You are on page 1of 24

A kinetic model for silica-filled rubber reinforcement

Jean-Charles Majesté and Frédéric Vincent

Citation: Journal of Rheology (1978-present) 59, 405 (2015); doi: 10.1122/1.4906621


View online: http://dx.doi.org/10.1122/1.4906621
View Table of Contents: http://scitation.aip.org/content/sor/journal/jor2/59/2?ver=pdfcov
Published by the The Society of Rheology

Articles you may be interested in


The effect of polymer-induced attraction on dynamical arrests of polymer composites with
bimodal particle size distributions
J. Rheol. 57, 1669 (2013); 10.1122/1.4822254

Dynamics in coarse-grained models for oligomer-grafted silica nanoparticles


J. Chem. Phys. 136, 204904 (2012); 10.1063/1.4719957

A mesoscopic rheological model of polymer/layered silicate nanocomposites


J. Rheol. 51, 1189 (2007); 10.1122/1.2790461

A microscopic look at the reinforcement of silica-filled rubbers


J. Chem. Phys. 124, 174908 (2006); 10.1063/1.2191048

Nonlinear rheology of styrene-butadiene rubber filled with carbon-black or silica particles


J. Rheol. 50, 115 (2006); 10.1122/1.2167448

Redistribution subject to SOR license or copyright; see http://scitation.aip.org/content/sor/journal/jor2/info/about. Downloaded to IP:


134.153.184.170 On: Sat, 31 Jan 2015 02:04:18
A kinetic model for silica-filled rubber reinforcement

Jean-Charles Majestea) and Frederic Vincent

Universite de Lyon, F-42023, Saint Etienne, France; CNRS, UMR 5223,


Ingenierie des Materiaux Polymères, F-42023, Saint Etienne, France;
and Universite de Saint Etienne, Jean Monnet,
F-42023, Saint Etienne, France

(Received 21 May 2014; final revision received 8 January 2015; published 30 January 2015)

Synopsis
The competition between filler-filler interaction and filler-rubber interaction during the dispersion
process of silica-filled rubber has been investigated. Several complementary techniques were
carried out going from local observations of the dispersion to a global view given from linear and
nonlinear rheological measurements in order to lead to a better estimation of the dispersion kinetics.
It has been shown that reinforcement evolves with mixing time. A direct link between the bound
rubber amount and reinforcement indicators was found, revealing a replacement of strong
filler-filler interactions by weak rubber-filler ones. As a result, rheological reinforcement can be
cast under the form of a universal power law by introducing an effective interacting surface
between fillers. Finally, a kinetic model of the rubber reinforcement has been developed on the
basis of the competition between filler dispersion mechanism and rubber physical
adsorption. V C 2015 The Society of Rheology. [http://dx.doi.org/10.1122/1.4906621]

I. INTRODUCTION
Incorporation of nanofillers in a polymer matrix is a well-known technique to improve
polymer properties. In particular, nanofiller addition to an elastomeric matrix provides a
significant commercial importance because it permits a strong improvement in final prop-
erties such as increasing the mechanical reinforcement [Heinrich et al. (2002)].
Reinforcement of the rubber material has been defined as increased stiffness, modulus,
rupture energy, tear strength, tensile strength, cracking resistance, fatigue resistance, and
abrasion resistance [Dannenberg (1975)]. These properties, required for a great variety of
modern rubber products, depend on the quality of both dispersion and distribution states
[Hess (1991); Hess et al. (1984)]. Generally, the better the dispersion and distribution of
the fillers in the matrix are, the better the nanocomposite properties are [Leblanc (1996)].
For instance, in tire industry, generally speaking, dispersion of the fillers impacts the
viscoelastic energy losses of the rubber compounds and consequently the rolling resist-
ance [Schuring and Futamura (1990)].
A large variety of inorganic nanofillers have already been incorporated into polymeric
matrices, ranging from carbon black to various types of silica or even soft and hard clays.
As they are of high industrial interest [Guy et al. (2005); Guy et al. (2009)], silica

a)
Author to whom correspondence should be addressed; electronic mail: majeste@univ-st-etienne.fr

C 2015 by The Society of Rheology, Inc.


V
J. Rheol. 59(2), 405-427 March/April (2015) 0148-6055/2015/59(2)/405/23/$30.00 405

Redistribution subject to SOR license or copyright; see http://scitation.aip.org/content/sor/journal/jor2/info/about. Downloaded to IP:


134.153.184.170 On: Sat, 31 Jan 2015 02:04:18
406  AND F. VINCENT
J.-C. MAJESTE

nanocomposites have mobilized many research efforts both in academic and more
applied projects. More and more, for safety convenience, silica fillers are used as micro-
pellets with typical sizes around 100 lm. When dispersed as aggregates into a rubber
matrix, their large surface area (50–400 m2/g) emphasizes interparticle interactions and
fractal cluster creation, which consequently leads to a major impact on the rheological
and reinforcement properties of nanocomposites. This fractal structure added to the high
specific area and polar surface cause in most cases self-aggregation and can consequently
form a network of connected or interacting particles in the molten polymer. Rubber
reinforcement stems from such 3D structures and then will strongly depend on the nature
of the silica and the processing conditions [Choi et al. (2004)]. In brief, reinforcement is
thoroughly linked to the quality of the filler dispersion in the rubber matrix.
Measuring the quality of dispersion in a nanocomposite is not an easy thing to do.
Generally, one can distinguish two methods of evaluation. First, there is a
“morphological” approach which is mainly based on direct observations of both the sizes
and the spatial distribution of fillers. This approach covers a wide variety of techniques:
scanning electron microscopy (SEM) [White et al. (2006)], field emission SEM
(FESEM), transmission electron microscopy (TEM) [Aranguren et al. (1997); Corte et al.
(2005)], and scattering techniques, such as x-ray diffraction (XRD) or small angle x-ray
scattering (SAXS) [Schaefer and Justice (2007)]. Although this approach undoubtedly
brings precious information on the dispersion (mostly on the local microstructure), none
of the techniques previously mentioned is able to capture suitably all the levels of disper-
sion and structure. The alternative is to use the second method of evaluation which is
rather based on the properties (mechanical, rheological, electrical, and permeability) of
the material. Measuring its macroscopic properties offers a global view of the material’s
performance. In particular, linear rheology is a technique generally used to assess the
state of dispersion of nanocomposites directly in the melt state [Galindo-Rosales et al.
(2011); Cassagnau (2008)].
Using equally one of the two approaches, numerous studies have focused essentially
on the parameters affecting the final distribution and dispersion of nanofillers. The mix-
ing process is generally the first lever that can control the final dispersion state. By apply-
ing a sufficient stress to overcome the cohesive forces of silica agglomerates, the size of
disperse silica particles reaches the nanoscale (10–100 nm) ensuring that the interparticle
distance is small enough to induce reinforcing effects [Rumpf (1962); Mele et al. (2002);
Horwatt et al. (1992a, 1992b)]. Despite the good level of dispersion reached by optimiz-
ing the process parameters (shear rate, stress, fillers concentration, and mode of incorpo-
ration of silica), final properties are often poor due to the weak filler–rubber interactions
[Wolff and Wang (1992); Wolff et al. (1993); Park and Cho (2003)] (low London disper-
sive component of surface energetics of silica). Therefore, various surface treatments
{going from thermal, chemical, and electrochemical to coupling agent treatments [Dugas
and Chevalier (2003); Park and Cho (2003); Castellano et al. (2005); Scurati et al.
(2002); Mathew et al. (2004)]} are needed to improve the reinforcement of the compo-
sites. Generally speaking, reinforcement is influenced by the thermodynamic interaction
between the matrix polymer and the bounded molecules which physically or chemically
attach themselves to the filler surface during the mixing process [Leblanc (2002);
Frohlich et al. (2005); Bohm et al. (2010)].
In brief, if there is experimental evidence that reinforcement originates from both
filler-filler interaction and filler-rubber interaction, the competition between the two
mechanisms is not completely highlighted. Most of the previous works have focused on
the final dispersion state or essentially have described kinetics of agglomerates’ size
reduction.

Redistribution subject to SOR license or copyright; see http://scitation.aip.org/content/sor/journal/jor2/info/about. Downloaded to IP:


134.153.184.170 On: Sat, 31 Jan 2015 02:04:18
A KINETIC MODEL FOR RUBBER REINFORCEMENT 407

In the present work, the kinetics of rubber reinforcement by silica nanofillers without
any surface treatment during mixing process is investigated. The goal is to put clearly in
evidence the competition between filler-filler and filler-rubber interaction during the dis-
persion process and consequently to show its impact on the final structure and reinforce-
ment of the nanocomposite. To this end, silica/rubber nanocomposites were created and
analyzed at different mixing times. The structure of the filler network and the progress in
dispersion process were obtained by coupling several characterization techniques going
from microscopy (for local assessment of the structure) to rheological measurement (for
global view and reinforcement). Then, a strong correlation between the amount of physi-
cally bounded rubber and the level of reinforcement will be shown. Finally, a kinetic
model of the rubber reinforcement is proposed on the basis of the competition between
filler dispersion mechanism and rubber physical adsorption.

II. EXPERIMENTAL
A. Materials
Neat styrene-butadiene-rubber (SBR) provided by Michelin was used as dispersive
matrix. NMR characterization indicates a mass composition of 25% styrene and 58% bu-
tadiene in the trans 1,2 form and 17% in the trans 1,4 form. Density at room temperature
is 0.938, glass transition temperature is 30  C, molecular weight is 310 000 g/mol, and
Ip ¼ 1.5. SBR matrix remains stable up to 400  C.
The two fillers chosen in this study are highly dispersible precipitated silica (Zeosil
Z1115MP and Zeosil Z1165MP) provided by Rhodia/Solvay; both silicas were received
as micropellet powder with average diameters around 250–300 lm. These silicas differ
from each other by their specific area and consequently by their porosity and mean aggre-
gates’ diameter. The density of hydroxyl groups at the surface was estimated for each
silica but one has to consider carefully these values owing to the high uncertainties even
if that reveals slight differences in the surface chemistry. The amount of filler has
been set to 20, 25, 30, and 30 wt. %. Characteristics of the two silicas are summarized in
Table I.

B. Sample preparation and mixing


R
Nanocomposites were prepared by using a Rheomix HAAKE 600V internal batch
mixer with roller rotor. SBR was first introduced and masticated at 80  C and 28.3 s1
(50 RPM) for 1 min. Then dried silica was second introduced. Filling time was around
30 s. No antioxidant was added into the blends to avoid preventively interaction with
silica surface. Rubber degradation was investigated by SEC. Change of molar mass is
observed during the earliest stage of mixing (decrease of SBR molar mass less than 10%
TABLE I. Characteristics of both precipitated silicas Z1115MP and Z1165MP.

Technique Measure Z1115MP Z1165MP

Humid particle size analysis Median diameter D50 (lm) 283 6 2 279 6 2
BET Specific area (m2/g) 111 153
CTAB Specific area (m2/g) 107 150
Porosimetry Hg Average diameter of pores (nm) 48 29
BET þ Pycnometry He Aggregates diameter 24 17
ATG Humidity—2 h/105  C (%) 7.6 6.5
— -OH number per nm2 8–10 6–8

Redistribution subject to SOR license or copyright; see http://scitation.aip.org/content/sor/journal/jor2/info/about. Downloaded to IP:


134.153.184.170 On: Sat, 31 Jan 2015 02:04:18
408  AND F. VINCENT
J.-C. MAJESTE

from 310 000 to 277 000 g/mol) and remains constant beyond 2 min. The mixer chamber
(50 cm3) was filled up to 70%. Afterward, the whole composite was masticated at
130 6 5  C and 28.3 s1. Thus, the conversion of the rotation speed N (in RPM) into a
shear rate value has been done according to the calculation method proposed by
Bousmina et al. (1996). For our mixing device, we have found c_ ¼ 0:566 N. Temperature
was controlled as far as possible (65  C by alternating cooling and heating in the cham-
ber of the internal mixer) taking into account self-heating of this high viscosity system
during the first minute of mixing. Time zero of the dispersion process was set at the end
of silica introduction in the mixer (Fig. 1). Mixing time varied from 1 to 50 min keeping
the shear rate constant. Although such batch times have no industrial reality (almost
5 min in internal mixer for tire industry followed by calendering operation), the mixing
process has been continued to be sure that a stationary state is reached.
For comparison purpose, a silane coupling agent (TESPT: bifunctional silane
bis(3-triethoxysilylpropyl)-tetrasulfide) has been introduced in the mixing process for
one unique compound filled with 30 wt. % of silica Z1165MP. Classically, the silane cou-
pling agent of 8.0 wt. % of the silica content was first added in the dry silica. Then,
TESPT/silica mixture was introduced in the rubber according to the same aforementioned
procedure as for untreated silica.

C. Bound rubber measurement


When an elastomer and reinforcing filler are mixed together, interactions occur
between the two components leading to the formation of an immobilized (or weakly im-
mobilized) polymer layer at the filler surface. Hence, a good solvent of the elastomer can
extract only a free rubber portion, resulting in a partial loss of solubility and leaving a
highly swollen rubber-filler gel. The well-known “Bound rubber” is, by definition, the
rubber content of that gel. Leblanc (2000) defined the bound rubber content by
mboundrubber
%BdR ¼ ; (1)
msample  /rubber
where mbound rubber is the mass of rubber still left after complete extraction of the free rub-
ber matrix. msample is the mass of the whole sample and /rubber stands for the mass frac-
tion of rubber in the blend.

FIG. 1. Evolution of measured internal mixer torque and chamber temperature over mixing time for SBR filled
with 30 wt. % of silica Z1165MP.

Redistribution subject to SOR license or copyright; see http://scitation.aip.org/content/sor/journal/jor2/info/about. Downloaded to IP:


134.153.184.170 On: Sat, 31 Jan 2015 02:04:18
A KINETIC MODEL FOR RUBBER REINFORCEMENT 409

The following protocol was defined to obtain the value of the bound rubber content.
This quantity was estimated by difference of the mass of filler introduced in the mixture
and that recovered after extraction of free SBR molecules [Choi (2002)]. Five successive
extractions were performed using tetrahydrofurane (THF). Each of them went on around
24 h in order to optimize the dissolution of the free SBR in the nanocomposite. By means
of centrifugation, the solution of unbound was extracted and the remaining solid phase
was dissolve again in THF during 24 h. Bound rubber content (%BdR) was then obtained
for each nanocomposite sample and determined with a Fisher-Student law (certitude
interval at 95%).

D. TEM
TEM observations were performed using a HITACHI H800-3 Transmission Electron
Microscope. Pictures were obtained in bright field with an acceleration voltage of
200 kV. Nanocomposites were first frozen at a temperature near 90  C (far lower than
the SBR glass transition, 30  C) and cut in ultrathin films (<80 nm). Thin films were
then stuck onto the TEM grids waiting for observations. The range of the magnification
spreads from  5000 to  200 000. Examples of images obtained from the TEM observa-
tions are shown in Fig. 2 after following the mixing procedure previously described.
Dark zones are identified as silica particles while the matrix looks bright. According to
the state of the dispersion, agglomerates (1–100 lm) or aggregates (40–100 nm) can be
observed. At the nanometric scale, the brightest domain (i.e., the thinnest zone) has been
systematically chosen. In this way, false interpretations due to the superposition of aggre-
gates along the thickness are limited. As the average size of aggregates was always found
lower than the thickness of the film, the Saltikov correction was not used. Size characteri-
zation of the particles was performed using IMAGEJ coupled with GIMP FREE software. First,
threshold and binarization of all the pictures were manually performed by creating masks
over the isolated particles (aggregates or agglomerates). This has been done in numerous
and various domains of the sample to satisfy statistical criterions. Second, using the
threshold images, the radius Ri of each object was determined using image processing
analysis provided by IMAGEJ. Then the average radius number of the filler system is given
by

FIG. 2. TEM micrographs of SBR filled with 30 wt. % of silica Z1165MP as a function of mixing time.
Magnification  10 000 (upper row) to highlight agglomerates’ size reduction,  150 000 (lower row) to show
the creation of aggregates.

Redistribution subject to SOR license or copyright; see http://scitation.aip.org/content/sor/journal/jor2/info/about. Downloaded to IP:


134.153.184.170 On: Sat, 31 Jan 2015 02:04:18
410  AND F. VINCENT
J.-C. MAJESTE

X
Ni  Ri
i
Rn ¼ X : (2)
Ni
i

Uncertainties were estimated at 20%.

E. Dynamic spectroscopy
Dynamic rheological measurements were performed using a rheometer MCR301 from
Anton Paar equipped of parallel plate geometry. Two kinds of measurements were
carried out. In order to extract information from the filler network structure, frequency
sweep measurements were carried out at 100  C from 100 to 0.004 rad/s with a deforma-
tion set to 0.1%, to be sure to remain in the linear viscoelastic regime. No time-
temperature superposition has been used. The axial force was kept constant around 5 N.
Measurements were reproduced twice. No differences between increasing and decreasing
frequency sweeps show the stability of the filler network (flocculation mechanism, for
instance) during the test as already observed for such system at high temperature under
oscillatory flow [Bohm et al. (2010); St€ockelhuber et al. (2011)]. Moreover, under resting
conditions (without flow) at room temperature, evolution of the mixture following mixing
has been checked. It was found that until 8 days (at least) the rheological properties
remain very stable. Nevertheless, all the samples were even though stored in a cool and
dry place after the mixing process.
Amplitude sweep measurements were carried out at increasing strain, 100  C and
1 rad/s in order to assess the nonlinear rheological behavior. As for the previous test, the
axial force was controlled. For both kinds of experiment, the sample preparation was
performed by compression molding in the same conditions for all the samples. Loading
of the sample into measuring cell was also similar as much as possible.

III. RESULTS AND DISCUSSION


A. Silica dispersion
In a classical way [White et al. (2006)], evolution of the dispersion state was first
estimated using TEM observations. Both systems (Z1115MP/SBR and Z1165MP/SBR)
were examined at different mixing times. As indicated in Sec. II, the time zero was set
after total introduction of the silica amount. Then, the internal mixer was stopped at the
desired mixing time. The whole batch was removed from the chamber and was stored in
a cool and dry place. Figure 2 shows TEM observations for Z1165MP silica versus mix-
ing time. Two scales of observation are given. On the large scale, the evolution of the
agglomerates is visible. On the small scale, changes in the aggregates’ size and structure
can be observed. A first look clearly shows the disappearance of the largest agglomerates
with the increase of the mixing time. After 25 min of mixing, less than 1 vol. % of unbro-
ken objects still remain. As the mixing time increases, the initial agglomerates have been
dispersed into isolated aggregates and/or clusters of aggregates. For more quantitative
analyses, by following the procedure described in Sec. II, we estimated the change in the
size of isolated aggregates and gathered the results in Table II as a function of the mixing
time.
The size of aggregates stabilizes after 2 and 7 min, respectively, for Z1115MP and
Z1165MP silicas indicating the completion of the dispersion. However, these indicators
will be compared to rheological data to confirm the results.

Redistribution subject to SOR license or copyright; see http://scitation.aip.org/content/sor/journal/jor2/info/about. Downloaded to IP:


134.153.184.170 On: Sat, 31 Jan 2015 02:04:18
A KINETIC MODEL FOR RUBBER REINFORCEMENT 411

TABLE II. Evolution of the average radius number Rn (agglomerates and aggregates) versus mixing time for
both silicas Z1115MP and Z1165MP.

Silica Mixing time (min) 1 5 10 50

Z1115MP Average radius Rn aggregates (nm) 55 43 38 37


Average radius Rn agglomerates (lm) 1 0.55 0.37 0.25
Z1165MP Average radius Rn aggregates (nm) 90 50 40 41
Average radius Rn agglomerates (lm) 1 0.5 0.5 0.5

B. Nonlinear rheology
According to Leonov (1990), a bump observed on the evolution of G00 (c), at constant fre-
quency, is associated with a mechanism of desagglomeration/organization of the silica
induced by the strain. This is particularly the area under the bump (visible for G00 ) which
reflects a variation in the size of agglomerates, as confirmed by showing the disappearance
of bumps once desagglomeration is complete [Galindo-Rosales et al. (2011)]. The existence
of agglomerates (flocks of particles) causes additional dissipative processes in such systems.
Experiments within polymer matrices of the same chemical composition and nature but hav-
ing very different rheological behaviors and therefore different dispersion capabilities have
shown the correlation between agglomerates and dissipation within the system at low fre-
quency [Carrot et al. (2010); Majeste et al. (2012)]. Dissipation of energy was attributed to
the rupture of agglomerates and reorganization into the flocks. According to the same
authors, the plateau observed at low frequency on the loss modulus [Fig. 4(b)] comes under
the same physics and reveals the kinetics of the dispersion mechanism.
Such nonlinear rheological measurements have been performed at different mixing
times. Figure 3 shows the variations of the loss modulus versus the applied strain for
Z1165MP silica-filled system. Clearly, a bump is observed for short mixing times. This
phenomenon vanishes steadily until it disappears after 7 min. These critical times indicate
the disappearance of large agglomerates in the system and are in agreement with TEM
observations. As the mixing time increases, both the number and the size of the agglom-
erates decrease and, consequently, so does the associated energy dissipation. As a result,
this peculiar rheological signature is then a very interesting qualitative marker of the dis-
persion, for silica-filled rubber at least.

FIG. 3. Loss modulus at a constant frequency (x ¼ 0.1 rad/s) as a function of strain amplitude for SBR filled
with 30 wt. % of silica Z1165MP for various mixing times under the conditions defined in Sec. II.

Redistribution subject to SOR license or copyright; see http://scitation.aip.org/content/sor/journal/jor2/info/about. Downloaded to IP:


134.153.184.170 On: Sat, 31 Jan 2015 02:04:18
412  AND F. VINCENT
J.-C. MAJESTE

FIG. 4. Storage (a) and loss (b) moduli of Z1165MP nanocomposites versus mixing time. The plateau Ge indi-
cating the presence of a silica particle network is visible at very low frequencies. The content of silica
Z1165MP is 30 wt. %.

In order to highlight the reinforcement effect, theoretical hydrodynamic contribution


has been calculated using work of Domurath et al. (2012) on stress and strain amplifica-
tion. Figure 3 shows that, even for long mixing time, calculated hydrodynamic contribu-
tion is lower revealing additional contributions to the whole rheological behavior. All
these effects evolve suggesting changes in the filler network. In Sec. III.C, by investigat-
ing the linear viscoelastic behavior, we get further information on the structure evolution
of these systems during the mixing process.

C. Linear viscoelasticity under oscillatory shear


Small-amplitude oscillatory shear flow in the melt is often claimed to be a suitable
method for the characterization of the state of dispersion of particulate fillers in a poly-
mer matrix. In particular, the occurrence and the level of a secondary plateau at low

Redistribution subject to SOR license or copyright; see http://scitation.aip.org/content/sor/journal/jor2/info/about. Downloaded to IP:


134.153.184.170 On: Sat, 31 Jan 2015 02:04:18
A KINETIC MODEL FOR RUBBER REINFORCEMENT 413

frequency are considered to be the trace of a finely and well-dispersed viscoelastic solid
phase in a molten matrix. Indeed, in oscillatory shear flow, the storage and loss moduli of
homogeneous polymer melts can be well described using the relaxation modulus of the
melt
ð1 ð1
G0 ðxÞ ¼ x GðtÞ sin ðxtÞdt and G00 ðxÞ ¼ x GðtÞ cos ðxtÞdt: (3)
0 0

In addition, G(t) tends to zero for infinitely long times. Both the existence of a
Newtonian viscosity and the zero value of G(t) define a liquidlike behavior.
Networking yields the conversion of a liquidlike behavior toward a solidlike behavior
with a relaxation modulus reaching a constant value Ge at time infinity. Therefore, the
real part of the complex moduli G0 become
ð1 X
0 ðxki Þ2
G ðx Þ ¼ G e þ x ½GðtÞ  Ge sinðxtÞdt or G0 ðxÞ ¼ Ge þ gi 2
: (4)
0 i 1 þ ðxki Þ

Figure 4 shows both the variations of the storage and loss modulus measured for the
Z1165MP silica-filled samples taken off from the mixing device at different mixing
times. From the beginning of the dispersion, a secondary plateau is being observed on
both the storage and loss modulus in a low frequency zone indicating the presence of a
silica particle network. Particularly, for G0 , the value of the modulus Ge characterizing
the plateau changes with the mixing time. Moreover, the plateau seems to vanish for the
longest times. In order to follow more accurately this evolution, values of Ge were
inferred from the determination of the relaxation spectrum using Marquad-Levenberg
minimization algorithm and Eq. (4) with regularization method and are gathered in
Fig. 5.
Surprisingly, the value of the plateau Ge increases with the mixing time up to 2 or
7 min, respectively, for Z1115MP and Z1165MP silicas and then decreases and levels off

FIG. 5. Evolution of the values of Ge as a function of the mixing time for both Z1115MP (䊏) and Z1165MP
(䊉) silica nanocomposites with 30 wt. % of silica content. Full lines indicate the prediction of Eq. (20) with the
coefficients of Table III. Dashed line represents the evolution of Ge for Z1165MP with a surface modified in the
mixing process by a bifunctional silane bis(3-triethoxysilylpropyl)-tetrasulfide (TESPT, Si 69).

Redistribution subject to SOR license or copyright; see http://scitation.aip.org/content/sor/journal/jor2/info/about. Downloaded to IP:


134.153.184.170 On: Sat, 31 Jan 2015 02:04:18
414  AND F. VINCENT
J.-C. MAJESTE

for the longer time. The increase of the modulus Ge at the beginning of the dispersion
process can be well explained in the frame of a filler network. The continuous breakup of
silica agglomerates increases the number of small aggregates which give structure to the
solidlike network. However, the value Ge of the plateau decreases after a critical time
correlated with the values found for TEM visualization or nonlinear rheological measure-
ments. This decrease cannot be related to changes in the dispersion state. In other words,
beyond the maximum observed in Fig. 5, TEM observations clearly show that the size
and then the number of aggregates remain constant. The decrease of Ge is thus associated
with another parameter which affects the nature of the particle network. Specific interac-
tions between the silica aggregates (filler/filler interaction) will then be investigated along
with their evolution with the mixing time. The next section will show how the variation
of the bound rubber adsorbed at the aggregate surface can affect the filler network.

D. Impact of bound rubber on reinforcement


The filler-rubber interaction is one of the dominant parameters influencing the disper-
sion of fillers. These interactions generally depend both on the topology and the chemis-
try of the filler surface and on polymer chemistry [Wolff and Wang (1992); Wolff et al.
(1993); Choi (2001)]. This interaction between filler and matrix is responsible for the
physical adsorption of rubber chains {loosely bound rubber, [Choi (2002)]}, as well as
the infiltration of the polymer in silica agglomerates (occluded rubber or trapped rubber).
Figure 6 shows evolution of the bound rubber content versus the mixing time for both
silicas and filler content of 30 wt. %.
For both silicas, the bound rubber content increases continuously with the mixing time
and reaches an asymptote for the long times. However, clear differences between the two
silicas used are visible. First, the maximum bound rubber value obtained for long time is
around 35% for Z1165MP, while it is 25% for Z1115MP. Two main parameters affect
this value. The amount of adsorbed polymer (rubber shell) is directly linked to the free
available surface and therefore to the specific surface and porosity when complete disper-
sion is achieved. Moreover, for a specific given surface, the maximum amount of bound
rubber depends on the polymer’s affinity with the aggregates’ surface. The chemical
nature of the surface (for instance number of hydroxyl groups OH per unit surface) drives

FIG. 6. Evolution of the bound rubber versus mixing time for both Z1115MP (䊏) and Z1165MP (䊉) silica
nanocomposites with 30 wt. % of silica content. Lines indicate the evolution functions fitted according to
Eq. (5) with the coefficients of Table III.

Redistribution subject to SOR license or copyright; see http://scitation.aip.org/content/sor/journal/jor2/info/about. Downloaded to IP:


134.153.184.170 On: Sat, 31 Jan 2015 02:04:18
A KINETIC MODEL FOR RUBBER REINFORCEMENT 415

the density of adsorbed polymer chains [Wolff and Wang (1992); Wolff et al. (1993)].
Hence, for these systems with similar surface activities, the ratio of specific surfaces is
quite close to that of the maximum bound rubber values. Besides, the maximum bound
rubber value increases with filler content in agreement with the increase of free specific
surface [Choi (2001, 2002)] (%BdR(/) ¼ 4.25 /2.09 for Z1165MP and %BdR(/) ¼ 5.12
/2.52 for Z1115MP). However, for both silicas, exponent of the power law is twice the
value expected if equilibrium aggregate size is not dependent of filler content (specific
area should then scale as //Ragg). This difference is likely related to variations of the final
aggregate size with filler content in agreement with the corresponding increase of the sur-
rounding hydrodynamic stresses [Boudimbou (2011)].
Second, the kinetics of the adsorption is different for both silicas. The kinetics depends
on chain diffusion time and the variation of the amount of free available silica surface. In
our mixing conditions, the diffusion time is determined by the temperature and then
remains constant if the temperature is controlled. On the other hand, the kinetics of the
silica dispersion drives the amount of free available surface. Hence, the evolution of the
bound rubber content at the beginning of the mixing characterizes how the silica is dis-
persed (rupture or erosion). A rapid increase in the bound rubber content at short times
(Z1115MP) means a rupture in large fragments while a more progressive and slower ero-
sion mechanism (case of Z1165MP) results in a later increase. Our results fairly confirm
that erosion mechanism (Z1165MP) is slower than rupture mechanism (Z1115MP)
[Boudimbou (2011)].
The overall evolution of the bound rubber %BdR with the mixing times is well cap-
tured by the following expression using two parameters (%BdRmax) and (CBR)

%BdRðtÞ ¼ %BdRmax ð1  exp ðCBR  tÞÞ; (5)

where (%BdRmax) is the maximum bound rubber and (CBR) is a kinetic constant depend-
ing on the physical adsorption process. Moreover, interestingly the kinetic constant (CBR)
scales as the strain rate c_ (Fig. 7)

FIG. 7. Variation of the kinetic parameter CBR as a function of the shear rate c_ for Z1115MP (䊏) and Z1165MP
(䊉).The content of silica Z1165MP is 30 wt. %. Straight line is only guide for the eyes.

Redistribution subject to SOR license or copyright; see http://scitation.aip.org/content/sor/journal/jor2/info/about. Downloaded to IP:


134.153.184.170 On: Sat, 31 Jan 2015 02:04:18
416  AND F. VINCENT
J.-C. MAJESTE

CBR ¼ kBR  c_ : (6)

Values of both parameters %BdRmax and kBR are given in Table III.
Bounding kinetics appears to be mainly related to dispersion kinetics, i.e., the increase
of free accessible silica surface during the mixing process. Therefore, for both silicas, the
bound rubber content increases during the mixing and consequently must replace filler-
filler interactions by filler-rubber ones. The strength of the filler network (the modulus
Ge) is consequently affected by this adsorption.
A way to investigate the impact of bound rubber on the filler network is to estimate
the variation of the number of filler-filler contacts with mixing time. Unfortunately, this
quantity is absolutely not available because of its tricky dependence on the various struc-
tures that filler network can constitute and also on the specific surface of interacting
particles. In other words, for an aggregate, the coordination number with close neighbors
can vary depending on the kind of structure (fractal dimension, compactness, etc.) and
the whole number of contact per unit volume depends on the specific area. However, con-
sidering the replacement of filler-filler contacts by filler-rubber ones, the number of con-
tacts (nc) must scale with the quantity of free surface (the area without adsorbed rubber)
as described by the following equation:
 
0 %BdR
nc ¼ nc 1  ; (7)
%BdRmax

where nc0 is the number of contacts between fillers that should be observed for unwetted
network and nc0 depends on the structure of the network and the state of dispersion.
In Fig. 8, when plotting the modulus Ge versus (nc) in double logarithmic manner, for
times beyond the maximum of Ge, i.e., constant nc0, a very good correlation is found for
both silicas following the general power law:
 
0 %BdR b
Ge ¼ f ðnc Þ ¼ Ge ð/Þ  1  : (8)
%BdRmax

Both G0e and (b) are parameters linked to the structure of filler network. G0e corresponds
to the modulus generated by unwetted silica network {average elastic bending-twisting
modulus of the different kinds of angular deformations of the bonds between particles
[Kl€uppel (2003)]}, without any rubber adsorption and with the same structure. G0e must
be dependent on filler specific surface and filler fraction u.
Actually, the last expression has similarities with usual scaling laws encountered in lit-
erature. Heinrich et al. (2002) have already taken into account the impact of bound rub-
ber content on the modulus Ge and proposed the following equation which can be cast
under a more general form:

Ge ¼ G0  /aef f ; (9a)

TABLE III. Parameters of rubber adsorption kinetics and parameters of dispersion kinetics for the two kinds of
silica.

Silica kBR %BdRmax (%)a G0 (MPa) GH (MPa)a cd


4
Z1115MP 1.02  10 25 394 0.032 2200 6 100
Z1165MP 8.8  105 33 2400 0.036 6000

a
These values of %BDRmax and GH are given for / ¼ 30 wt. %.

Redistribution subject to SOR license or copyright; see http://scitation.aip.org/content/sor/journal/jor2/info/about. Downloaded to IP:


134.153.184.170 On: Sat, 31 Jan 2015 02:04:18
A KINETIC MODEL FOR RUBBER REINFORCEMENT 417

 
FIG. 8. Variation of the value of Ge as a function of 1  %% BdR
BdRmax
, the fraction of free surface, for Z1165MP
nanocomposites (a) and Z1115MP nanocomposites (b) for various amounts of silica in the rubber matrix (vol-
ume fractions are indicated on graphs). Lines are just visual guides. Inset: Complete rescaling of the data by
plotting Ge versus the effective volume fraction /ef f .

with

3 þ df ;B
a¼ (9b)
3  df

and

ðd þ 2DÞ3 –6dD2
/ef f ¼ /: (9c)
d3

The exponent (a) reflects the characteristic structure of a fractal filler cluster and of the
corresponding filler network. (a) is related to the fractal dimensions df and df,B which
characterize the fractal structure and their backbone, respectively, of the filler clusters.

Redistribution subject to SOR license or copyright; see http://scitation.aip.org/content/sor/journal/jor2/info/about. Downloaded to IP:


134.153.184.170 On: Sat, 31 Jan 2015 02:04:18
418  AND F. VINCENT
J.-C. MAJESTE

Generally, the exponent (a) lies between 4.2 and 7.2 depending on the nature of the filler
[Cassagnau (2008)]. (d) is the size of an aggregate in the cluster and (D) is the thickness
of the bound rubber layer.
In expression (9), Kl€uppel [Heinrich et al. (2002); Kl€uppel (2003)] introduce the effect
of bound rubber through a modification of the effective filler volume fraction. This equa-
tion considers the mechanically effective solid volume of the clusters, approximately,
by enlarging the particle diameter from d to d þ 2D and subtracting the volume
V ¼ 2p=3D 2 ð3 ðd=2 þ D Þ –D Þ that results from the intersections of the layers of thick-
ness D at the contact points between two neighboring particle. This modification leads to
a nonscaling invariant relation for the concentration dependency of the scaling modulus.
The relation is then dependent on filler particle size. Moreover, Eq. (9) implies a constant
surface covering by the bounded rubber and only a variation in the thickness of the im-
mobilized rubber layer. For silica-filled systems, this hypothesis does not hold because
the presence of an immobilized rubber layer is not really observed.
Figure 8 rather suggests that the modulus is linked to the number of contacts (nc)
between aggregates inside the solid network owing to the certain uncorrelated behavior
between the amount of free surface and the volume fraction of filler. Actually, for a given
fraction of free surface on the aggregates, the total numbers of hard contacts (nc) must
scale proportionally to the effective volume fraction of fillers in the same way as Ge [Eq.
9(a)]. The effective volume fraction measures the effective interacting surface.
Thus, Fig. 8 suggests that the modulus Ge can be effectively described by Eq. 9(a) but
using the following expression of the effective volume fraction:
 1=3
%BdR
/ef f ¼/ 1 : (10)
%BdRmax

The exponent 1/3 is deduced from statistical arguments on the contacts between partially
covered aggregates. Three kind of contact are possible: covered surface/covered surface,
covered surface/bare surface, and bare surface/bare surface. Among these three possibil-
ities, only one leads to hard interaction: bare surface on bare surface. The exponent 1/3
describes this effect.
Thus, the modulus Ge can be expressed as
"   #a
%BdR 1=3
Ge ¼ G0e  / 1 : (11)
%BdRmax

Plotting the values of the modulus Ge versus the volume fraction of filler, the silicas
Z1115MP and Z1165MP follow Eq. 9(a) with exponents, respectively, equal to 6.2 and
4.47 (Fig. 9).
According to Fig. 8 and Eq. (11), the exponent b must be equal to a/3. We found
b ¼ 2.04 and b ¼ 1.49, respectively, for Z1115MP and Z1165MP. The results perfectly
agree with Eq. (11).
The value of the plateau modulus at low frequency for the storage modulus G0 is usu-
ally known as a sensitive and very interesting tool to characterize the filler network.
However, it must be used with care as its value depends strongly on the loosely bound
rubber amount. Thus, using Ge, the dispersion states have to be compared with one
another at the same bound rubber amount in order to infer real structure information.
Otherwise, the alternative way is to use G0e which only depends on structural
parameters.

Redistribution subject to SOR license or copyright; see http://scitation.aip.org/content/sor/journal/jor2/info/about. Downloaded to IP:


134.153.184.170 On: Sat, 31 Jan 2015 02:04:18
A KINETIC MODEL FOR RUBBER REINFORCEMENT 419

FIG. 9. Power law fits of the plateau Ge as a function of filler volume fraction / for both Z1115MP (䊏) and
Z1165MP (䊉) nanocomposites. The values of Ge has been measured at the same bound rubber fraction (%BdR/
%BdRmax ¼ 0.4).

IV. SCENARIO OF DISPERSION


Now, compiling all the characterization results, we are able to propose a scenario for
the silica dispersion mechanisms in the BR matrix. Figures 5 and 6 suggest a universal
behavior with only different kinetics. We will then focus on the marker Ge which
includes the evolution of all the parameters (%BdR, filler content, aggregate size, etc.)
involved in the dispersion mechanism and additionally the filler-filler or filler-rubber
interactions. Looking at Fig. 5, three main domains are clearly noticeable.
In the first zone, for low mixing times, the modulus Ge increases. Referring to the
TEM observations and the strain sweep measurements, this stage must be associated with
the primary dispersion of the silica in the rubber matrix. The size and the number of
agglomerates decrease. A great amount of free silica surface is created and the loosely
bound rubber content increases rapidly but without covering the whole surface yet as
shown in Fig. 10(b). This stage is achieved around the maximum of Ge.
In the second zone, the modulus Ge decreases. TEM observations and strain sweep
converge to show that agglomerates’ breakup is completed. The number of aggregates
remains constant. This step is only driven by rubber physical adsorption on the silica

FIG. 10. Schematic representations of the competition between the filler dispersion and the rubber physical
adsorption processes. (a) At the beginning of the dispersion, the system consists of large micropellets in the rub-
ber matrix; (b) the first big agglomerates are created just like clusters of individual aggregates covered by an
external layer of bound rubber; (c) the final number of individual aggregates is reached but chains of hard filler-
filler interactions remain. The bound rubber amount increases as the filler chains decrease; (d) all the aggregates
are covered by bounded rubber and interact via the immobilized rubber layer.

Redistribution subject to SOR license or copyright; see http://scitation.aip.org/content/sor/journal/jor2/info/about. Downloaded to IP:


134.153.184.170 On: Sat, 31 Jan 2015 02:04:18
420  AND F. VINCENT
J.-C. MAJESTE

surface. In this zone, we found the correlation between the bound rubber content and the
modulus illustrated by Fig. 8. The modulus Ge is given by chains of hard filler contacts
[Fig. 10(c)]. The flow’s kinematic induces rupture of the filler-filler contact and gives
opportunity to the rubber to bind on the free available surface. The kinetics of this step is
therefore linked to the mixing rate.
In the last zone, a stable state is observed. Both aggregate’s size and bound rubber
amount remain at their equilibrium value [Fig. 10(d)]. The mixing process does not mod-
ify the structure of the filler network. The weaker reinforcement is due to interaction
between the interphase of bound rubber in the vicinity of the silica particles in addition to
hydrodynamic interactions leading to the roughly the same modulus Ge whatever the
nature of silica-filler and its specific surface.
Thus, the maximum of the modulus Ge and the bump observed for G00 (c) curves are
interesting markers of the end of breakup processes during the mixing but not of a
fulfilled dispersion state. The very end of the whole dispersion act is reached after com-
plete rubber physical adsorption. Increasing the mixing time does not improve the filler
network but weakens it. By replacing filler-filler interaction with filler-rubber ones, the
reinforcement decreases along with energy dissipation. The mixing time can control the
final properties.
However, such kinetics is not compatible with industrial productiveness. The use of
coupling agents is generally the way to achieve a good dispersion and to increase work of
adhesion of rubber on the silica surface. For comparison purpose, using again the evolu-
tion of Ge as a marker of the dispersion kinetics, the dispersion of Z1165MP with a sur-
face modified by a coupling agent (TESPT) has been investigated. Figure 5 shows clearly
that, even for modified silica, all the different stages of the dispersion process remain:
first increase of Ge until a maximum, then a decrease until a same asymptotic value of
the modulus. The main and striking difference is the increase of the kinetics. Obviously,
the very rapid chemical reaction of the coupling agent on the surface of silica leads to
modification of the energy of interaction between aggregates very early in the mixing
process. As a consequence, the value of the maximum of Ge is lower compared to that
founded for untreated Z1165MP silica (even if it is clear that the surface modification
was not maximized). Interestingly, it seems that some free untreated surface of silica still
remains giving the opportunity to bound rubber to take place. Finally, whatever the
nature of the molecule covering the surface, at the end of the dispersion process, the
same modulus (i.e., same energy of interaction between aggregates) is found. Chemical
bounding is more rapid or more efficient compared to the physical one (likely due to
equilibrium between bounding and tearing of rubber chains on the silica surface) but the
scenario of the dispersion is the same.
From these results, it appears that dispersion is almost a double kinetic problem. In the
next section, we will propose a semiphenomenological model in order to capture this
competition between agglomerates’s breakup and rubber physical adsorption.

A. Modeling the reinforcement kinetics


Actually, the modulus Ge represents the elastic energy stored and released by the filler
network during one period of harmonic strain or stress applied on the filled sample. This
energy originates from the elastic interaction between fillers structured in a 3D network
[Kl€
uppel (2003)]. Macroscopic deformation of this structure causes local solicitation at
the aggregate level. The whole network’s response is the sum of each local contribution.
Therefore, the modulus Ge can be expressed as the product of the local energy stored

Redistribution subject to SOR license or copyright; see http://scitation.aip.org/content/sor/journal/jor2/info/about. Downloaded to IP:


134.153.184.170 On: Sat, 31 Jan 2015 02:04:18
A KINETIC MODEL FOR RUBBER REINFORCEMENT 421

times the number of interactions per unit volume. The simple following equation summa-
rizes this approach:

Ge ¼ N  E; (12)

where N is the number of interacting aggregates per unit volume and E is the energy
involved in the filler-filler interaction for one particle and its close neighbors.
As previously shown in the paper, both N and E are time dependent quantities through
the kinetics of agglomerates’ breakup and the kinetics of rubber physical adsorption.
Thereafter, we propose and detail the two kinetic models and compare our predictions to
the experimental variations of the modulus Ge during the mixing time.
First, the evolution of the number of interacting agglomerates (N) per unit volume as a
function of the mixing time is given. We focus on the partitions occurring in the initial
silica micropellets during the mixing process. We define the fraction of partition (n) as
the ratio of the number of agglomerates (N) and the number of aggregates (Na) constitut-
ing the amount of silica in the system per unit volume

N
n¼ : (13)
Na

Ideally, the fraction of partition n reaches unity at the end of the mixing process, e.g.,
when the number of dispersed aggregates equals (Na). For compact arrangement, this
number is related to the silica concentration and specific area according to

Na  /  A3sp q3a ; (14)

where Asp and qa are, respectively, the specific area and the specific mass of an aggregate.
Throughout the mixing process, the fraction of partition n increases according to the
amount of energy provided by the mixing device. The elementary amount of energy dW
will increase the relative number of partitions according to the following expression:

dn
/ dW; (15)
1n

dW represents the work undergone by the elastic network of aggregates and thus can be
rewritten as the work of an elastic stress

dW ¼ rdc  cdc: (16)

By integration over the whole mixing process, the previous equations finally lead to the
expression of the number of agglomerates as a function of the total strain
"  2 #
c
N ðtÞ ¼ Na 1  exp : (17)
cd

The previous equation can be rewritten introducing more tractable parameters, namely,
the mixing time and the strain rate
"  #
_c t 2
N ðtÞ ¼ Na 1  exp : (18)
cd

Redistribution subject to SOR license or copyright; see http://scitation.aip.org/content/sor/journal/jor2/info/about. Downloaded to IP:


134.153.184.170 On: Sat, 31 Jan 2015 02:04:18
422  AND F. VINCENT
J.-C. MAJESTE

The unique fitted parameter c2d stands for the total elastic energy needed to complete
the full dispersion of the Na aggregates. Actually, it represents the ability of a type of
silica to disperse. The higher its value is, the more difficult the silica will be to disperse.
The previous expression shows that the key parameter is the elastic energy stored in
the silica micropellets at rest (the sum of attractive interaction energy between aggre-
gates). By applying a flow on the filled system, the hydrodynamic stress generated by the
sheared elastomer will break the initial network of elastic interactions. Elastic energy is
transferred from silica network to elastomer matrix and complete dispersion should be
achieved when the amount of mixing energy is Oðc2d Þ the stored elastic energy. This point
will be discussed further in the paper.
Now we will focus on the evolution of the average energy of interaction between
neighboring aggregates in the elastic network of fillers. As observed in Sec. III.D,
this energy is obviously strongly related to the amount of loosely bound rubber
adsorbed on the surface of aggregates. We are in the presence of roughly two levels
of energy of interaction: strong energy when the aggregates interact via their neat
surface and weak energy when adsorbed rubber lubricates the contact between
particles.
Averaging over the whole aggregates, the energy of interaction scales proportionally
to that of the modulus Ge when the number of aggregates remain constant (i.e., complete
dispersion). Considering the scaling law for Ge removes any / contribution (actually, the
contribution of the number of particles in the network), the evolution of the energy of
interaction is then given by the following expression:
 a=3
%BdRðtÞ
EðtÞ ¼ E0  1  ; (19)
%BdRmax

where %BdR(t), %BdRmax, and a are available from direct measurements (Figs. 8 and 9).
Only E0 which represents the average energy of interaction for a network of neat aggre-
gates is unknown {it is likely that Atomic Force Microscopy (AFM) measurement onto
the neat surface of aggregates could assess these energies [Jones et al. (2003)]}. This phe-
nomenological approach shows that the covering rate of aggregates is the main parameter
governing the kinetics and intensity of the aggregates’s interactions.
Now, according to Eq. (12), by combining the contribution of the number of aggre-
gates [Eq. (18)] and the energy of interaction between aggregates [Eq. (19)], the evolu-
tion of the network’s strength, represented by Ge, can be described by the following
expression:
" !#  
2 2
0 _c  t a
ð Þ
Ge t ¼ GH ð/Þ þ Ge  1  exp  exp  kbr  c  t ;
_ (20)
c2d 3

where GH(/) is the contribution of the dispersed fillers by rubber/rubber interaction and/
or hydrodynamic interaction when the bound rubber amount is maximum.
G0e is easily obtained using Fig. 8. G0e is plainly the extrapolation of Ge for no bounded
rubber and can be defined by

G0e ¼ Na  E0 (21a)

or in a more tractable form

G0e ¼ G0  /a : (21b)

Redistribution subject to SOR license or copyright; see http://scitation.aip.org/content/sor/journal/jor2/info/about. Downloaded to IP:


134.153.184.170 On: Sat, 31 Jan 2015 02:04:18
A KINETIC MODEL FOR RUBBER REINFORCEMENT 423

Equation (20) has been successfully compared with experimental data for c_ ¼ 28.3 s1
as is shown in Fig. 5. Only cd is not available by direct measurement and has been fitted
to reach the best agreement. All the other parameters have been inferred from experimen-
tal data and they are gathered together in Table III. Equation (20) has been applied for
different filler amount from 20 up to 35 wt. %. It comes that cd is not significantly
dependent on the filler concentration at least in the range of filler amount we
investigated.
To go further, in order to check the c_ 2 dependence of the dispersion kinetics predicted
by Eq. (20), the same analysis was performed for silica Z1115MP applying two addi-
tional strain rates. Figure 11 shows the impact of the strain rate on the kinetics of disper-
sion. Decreasing the mixing rate shifts the maximum of Ge(t) toward the long times
(10 nm) while the dispersion is more shortly completed for higher strain rates. The
more noteworthy result is that for all the strain rates Eq. (20) captures very well the whole
evolution of the modulus Ge with the same set of parameters and confirms the energy
dependence of the dispersion mechanism.
In Eq. (20), two classes of parameters can be distinguished: on the one hand, kbr which
drives the kinetics of rubber adsorption onto the surface of aggregate. It must be related
to the self-diffusion coefficient and therefore, kbr surely should depend on the tempera-
ture but this effect was not investigated in this paper. On the other hand, Eq. (20) introdu-
ces filler parameters (G0e , cd , and a) which depend on the nature and characteristics of the
silica aggregates and silica solid network (from the nature of the surface to the structure
of the network). There is probably a link between the three parameters as they arise from
the structure and elastic energy of the particle network. (a) is mainly a characteristic of
the structure of the network (fractal behavior or not). G0e , which really stands for the best
specific descriptor of the solid network, comprises both fundamental contributions as
expressed by Eq. (21). Finally, by a phenomenological way, cd introduces the ability of a
silica to disperse and its values clearly agree with what was generally observed for both
Z1115MP and Z1165MP silicas [Boudimbou (2011)]. cd should comprise the same
structural and energetic contributions as G0e but the large values obtained suggest the low
efficiency of the dispersion process at the local scale. An estimation of the average strain
needed to separate two interacting aggregates can be obtained from the work of the

FIG. 11. Evolution of the values of Ge as a function of the mixing time for Z1115MP silica nanocomposites
with 30 wt. % of silica content for various shear rates: (䊏) 14.1 s1; (䊉) 28.3 s1; (D) 45.3 s1. Lines indicate
the prediction of Eq. (20) with the same coefficients (Table III) for all the shear rates.

Redistribution subject to SOR license or copyright; see http://scitation.aip.org/content/sor/journal/jor2/info/about. Downloaded to IP:


134.153.184.170 On: Sat, 31 Jan 2015 02:04:18
424  AND F. VINCENT
J.-C. MAJESTE

attractive forces involved in the cohesive structure of aggregates. The maximum elastic
Ð per unit volume that a network is able to store before rupture at c0 is then given
energy
by G0e cdc ¼ 12 G0e c20 . This energy has to be compared with the bound energy involved at
the local scale between the surfaces of aggregates. Considering the diversity of interac-
tions (from Van der Walls to hydrogen bonds), this energy lies between 10 and 100 kJ/
mol. Using the value of Ge0 from Fig. 8 and the number of moles of bounds between
aggregates per unit volume (as a first approximation, assumed to be very close to the
number of aggregates by unit volume), the maximum deformation needed to break the
particle network must lie between 1% and 5%. Hence, whatever the nature of the interac-
tions between aggregates, the strain needed to break the elastic attraction between two
aggregates is far below the values of cd .
A likely explanation can be proposed hereafter. When submitted to high shear, the
solid network is not fully dispersed in one pass. Only a fraction of the network is affected
by the local strain or stress. This fraction increases with the mixing time. Nevertheless,
the value of cd has been compared with the total strain cT undergone by the filled elasto-
mer in our mixing device. During the mixing process, the whole sample successively
passes through high shear zones (HSZ) and low shear zones. There are three HSZ for
each rotor. Consequently, any volume element has to pass three times by rule through
HSZs in which the strain chs undergone by the material is calculated from classical rela-
tions of simple shear flow chs ¼ 2L Hhs (Lhs ¼ 3 mm and Hhs ¼ 1 being, respectively, the
hs

length and the thickness of the HSZ) and estimated to 6 in our device. Therefore, the total
strain for complete dispersion of silica can be calculated using the following equation:

cT ¼ tdisp  X_  3  chs ; (22)

where X_ is the rotation speed (expressed in RPM) and tdisp is the time (in min) needed for
complete dispersion measured in Fig. 5.
For Z1165MP silica, an assessment of the dispersion time tdisp is found around 7 min
leading to a value of cT  6300 very close to cd ¼ 6000 given from Eq. (22). The same
conclusion is found for Z1115MP silica.
Therefore, cd is not the strain needed to break the physical bounds between aggregates
inside the solid network but, as discussed above, represents actually the amount of strain
needed to ensure that the whole solid particle network has been sufficiently shattered
under high shear conditions. As a consequence, even if the value of cd reflects effectively
the ability to disperse (actually a kinetic parameter), it seems very tricky to develop a
reliable link with intrinsic structural parameters as expected for G0e . Moreover, even if
the same total strain can be achieved with different mixing devices, it is likely that the ef-
ficiency of the local flow will not be the same and consequently neither will be the level
of dispersion.

V. CONCLUSION
The present study was undertaken to give a more detailed analysis of the competition
between filler-filler interaction and filler-rubber interaction during the dispersion process
and to show its impact on the final structure and reinforcement of the nanocomposite. By
coupling local observations of the dispersion and global views from linear and nonlinear
rheological measurements, a better estimation of the dispersion kinetics was obtained.
Mainly based on the measure of rheological properties and the level of the secondary pla-
teau Ge at low frequencies, it was shown that reinforcement evolves with mixing time. It

Redistribution subject to SOR license or copyright; see http://scitation.aip.org/content/sor/journal/jor2/info/about. Downloaded to IP:


134.153.184.170 On: Sat, 31 Jan 2015 02:04:18
A KINETIC MODEL FOR RUBBER REINFORCEMENT 425

can be stated that a decrease in filler networking is due to a better filler–polymer interac-
tion and consequently a replacement of strong filler-filler interactions by weak rubber-
filler ones. Thus, a direct link between the bound rubber amount (more precisely the cov-
ering rate of the surface of aggregates by physically adsorbed rubber) and the modulus
Ge was shown, which can be cast under the form of a universal power law by introducing
an effective interacting surface between filler. Finally, a kinetic model of the rubber rein-
forcement was developed on the basis of the competition between filler dispersion mech-
anism and rubber physical adsorption. Besides, it has been proposed that the dispersion
mechanism is energy dependent in great agreement with the c2 or c_ 2 dependence of the
dispersion kinetics. With only one adjusted parameter, the model is able to capture the
variation of Ge whatever the shear conditions. Hence, filler–filler and filler–rubber inter-
actions are two competitive processes that have to be accounted for in the dispersion pro-
cess even for reactive surface treatment by organic coupling agent.

Acknowledgments
This work was sponsored by MFP MICHELIN, RHODIA, and the competitive cluster
AXELERA for which grateful acknowledgement is made. The authors also acknowledge
the Centre de Microscopie Electronique Stephanois (Saint Etienne) for its support in
carrying out TEM acquisition. The authors would like to thank anonymous reviewers for
their constructive comments.

References
Aranguren, M. I., E. Mora, and C. W. Macosko, “Compounding fumed silicas into polydimethylsiloxane:
Bound rubber and final aggregate size,” J. Colloid Interface Sci. 195, 329–337 (1997).
Bohm, G. A., W. Tomaszewski, W. Cole, and T. Hogan, “Furthering the understanding of the non linear
response of filler reinforced elastomers,” Polymer 51, 2057–2068 (2010).
Boudimbou, I., “Mecanismes El ementaires De Dispersion De Charges De Silice Dans Une Matrice Elastomère,”

Ph.D. thesis (Ecole National Superieure des Mines de Paris, 2011).
Bousmina, M., A. Ait-Kadi, and J. B. Faisant, “Determination of shear rate and viscosity from batch mixer
data,” J. Rheol. 43, 415–433 (1996).
Carrot, C., J. C. Majeste, B. Olalla, and R. Fulchiron, “On the use of the model proposed by Leonov for the ex-
planation of a secondary plateau of the loss modulus in heterogeneous polymer-filler systems with agglom-
erates,” Rheol. Acta 49, 513–527 (2010).
Cassagnau, P., “Melt rheology of organoclay and fumed silica nanocomposites,” Polymer 49, 2183–2196
(2008).
Castellano, M., L. Conzatti, G. Costa, L. Falqui, A. Turturro, B. Valenti, and F. Negroni, “Surface modifica-
tion of silica: 1. Thermodynamic aspects and effect on elastomer reinforcement,” Polymer 46, 695–703
(2005).
Choi, S. S., “Filler-polymer interactions in filled styrene-butadiene rubber compounds,” Kor. Polym. J. 9, 45–50
(2001).
Choi, S. S., “Difference in bound rubber formation of silica and carbon black with styrene-butadiene rubber,”
Polym. Adv. Technol. 13, 466–474 (2002).
Choi, S. S., B. H. Park, and H. Song, “Influence of filler type and content on properties of styrene-butadiene
rubber (SBR) compound reinforced with carbon black or silica,” Polym. Adv. Technol. 15, 122–127
(2004).
Corte, L., and L. Leibler, “Analysis of polymer blend morphologies from transmission electron micrographs,”
Polymer 46, 6360–6368 (2005).
Dannenberg, E., “The effects of surface chemical interactions on the properties of filler-reinforced rubber,”
Rubber 48, 410–430 (1975).

Redistribution subject to SOR license or copyright; see http://scitation.aip.org/content/sor/journal/jor2/info/about. Downloaded to IP:


134.153.184.170 On: Sat, 31 Jan 2015 02:04:18
426  AND F. VINCENT
J.-C. MAJESTE

Domurath, J., M. Saphiannikova, G. Ausias, and G. Heinrich, “Modelling of stress and strain amplification
effects in filled polymer melts,” J. Non-Newton. Fluid Mech. 171–172, 8–16 (2012).
Dugas, V., and Y. Chevalier, “Surface hydroxylation and silane grafting on fumed and thermal silica,”
J. Colloid Interface Sci. 264, 354–361 (2003).
Frohlich, J., W. Niedermeier, and H. D. Luginsland, “The effect of filler-filler and filler-elastomer interaction on
rubber reinforcement,” Composites A 36, 449–460 (2005).
Galindo-Rosales, F. J., P. Moldenaers, and J. Vermant, “Assessment of the dispersion quality in polymer nano-
composites by rheological methods,” Macromol. Mater. Eng. 296, 331–340 (2011).
Guy, L., S. Daudey, P. Cochet, and Y. Bomal, “New insights in the dynamic properties of precipitated silica
filled rubber using a new high surface silica,” Kautsch. Gummi Kunstst. 62, 383–391 (2009).
Guy, L., Y. Bomal, and L. Ladouce-Stelandre, “Elastomers reinforcement by precipitated silicas,” Kautsch.
Gummi Kunstst. 58, 43–49 (2005).
Heinrich, G., M. Kluppel, and T. A. Vilgis, “Reinforcement of elastomers,” Curr. Opin. Solid State Mater. Sci.
6, 195–203 (2002).
Hess, W. M., “Characterization of dispersions,” Rubber Chem. Technol. 64, 386–449 (1991).
Hess, W. M., R. A. Swor, and E. J. Micek, “The influence of carbon-black, mixing, and compounding variables
on dispersion,” Rubber Chem. Technol. 57, 959–1000 (1984).
Horwatt, S. W., D. L. Feke, and I. Manas-Zloczower, “The influence of structural heterogeneities on the cohe-
sivity and breakup of agglomerates in simple shear-flow,” Powder Technol. 72, 113–119 (1992a).
Horwatt, S. W., I. Manas-Zloczower, and D. L. Feke, “Dispersion behavior of heterogeneous agglomerates at
supercritical stresses,” Chem. Eng. Sci. 47, 1849–1855 (1992b).
Jones, R., H. M. Pollock, D. Geldart, and A. Verlinden, “Inter-particle forces in cohesive powders studied by
AFM: Effects of relative humidity, particle size and wall adhesion,” Powder Technol. 132, 196–210
(2003).
Kl€uppel, M., “Elasticity of fractal networks in elastomers,” Macromol. Symp. 194, 39–45 (2003).

Leblanc, J., Rheologie Des Elastomères Et Leur Mise En Forme, edited by N. Artel (Namur, 1996).
Leblanc, J. L., “Elastomer-filler interactions and the rheology of filled rubber compounds,” J. Appl. Polym. Sci.
78, 1541–1550 (2000).
Leblanc, J. L., “Rubber-filler interactions and rheological properties in filled compounds,” Prog. Polym. Sci. 27,
627–687 (2002).
Leonov, A. I., “On the rheology of filled polymers,” J. Rheol. 34, 1039–1068 (1990).
Majeste, J. C., C. Carrot, B. Olalla, and R. Fulchiron, “Internal reorganization of agglomerates as an explanation
of energy dissipation at very low strain for heterogeneous polymer systems,” Macromol. Theory Simul. 21,
113–119 (2012).
Mathew, G., M. Y. Huh, J. M. Rhee, M. H. Lee, and C. Nah, “Improvement of properties of silica-filled styrene-
butadiene rubber composites through plasma surface modification of silica,” Polym. Adv. Technol. 15,
400–408 (2004).
Mele, P., S. Marceau, D. Brown, Y. De Puydt, and N. D. Alberola, “Reinforcement effects in fractal-structure-
filled rubber,” Polymer 43, 5577–5586 (2002).
Park, S. J., and K. S. Cho, “Filler-elastomer interactions: influence of silane coupling agent on crosslink density
and thermal stability of silica/rubber composites,” J. Colloid Interface Sci. 267, 86–91 (2003).
Rumpf, H., The Strength of Granules and Agglomerates, edited by W. Knepper (Agglomeration Interscience,
New York, 1962).
Schaefer, D. W., and R. S. Justice, “How nano are nanocomposites?,” Macromolecules 40, 8501–8517
(2007).
Schuring, D. J., and Futamura, S., “Rolling loss of pneumatic highway tires in the eighties,” Rubber Chem.
Technol. 63, 315–367 (1990).
Scurati, A., I. Manas-Zloczower, and D. L. Feke, “Influence of powder surface treatment on the dispersion
behavior of silica into polymeric materials,” Rubber Chem. Technol. 75, 725–737 (2002).
St€ockelhuber, K. W., A. S. Svistkov, A. G. Pelevin, and G. Heinrich, “Impact of filler surface modification on
large scale mechanics of styrene butadiene/silica rubber composites,” Macromolecules 44, 4366–4381
(2011).

Redistribution subject to SOR license or copyright; see http://scitation.aip.org/content/sor/journal/jor2/info/about. Downloaded to IP:


134.153.184.170 On: Sat, 31 Jan 2015 02:04:18
A KINETIC MODEL FOR RUBBER REINFORCEMENT 427

White, J. L., D. Liu, and S. H. Bumm, “Development of dispersion in rubber-particle compounds in internal and
continuous mixers,” J. Appl. Polym. Sci. 102, 3940–3943 (2006).
Wolff, S., and M. J. Wang, “Filler elastomer interactions. IV. The effect of the surface energies of fillers on elas-
tomer reinforcement,” Rubber Chem. Technol. 65, 329–342 (1992).
Wolff, S., M. J. Wang, and E. H. Tan, “Filler elastomer interactions. VII. Study on bound rubber,” Rubber
Chem. Technol. 66, 163–177 (1993).

Redistribution subject to SOR license or copyright; see http://scitation.aip.org/content/sor/journal/jor2/info/about. Downloaded to IP:


134.153.184.170 On: Sat, 31 Jan 2015 02:04:18

You might also like