ture and their surface activity [1,17] and the reinforcement effect is
crease of the mechanical properties, i.e. elastic modulus and tensile greatly affected by these parameters, e.g. [3,12].
strength, as well as a sharp increase of the hysteresis, e.g. [1]. This The reinforcement effect is classically split into different contri-
reinforcement effect comes from the filler–filler and filler–rubber butions [1–3,7]: (i) the hydrodynamical effect, which corresponds
interactions, both chemical and physical, which take place at dif- to the reinforcement obtained by inclusions dispersed within a
ferent length scales due to the specific structure of carbon black. matrix, (ii) the strong interactions between the polymer and the
Carbon black particles are strongly bonded to other elementary filler, leading to the modified rubber layer and (iii) the agglomerate
particles (10 up to 100) to form an aggregate, the size of which is structure contribution, above a percolating threshold. Different
of the order of 100 nm [2]. Due to the branched structure of the models have been proposed to predict, mainly from the filler vol-
aggregate, part of the rubber is trapped inside the aggregate and ume fraction, the elastic properties of the composite. Recently,
is shielded from macroscopic deformation [3]; the occluded rubber micromechanical model developed within a self-consistent ap-
forms with carbon black particles the core of the aggregates and its proach have shown a good correlation with experimental results,
main contribution is to lead to an effective filler volume fraction. It e.g. [18]. These models rely on the definition of a morphological
is commonly accepted that the occluded rubber proportion can be pattern which is representative of the microstructure of the mate-
measured from dibutyl-phtalate (DBP) absorption [4]. Due to the rial. Mélé et al. [19] predicted, from rheological experiments on a
filler–rubber interactions, polymer chains are strongly linked on styrene–butadiene matrix elastomer reinforced by silica, the rheo-
carbon black surface and a layer of rubber with modified proper- logical properties of the percolating network and the bound rub-
ties, also called bound rubber, surrounds the filler particles [5–8]. ber. Gauthier et al. [20] proposed a 3-phase model applied in
A new approach has been recently proposed [9–11], that takes into three successive steps, for a silica reinforced SBR matrix, in order
account a glass transition gradient around the inorganic particles. to highlight occluded rubber and bound rubber interactions, then
This gradient yields a spatial variation of the elastic modulus in filler interactions with the previous media, leading to an aggregate
the vicinity of the particles which can strongly increase the modu- and finally bulk rubber with the aggregate. Such approaches have
lus of the matrix. The aggregates tend to agglomerate, and at high also been extended to the non-linear field, in the case of carbon
concentrations, can form a percolating three-dimensional network black reinforced natural rubber [21].
The aim of this article is to present the prediction of the elastic
properties (both Young’s and bulk moduli) of a natural rubber
* Corresponding author.
E-mail address: [email protected] (S. Thuillier). reinforced by carbon black with a self-consistent model. Section

1359-835X/$ - see front matter Ó 2008 Elsevier Ltd. All rights reserved.
doi:10.1016/j.compositesa.2008.04.003
1142 B. Omnès et al. / Composites: Part A 39 (2008) 1141–1149

2 describes the materials and the experiments performed to char-

acterize the microstructure and the mechanical behavior: swelling
tests to investigate the rubber state within the composite for dif-
ferent filler volume fraction, rheology of the green compounds to
determine the percolation threshold, investigation of the Payne ef-
fect with dynamic mechanical analysis and finally tension and
oedometric compression to measure the elastic properties. A mor-
phological pattern representative of the internal structure of the
material [22] is proposed in Section 3, which takes into account oc-
cluded rubber, carbon black, bound rubber and a percolating net-
work. The effective properties thus obtained are compared with
experimental results. Materials with three volume fractions of car-
bon black have been characterized and used to fully fix the mate-
rial parameters; therefore further work should be performed, with
other volume fractions, to validate the approach. The influence of
the specific surface of the filler is studied, by using N650 and
N330 carbon black types and is taken into account in the model.

2. Experiments

2.1. Materials

The material used in this study is elaborated on purpose, in or-

der to have two main components i.e. natural rubber and carbon
black, and a small number of additives like sulfur, accelerator
and lubricants. The composition for 100 g of natural rubber is the
following one: stearic acid 2.0 g, ZnO 5.0 g, sulphur 1.8 g, accelera-
tor CBS 2.5 g and carbon black is variable from 0 to 45 g.
The chemical properties of natural rubber were determined:
1 1
M n ¼ 3:2  105 g mol , M w ¼ 2:5  105 g mol , Ip = 2.1; its den-
3
sity is equal to 0.93 g cm . The volume fraction of the filler / is ta-
ken equal to 0%, to study the mechanical behavior of the pure
rubber, 2.4%, 6.8%, 12.8% and 18%. The corresponding part per hun-
dred rubber or phr, which corresponds to the mass of carbon per
100 g of rubber, varies in-between 0 to 45. Two types of carbon
black are used in this study, a N330 and N650, the characteristics
of which are given in Table 1. All samples are mixed and vulcan-
ised. The vulcanisation time is adapted for each formulation and
it corresponds to the time for which both a maximum stress level
and a minimum unreversible strain are obtained. This optimized
time decreases with increasing /, with a ratio of one third between
pure rubber and samples with a phr of 45. This time is roughly the
same for both types of carbon black.
Strips of materials of thickness 2 mm were cryo-fractured and
the surfaces were observed with a scanning electron microscope,
after spraying a thin metallic layer, under an accelerating voltage
of 20 kV. The aggregate distribution for the volume fractions of
12.8% and 18% and carbon black N330 is shown in Fig. 1a and c. It
can be seen that the filler is rather homogeneously dispersed in
the matrix. The influence of the type of carbon black is shown from
micrographs (a) and (b): for the same volume fraction of 18%, the
aggregates have a larger size for N650 than for N330, according to
the specific surface Sp given in Table 1. Image analysis was carried Fig. 1. SEM micrographs of the microstructure of the materials of the study: (a)
out to determine the mean size of the aggregates. It was found that 30 phr with N330; (b) 30 phr with N650; (c) 45 phr with N330.
this size is rather independent of the volume fraction of the filler
and when assuming a spherical shape, a diameter of 0.06 lm for
N330 is calculated versus a diameter of 0.1 lm for N650. Strain-induced crystallization in tension is not considered in
this study; indeed, the tensile tests where limited to an elongation
of 1.5, and within this range, and for a maximum volume fraction
Table 1
Characteristics of N330 ans N650 carbon blacks
of 0.18, the strain-induced crystallisation of natural rubber can be
reasonably neglected [23].
Carbon black Mean particle size Specific surface DBP (ml/
type (nm) (m2 g1) 100 g)
2.2. Swelling properties
N330 29 84 101
N650 50 27 122
The swelling properties of the vulcanized material have been
The density is taken equal to 1.8 g cm3. investigated by setting natural rubber samples in toluene. Firstly,
 B. Omnès et al. / Composites: Part A 39 (2008) 1141–1149 1143

the samples are weighted in the dry condition (initial mass mi) N330

Storage modulus G’(MPa)

then immersed in the solvent during 6 days at room temperature. N650
1
The aim was to characterize the state of the rubber in the compos- 18%
ite. The mass in the swollen state is measured (ms) after a slight re-
moval of the solvent on the free surfaces of the sample.
13%
Subsequently, the sample is dried in a vacuum oven at a tempera- 18%
ture of 60 °C during 12 hours to evaporate the solvent. The dried
mass (md) is then measured. In order to test the reproducibility
of the measures, three samples were characterized for each filler 0.1
content. The swelling ratio G is calculated from 13%
NR
qrubber ms  md 1 7%
G¼1 ð1Þ
qsolvent md 1  /m 30 60 90 120 150
Temperature (degre C)
where qrubber and qsolvent are the density of the pure rubber and the
solvent, respectively, and /m the weight fraction of the filler. The Fig. 3. Storage modulus versus temperature for filled natural rubber with different
maximum swelling ratios obtained for both types of carbon black volume fractions, for both carbon black N330 and N650. At high temperatures, the
are plotted versus the filler volume fraction in Fig. 2. For both car- existence of a plateau can be related to a percolating network.

bon blacks, this ratio decreases with an increasing filler volume

fraction [24]. A slight (3%) difference can be noticed for the two
highest volume fractions between N330 and N650 carbon black. main in the case of uncured materials. The aim was to determine
The average molar mass Mc in-between reticulation points is calcu- the volume fraction above which carbon black forms a continuous
lated with the Flory–Rehner equation, as presented in [25,26]: percolating network structure, as evidenced in [15]. The viscoelas-
tic measurements were carried out with a Rubber Process Analyzer
2ððV rubber Þ1=3  0:5V rubber Þqrubber V toluene
Mc ¼ ð2Þ (RPA 2000), which is a torsional dynamic rheometer [28]. The evo-
lnð1  V rubber Þ þ V rubber þ vðV rubber Þ2 lution of the storage modulus G0 , real part of G*, as a function of the
with Vtoluene the molar volume of toluene (106.1 cm3 mol1), and temperature was obtained for the uncured filled elastomers; vulca-
v = 0.43 the Huygins natural rubber–toluene interaction parameter nisation was avoided by removing the sulphur in the formulation,
[27], Vrubber the volume fraction of rubber in the swollen material. in order not to have a percolating 3D network from the chemical
The average molar mass Mc in-between reticulation points is given links between the macromolecular chains. The domain of linear
in Table 2. These measures indicate that the reticulation state of the viscoelasticity was determined for the high volume fraction com-
rubber is modified due to an increasing volume fraction of carbon posites. In the linear regime, the viscoelastic parameters are inde-
black. However, and as a first step, this phenomenon will not be ta- pendent of the strain amplitude c. Strain sweep experiments at a
ken into account in the model. fixed temperature of 90 °C and a frequency of 1 Hz were performed
on the samples with / = 18%. The results showed that the limit of
2.3. Rheological properties of the green compound the domain of linear viscoelasticity is equal to c = 0.0055. The opti-
mal strain was then fixed at this value and the frequency of 1Hz is
The evolution of the complex dynamic modulus G* of the filled kept for temperature sweep experiments. The storage modulus G0
natural rubber with temperature was investigated in the linear do- was measured within the temperature range (30–140 °C) for each
volume fraction and type of carbon black. The specimen were first
heated at 90 °C in order to have a good bonding in-between the
5.25 sample and the rheometer grips and then the temperature sweep
N330
N650 is performed. The evolution of the storage modulus versus temper-
5 ature is rather similar for volume fractions below or equal to 7% for
N330 (Fig. 3), it decreases regularly with an increasing tempera-
Swelling ratio

4.75 ture. For higher volume fractions, the curve exhibits a plateau
when the temperature is above 90 °C. This plateau can be related
to the existence of a percolating network. The corresponding
4.5
threshold value above which such a network exists lies in-between
7% and 13%. This value of the mechanical percolation threshold is
4.25 in good agreement with the values from [15], given between 5%
and 9%.
4 Similar trends are observed for N650 (Fig. 3), except that the
0 0.05 0.1 0.15 0.2 plateau is observed only for / = 18%. The threshold value lies there-
Φ fore in-between 12.8% and 18%, it is higher for N650 than for N330,
Fig. 2. Variation of the maximum swelling ratio with volume fraction of carbon
which is consistent with the higher specific surface of the N330
black. compared with N650 and which corresponds to a highly dispersed
state of the carbon black.

Table 2 2.4. Dynamic mechanical analysis

Average molar mass in-between reticulation points Mc

Filler volume Carbon black N330 Carbon black N650 In order to investigate the reinforcement effect for both types of
fraction (g mol1) (g mol1) carbon black, cyclic shear tests with a variable amplitude were per-
0 17740 ± 160 id. formed on double-gauge length samples made of vulcanised mate-
0.068 14550 ± 150 14370 ± 140 rial, at room temperature and a frequency of 1 Hz. The storage
0.128 12980 ± 150 11770 ± 80
moduli were calculated, within a shear strain range from 0.002
0.180 11150 ± 120 10050 ± 150
up to 0.8 and results are plotted in Fig. 4. The origin of the decrease
1144 B. Omnès et al. / Composites: Part A 39 (2008) 1141–1149

3.5 thickness) and cylindrical compression samples have a diameter

NR
N330 of 20 mm and a height of 25 mm. They were performed three times
Storage modulus G’(MPa)

3 18% N650 for each filler content and type of carbon black, to test the repro-
ducibility. The samples are stabilised under four loading–unload-
2.5
ing cycles up to the maximum displacement of the test. Force
2 18% and crosshead displacement are recorded and the logarithmic
strain (e) and Cauchy stress (r) are calculated by assuming an iso-
1.5 13% choric transformation. The strain rate is of the order of
2  102 s1. Strain–stress curves for a N330 type carbon black
1
(Fig. 6) show an increase of the elastic modulus, the hysteresis
7%
0.5 and the residual strain with the increase of / both in tension and
NR
compression. The same evolution is observed with a N650 type
0 carbon black. Tensile and compression results were used to calcu-
0.01 0.1 1
γ late the Young’s modulus E on the stabilised loading. The stress–
strain curves are fitted by a sixth-order polynomial function for a
Fig. 4. Evolution of the storage modulus with the shear strain for N330 and N650 strain jej < 0.09 both in tension and compression and Young’s mod-
carbon blacks. The magnitude of the Payne effect is higher for N330 carbon black
than for N650.
uli are given in Table 3. It was noted that the Young’s modulus
measured in tension during the first loading is higher than the
one measured during the fifth loading, with a gap varying around
3–4% at low volume fraction and up to 12% for phr 45. This differ-
2.5
N330 ence is less significant in compression. Moreover, a slight differ-
N650 ence between the modulus measured in tension and compression
2 is also noticed; this can come from the different sample geometries
and boundary conditions of tensile and compression tests. More-
ΔG’ (MPa)

1.5 over, there is a local strain measure with an extensometer for ten-
sile test whereas the global crosshead displacement is used in
compression.
1
Holownia [32] studied the variation of bulk modulus K and
Young’s modulus E with the volume fraction of carbon black. Pois-
0.5 son’s ratio is calculated from the two measured elastic constants K
and E. For this study, an oedometric device was developed from
0 Holownia’s work, to investigate the dependence of the bulk modu-
0 0.05 0.1 0.15 0.2 lus on the filler volume fraction, for each type of carbon black. The
Φ specimen are moulded cylinders of diameter d = 30.0 ± 0.2 mm and
height h = 13.0 ± 0.1 mm. The size of the samples depends on the
Fig. 5. Evolution of the maximum amplitude of the storage modulus with the filler
volume fraction for both types of carbon black.
formulation of the composite; indeed, the volume increases with
the volume fraction of carbon black.

of the storage modulus with the shear strain or Payne effect [29] is
still a matter of discussion, and can be explained as the breakdown 18%
either of the percolating network [14] or of unstable links between 1.6 13%
the polymer and the filler particles [30,31]. Due to its low specific
Cauchy Stress (MPa)

surface, given in Table 1, the amplitude of the Payne effect is lower 7%

by approximately 40% for N650 than for N330. The saturation val- 0.8 NR
ues at low and high strain amplitude were not reached during the
experiments and therefore a phenomenological expression given in 0
[30] of the decrease of G0 with c is assumed and G00 and G01 were
derived by fitting each curves for each volume fraction with -0.8
0 G0  G01
G ðcÞ ¼ G01 þ 0 ð3Þ
1 þ cc -1.6
-0.4 -0.2 0 0.2 0.4
where c is a material parameter. The storage amplitude, defined as Logarithmic strain
DG0 ¼ G00  G01 is calculated for each volume fraction and plotted in
Fig. 5. It can be seen that DG0 is four times higher for N330 than for Fig. 6. Tension and compression tests for filled elastomers with carbon black N330.
The stress level and the hysteresis increase with filler volume fraction.
N650, for / = 18%. The evolution of DG0 for both N330 and N650 is
consistent with the percolation threshold values found in the previ-
ous paragraph. Indeed, a significant decrease of the storage modu-
Table 3
lus with strain is recorded for volume fractions above the Variation of the Young’s modulus E with the filler volume fraction /
percolation threshold.
/ E (MPa) – N330 E (MPa) – N650

2.5. Young’s and bulk moduli measurements Tension Compression Tension Compression
0.000 1.565 ± 0.015 1.600 ± 0.030 – –
Static mechanical tests in uniaxial tension and compression 0.024 1.855 ± 0.015 1.787 ± 0.020 1.680 ± 0.020 1.795 ± 0.050
were carried out to determine the mechanical behavior within a 0.068 2.305 ± 0.005 2.432 ± 0.010 2.215 ± 0.035 2.299 ± 0.050
0.128 3.715 ±0.065 3.692 ± 0.060 3.285 ± 0.065 3.311 ± 0.020
moderate strain range of e = ±0.4. Bone-shaped tensile samples
0.180 5.630 ± 0.016 5.344 ± 0.020 4.415 ± 0.045 4.599 ± 0.070
have a gauged volume of 20  4  2 mm3 (length per width per
 B. Omnès et al. / Composites: Part A 39 (2008) 1141–1149 1145

0 N330
7%v N330 N650
0.4999
Cauchy stress (MPa)

-40

Poisson ratio
0.4998
-80
oedometric unaxial compression
compression

-120 0.4997
-0.12 -0.08 -0.04 0 0 0.05 0.1 0.15 0.2
Logaritmic strain Φ
Fig. 7. Stress–strain curve for a 7% filled rubber in uniaxial compression and oed- Fig. 10. Variation of the Poisson’s ratio for N330 and N650 filled elastomers with
ometric compression. the carbon black volume fraction.

The tests were then performed three times on three different

Fig. 8 shows the values of K calculated by assuming a hydrostatic
samples to test the reproducibility. The samples are stabilised
state, as a function of the volume fraction of carbon black. K in-
under four loading–unloading cycles between 1000 N and the
creases with increase of the content of carbon black between
maximum test force imposed by the maximum capacity of the
3130 ± 15 MPa for unfilled natural rubber up to 3450 ± 50 MPa
testing machine, i.e. 95,000 N. The height variation of the sample
for 18%. The type of filler has only a slight influence on the values
is measured from the crosshead displacement minus the displace-
of K. A small dispersion of the bulk modulus is observed for the low
ment corresponding to the machine and device stiffness. The
volume fractions. The Poisson’s ratio are calculated from the
logarithmic strain–Cauchy stress curves (Fig. 7) are fitted on the
Young’s modulus in compression (Table 3 and Fig. 9) and the bulk
stabilised loading with a linear function between 40 MPa <
modulus on the stabilised loading.
r < 120 MPa. The contribution of the uniaxial stress state is less
The values of Poisson’s ratio m are shown in Fig. 10. It decreases
than 1 MPa and is therefore neglected within this stress range.
with the volume fraction of carbon black and the type of filler does
not affect significantly its values. For unfilled natural rubber the
calculated Poisson’s ratio is 0.49991 and for a high volume fraction
3750
of 18%, is equal to 0.49972. These values are in good agreement
with the values found in the literature [32]: the Poisson’s ratio is
found to remain above 0.499 for a natural rubber matrix.
Bulk modulus (MPa)

N330
N650
3500
3. Self-consistent model

In a previous work [21], a self-consistent model has been used,

3250 both in the linear and non-linear field, in order to predict the rein-
forcement effect, with the filler volume fraction as an input data.
Such a model was developed within a generalized self-consistent
framework, with two phases embedded in the homogeneous
3000 equivalent medium (HEM). The aim of this work is to enrich the
0 0.05 0.1 0.15 0.2
morphological pattern in order to take into account the different
Φ
phases.
Fig. 8. Variation of the bulk modulus for N330 and N650 filled elastomers with the Generally speaking, the classical self-consistent scheme as-
carbon black volume fraction. sumes that each phase is embedded in a homogeneous equivalent
medium. The effective elastic tensor of the HEM is derived from an
implicit equation depending on elastic tensors and volume frac-
tions of the phases [33]. To account more precisely of the geomet-
6 rical phase distribution, a generalized self-consistent scheme was
Young modulus (MPa)

N330
N650 developped [34] with a composite sphere constituted of N phases
embedded in the HEM. Such an approach is used in this study. Ana-
4 lytical expressions of the HEM elastic tensor components are then
derived from the elastic moduli of the phases as well as the volume
fractions.
2 The idealized microstructure of the reinforced natural rubber is
shown in Fig. 11. There exists a dispersion of aggregates in a ma-
trix of rubber, as evidenced in Fig. 1. These aggregates are made of
0 carbon black and occluded rubber and are surrounded by a layer of
0 0.05 0.1 0.15 0.2 bound rubber. Moreover, as evidenced by the experimental results,
Φ there exists a percolating network above a given threshold. In this
Fig. 9. Variation of the Young’s modulus for N330 and N650 filled elastomers, m-
work, only one morphological pattern is developed with a perco-
easured in compression on the stabilised loading, with the carbon black volume lating network, with its volume fraction decreasing down to al-
fraction. most zero below the percolation threshold. The proposed model
1146 B. Omnès et al. / Composites: Part A 39 (2008) 1141–1149

Fig. 11. Schematic view of the microstructure of a reinforced carbon black: aggregates, made of an inner core of occluded rubber and carbon black are surrounded by a thin
layer of bound rubber; these aggregates are dispersed within a labile rubber matrix and a percolating network, the structure of which is not detailed, bears the load. The thick
black lines represents the percolating network and the small dots pointed by the arrows represent the aggregates. A magnified view of an aggregate is displayed, with the
elementary carbon black particles, the occluded and bound rubber.

representing this microstructure consists in performing the surrounded by a carbon black phase which is embedded in a mod-
homogenization in two successive steps, as shown in Fig. 12. An ified polymer layer.
aggregate is represented by a homogenized medium HEMag. This Phase a. Volume fraction of occluded rubber fa is calculated from
medium is by definition equivalent to a composite sphere made the absorption of DPB, as found in [4]. This author proposed an
of an internal core of occluded rubber (index a), an intermediate effective volume fraction of filler /e, given by
layer of carbon black (index b) and an external layer of bound rub-
1
ber (index c). This aggregate is one of the three phases of a second /e ¼ ð1 þ ð1 þ 0:02139DBPÞ=1:46Þ/ ð4Þ
2
HEM, equivalent to a composite sphere made of an internal core of
HEMag (index 1), an intermediate layer of free rubber (index 2) and Therefore, fa = /e  /.
an external layer of a percolating network, also called the reinforc- Phase b. The volume fraction of carbon black is a known param-
ing phase (index 3) [19]. The percolating network connects the eter for each material. It is assumed that during elaboration of the
outer surfaces of the sample, and therefore is chosen to be the out- composite, carbon black interacts strongly with rubber and it leads
er phase of the HEM morphological pattern. Analytical expressions to the formation of aggregates. When the distance between aggre-
of the effective moduli of HEMag and HEM, as given in [34], depend gates decreases below a critical value, the percolating network ap-
on the volume fractions of each phase as well as on their elastic pears. Therefore, the total filler volume fraction is considered here
moduli. The next paragraph details the characteristics of each to estimate the chemical composition of the aggregate. Then, fb = /.
phase. Phase c. Finally, the bound rubber volume fraction is only an
estimate. Indeed, this phase represents the polymer whose behav-
3.1. Microstructure representation ior is modified by the interactions with the filler. Previous works
[9–11] have shown that there is a shell of immobilized rubber that
Let us consider first the aggregate structure. It is made of three surrounds the filler, with a given thickness. Starting from an
phases (indices a, b and c), i.e. an internal core of occluded rubber average size of the aggregate, represented by a perfect sphere of
 B. Omnès et al. / Composites: Part A 39 (2008) 1141–1149 1147

Occluded rubber (a) Phase 2. When considering only a hydrodynamic contribution,

ag
HEM the labile rubber corresponds to the natural rubber in the sample
minus the occluded and bound rubber. When the percolating net-
work exists, some of the aggregates and the labile rubber are in-
Carbon black (b) volved in the building of this network and therefore do not exist
any longer as individual phases. To take into account this phenom-
enon, it is assumed in this study that aggregates and labile rubber
are taken away in the same proportion as the one existing in the
Bound rubber (c) absence of the percolating network, i.e. f2/f1 = (frubber  fa  fc)/
(i) (fa + fb + fc), where frubber is the volume fraction of rubber in the
composite. When the percolating phase volume fraction is differ-
ag ent from zero, f1 + f2 = 1  f3 with a constant ratio f2/f1. It should
HME (1) be emphasized that this is a rough approximation and the mecha-
HEM nism of the formation of the percolating network should be inves-
tigated further to give a physical meaning to this ratio.
Phase 1. It corresponds to the aggregates, the mechanical prop-
erties of which are those of the HEMag.
Unbound rubber (2)
3.2. Elastic properties of the phases

Reinforcing phase (3) Compression and oedometric tests performed on the pure rub-
(ii) ber give the elastic moduli for the unbound rubber. The simplifying
assumption is made that these elastic properties are the same
Fig. 12. Morphological pattern proposed in this work. (i) The HEMag represents the
whatever the filler volume fraction. However, the swelling ratio
microstructure of the aggregates, made of occluded rubber, carbon black and bound
rubber (ii) the second HEM represents the structure at a higher scale: the aggreg- presented previously is not constant and the gap between pure
ates are dispersed in a labile rubber matrix and the percolating network connects rubber and a sample with / = 18% of filler is 18% and 21% for
the free ends of the sample. N330 and N650, respectively. The average molar mass in-between
reticulation points decreases when the filler volume fraction in-
creases, as can be seen in Table 2 and the Young’s modulus is
diameter dag, and given a thickness e of bound rubber, the volume known to depend on this parameter. But as a first step, this differ-
fraction of bound rubber can be calculated. Indeed, starting from ence was not taken into account in the model. Occluded rubber is
the filler volume fraction and the specific surface, and for a given shielded from the macroscopic deformation and its mechanical
volume of material and assuming that the particles are spherical, role is limited; therefore, its mechanical properties are the same
the diameter is given by 6/(qCBSp) where qCB is the carbon black as the ones of the pure rubber. Carbon black used as a filler is found
density. It gives dag equals to 0.02 lm and 0.06 lm for N330 and under a powder form and is therefore difficult to characterize by
N650, respectively, which is lower than the experimental sizes itself. Moduli typical of graphite are then considered. The mobility
measured from SEM micrographs. of the polymer chains in the vicinity of the filler surface is reduced,
The volume fraction of bound rubber fc is given below: leading to an increased modulus of the bound rubber. As a first
choice, a ratio of 10 is chosen between the Young’s modulus of
fc ¼ e/qCB Sp ð5Þ
bound rubber and of pure rubber. The elastic properties of the
In this work, a value of e = 2.5 nm is used. This expression does not occluded phase, the carbon black and the bound rubber are,
take into account the effective filler volume fraction resulting from respectively: Ea = 1.6 MPa, Ka = 3120 MPa, Eb = 10000 MPa, Kb =
the occluded rubber and Wang et al [13], in order to improve this, 8330 MPa, Ec = 20 MPa, Kc = 1905 MPa.
proposed that dag = 6b1.43/(qCBSp) with b = /e//. This modification Considering the second HEM, the elastic properties of the inner
was not considered in this work. phase (phase 1) are those of the HEMag and the ones of phase 2 are
It should be emphasized that the volume fraction of the aggre- equals to those of pure rubber. The percolating network is built
gate hence calculated, i.e fag = fa + fb + fc is lower than one; the pre- from close aggregates and its modulus is adjusted to fit the exper-
vious volume fractions are then normalized by fag when imental results. The chosen values are E1 = 1.6 MPa, K1 = 3120 MPa,
considering HEMag calculations but are given their real values E3 = 15 MPa and K3 is such that the Poisson’s ratio of this phase is
when considering the whole model. equal to the one of the HEMag. It should be emphasized that due to
The second HEM is also made of three phases: the previous the crude assumption made for estimating the volume fraction of
aggregate is surrounded by a layer of unmodified rubber, which the percolating network, it was necessary to adjust the Young’s
is embedded in a percolating network, when it exists. modulus of this phase on the experimental results. There is indeed
Phase 3. The guiding idea is that the storage modulus at low a strong relationship between f3 and E3. Further experiments on
strain is dependent mainly on the additive contribution of both composites with different volume fractions should be performed,
the hydrodynamic effect, which corresponds to a dispersion of in order to validate this approach. However, a partial validation
aggregates in the labile rubber, and of a percolating network. At comes from the use of two carbon blacks; indeed, the parameters
high strains, there is a breakdown of this network and there re- were determined for N330 carbon black and the prediction of elas-
mains only the hydrodynamic contribution. Assuming that the tic constants of N650 composites comes only from the dependance
mechanical behavior can be represented with a simple law of mix- on the specific surface given by Eq. (5) and Fig. 5.
tures, the volume fraction of the percolating network can be esti-
mated from Fig. 5. The main advantage of this assumption is that 3.3. Results and discussion
an order of magnitude of the volume fraction of the percolating
network comes from experimental results on the material and Experimental evolution of Young’s and bulk moduli with the fil-
therefore it integrates the influence of the specific surface on the ler volume fraction is used in order to determine the unknown
reinforcement effect. parameters of the different phases. The comparison of the
1148 B. Omnès et al. / Composites: Part A 39 (2008) 1141–1149

predicted values with experimental ones is plotted in Figs. 13 and 0.4

14. It can be seen that the model predictions are really close to
experimental values. aggregate N330
percolating network N330
Within the assumptions presented above, the volume fractions 0.3 aggregate N650

Volume fraction
percolating network N650
for the different phases are calculated. For occluded rubber, the
values reached for / = 18% are equal to 11% and 14% for N330
and N650, respectively. For bound rubber, the values reached 7% 0.2
and 2% for N330 and N650, respectively. The difference between
N330 and N650 comes from the specific surface, which is higher
for N330 than for N650 (Table 1). It is obvious from Eqs. (4) and 0.1
(5) that the occluded and bound rubber volume fractions are pro-
portional to the carbon black volume fraction. This leads to a con-
0
stant chemical composition of the aggregate, whatever the / 0 0.05 0.1 0.15 0.2
values. In fact, the aggregate is made of 30% of occluded rubber, Φ
51% of carbon black and 19% of bound rubber.
Concerning the aggregates and the percolating network, their Fig. 15. Predicted volume fractions of aggregates and percolating network for N330
and N650 carbon blacks.
volume fractions are plotted in Fig. 15. It can be seen that there
is only a slight difference between the two carbon blacks in the
aggregate volume fractions and the main difference comes from
the percolating network. being derived from cyclic tests as it was done in this work, out of
The percolating network is only indirectly characterized, e.g. simplicity’s sake.
from rheological measurements of the uncured materials and cyc-
lic tests of vulcanised materials. It can be found in the literature
4. Conclusions
that this network exists when aggregates become very close one
to the other, depending on the filler volume fraction and the spe-
The aim of this work was to propose a representative pattern for
cific surface. Therefore, its volume fraction should be predicted
the microstructure of a natural rubber filled with carbon black, in
analytically as a function of the carbon black properties rather than
order to take into account the influence of both the filler volume
fraction and the filler specific surface on the reinforcement effect.
In a first step, the microstructure of the materials was investigated
with SEM observations, swelling experiments and rheological tes-
8
N330 tings of the uncured materials. Mechanical testings including cyclic
N650 shear tests and tensile, compression and oedometric tests were
Young modulus (MPa)

6 performed. Then an isotropic morphological pattern is proposed

to estimate the elastic properties within a generalised self-consis-
tent approach. This pattern captures different phases, such as oc-
4 cluded and bound rubber and the main influence of the specific
surface is represented by a percolating network. The homogeniza-
tion is performed in two steps, first an aggregate made of occluded
2 rubber, carbon black and bound rubber is considered and in a sec-
ond step, this aggregate is surrounded by labile rubber and a per-
colating network. The predicted Young’s and bulk moduli are very
0
0 0.05 0.1 0.15 0.2 close to the experimental results. A validation step is now neces-
Φ sary, by adding materials with other volume fractions of carbon
black, that were not in the database. Further work is also focused
Fig. 13. Predicted Young’s modulus versus filler volume fraction. N330 in square, on the extension of this model to the non-linear field, as proposed
N650 in circle.
in [21,35] for a two phase material.

References

[1] Donnet JB. Nano and microcomposites of polymers elastomers and their
3600 reinforcement. Compos Sci Technol 2003;63:1085–8.
N330 [2] Vilgis TA. Time scales in the reinforcement of elastomers. Polymer
Bulk modulus (MPa)

N650 2005;46:4223–9.
[3] Kohls G, Beaucage G. Rational design of reinforced rubber. Solid State Mater Sci
2002:183–94.
3400 [4] Medalia A. Morphology of aggregates: vi. effective volume of aggregates of
carbon black from electron microscopy; application to vehicle absorption and
to die swell of filled rubber. J Colloidal Interface Sci 1970;32:115.
[5] Meissner B. Bound rubber theory and experiment. J Appl Polym Sci
3200 1993;50:285–92.
[6] Leblanc JL. Rubber–filler interactions and rheological properties in filled
compounds. Prog Polym Sci 2002;27:627–87.
[7] Heinrich G, Kluppel M, Vilgis T. Reinforcement of elastomers. Curr Opin Solid
State Mater Sci 2002;6:195–203.
3000 [8] Léopoldès J, Barrès C, Leblanc JL, Georget P. Influence of filler–rubber
0 0.05 0.1 0.15 0.2
interactions on the viscoelastic properties of carbon-black-filled rubber
Φ compounds. J Appl Polym Sci 2004;91:577–88.
[9] Berriot J, Montès H, Lequeux F, Long D, Sotta P. Evidence for the shift of the
Fig. 14. Predicted bulk modulus versus filler volume fraction with. N330 in square, glass transition near the particles in silica-filled elastomers. Macromolecules
N650 in circle. 2002;35:9756–62.
 B. Omnès et al. / Composites: Part A 39 (2008) 1141–1149 1149

[10] Berriot J, Lequeux F, Monnerie L, Montès H, Long D, Sotta P. Filler–elastomer [23] Poompradub S, Tosaka M, Kohjiya S, Ikeda Y, Toki S, Sics I, et al. Mechanisms of
interaction in model filled rubbers, a H NMR study. J Non-Cryst Solids strain-induced crystallization in filled and unfilled natural rubber
2002;307–310:719–24. vulcanizates. J Appl Phys 2005;97:103529-1-9.
[11] Berriot J, Montès H, Lequeux F, Long D, Sotta P. Gradient of glass transition [24] Berriot J, Lequeux F, Montès H, Pernot H. Reinforcement of model filled
temperature in filled elastomers. Polymer 2002;43:6131–8. elastomers: experimental and theoretical approach of swelling properties.
[12] Wolff S, Wang MJ. Filler–elastomer interactions. part iv: the effect of the Europhys Lett 2003;64:50–6.
surface energies of fillers on elastomer reinforcement. Rubber Chem Technol [25] Boochathum P, Prajudtake W. Vulcanisation of cis- and trans-polyisoprene and
1992;65:329–42. their blands: cure characteristics and crosslink distribution. Eur Polym J
[13] Wang MJ, Wolff S, Tan EH. Filler–elastomer interactions. Part viii: the role of 2001;37:417–27.
the distance between filler aggregates in the dynamic properties of filled [26] Hwang WG, Wei KH. Mechanical, thermal and barrier properties of nbr/
vulcanizates. Rubber Chem Technol 1993;66:178–95. organosilicate nanocomposites. Polym Eng Sci 2004;44:2117–24.
[14] Kraus G. Mechanical losses in carbon-black-filled rubbers. J Appl Polym Sci [27] Ravichandran K, Natchimutu N. Vulcanisation characteristics and mechanical
1984;39:75–92. properties of natural rubber-scrap rubber compositions filled with leather
[15] Yurekli K, Krishnamoorti R, Tse MF, Mcelrath KO, Tsou AH, Wang HC. Structure particles. Polym Int 2005;54:553–9.
and dynamics of carbon black-filled elastomers. J Polym Sci 2001;39:256–75. [28] Leblanc JL. Rubber–filler interactions and rheological properties in filled
[16] Fröhlich J, Niedermeier W, Luginsland HD. The effect of filler–filler and filler– compounds. Prog Polym Sci 2001;27:627–87.
elastomer interaction on rubber reinforcement. Composites: Part A [29] Payne AR. The dynamic properties of carbon black loaded natural rubber
2005;36:449–60. vulcanizates. part ii. J Appl Polym Sci 1962;VI:368–72.
[17] Wang MJ, Wolff S, Donnet JB. Filler–elastomer interactions. part iii: carbon- [30] Maier PG, Gritz D. Molecular interpretation of the payne effect. Kaut Gummi
black-surface energies and interactions with elastomer analogs. Rubber Chem Kunstst 1996;49:18–21.
Technol 1991;64:714–36. [31] Cassagnau Ph, Mélis F. Non-linear viscoelastic behaviour and modulus
[18] Alberola ND, Benzarti K, Bas C, Bomal Y. Interface effects in elastomers recovery in silica filled polymers. Polymer 2003;44:6607–15.
reinforced by modified precipitated silica. Polym Compos 2001;22:312–25. [32] Holownia BP. Effect of carbon black on the elastic constants of elastomers.
[19] Mélé P, Marceau S, Brown D, de Puydt Y, Albérola N. Reinforcement effects in Rubber Ind 1974:157–60.
fractal-structure-filled rubber. Polymer 2002;43:5577–86. [33] Christensen RM, Lo KH. Solutions for effective shear properties in three phase
[20] Gauthier C, Raynaud E, Vassoille R, Ladouce-Stelandre L. Analysis of the non- sphere and cylinder models. J Mech Phys Solids 1979;27:315–30.
linear viscoelastic behaviour of silica filled styrene butadiene rubber. Polymer [34] Hervé E, Zaoui A. n-layered inclusion-based micromechanical modelling. Int J
2004;45:2761–71. Eng Sci 1993;1(31):1–10.
[21] Omnès B, Henri T, Thuillier S, Pilvin P. Modelling of the mechanical behaviour [35] B. Omnès, S. Thuillier, P. Pilvin, S. Gillet. Non-linear mechanical behaviour of
of a filled elastomer with a self-consistent homogenization technique. In: carbon black reinforced elastomers: experiments and self-consistent
Fourth European conference on constitutive models for rubber (ECCMR), June modelling. In: Third China–Europe symposium on processing and properties
27–29 2005, Stockholm (Sweden); 2005. p. 197–202. of reinforced polymers, June 11–15 2007, Budapest (Hungary); 2007.
[22] Guz IA, Rodger AA, Guz AN, Rushchitsky JJ. Developing the mechanical models
for nanomaterials. Composites: Part A 2007;38:1234–50.