New insight into the flocculation behavior of hydrophilic silica in styrene butadiene rubber composites

Abstract

The flocculation behavior of hydrophilic silica in styrene butadiene rubber composites has been carefully analyzed by rheology methodology. An evident increment of the elastic modulus (G′) can be observed over a critical temperature for unmodified composites due to a significant filler network composed of loose silica clusters, while the increment of G′ for modified composites is slight. Still, the flocculation behavior is confirmed by nonlinear dynamical strain sweeps and Atomic Force Microscopy (AFM). Thereafter, modified and unmodified silica filled composites, with varied processing temperature, are vulcanized respectively, and corresponding fatigue crack growth tests are implemented. A modified composite with a processing temperature of 130 °C possesses the smallest exponent law, b, and dc/dn at a given tearing energy (T). We deduce that fatigue crack growth changes from a local stress concentration mechanism originating from severe silica flocculation within unmodified composites to a crack deflection growth mechanism originating from moderate silica flocculation within modified composites. Based on the crack tip morphology investigations at T = 5, 10, 15, 20 kJ m⁻², it can be proposed that the crack tip morphology has a tear energy dependence, closely related to the flocculation behavior of silica.

---

1. Introduction

In recent years, silica has been established as the most important filler for car tires in reducing fuel consumption and CO₂ emissions, and the rolling resistance and wet grip performance can be improved significantly. Incorporation of silica particles within styrene butadiene rubber (SBR) is an efficient and practical method to reinforce passenger car tires.^1,2 The dispersion state of silica, silica content, etc. have been well investigated for composites, and many studies focus on the influence of the silane coupling agent content on the change of dynamic mechanical behaviors, i.e., Payne effect, Mullins effect, and network structure.^3–5 Liqun Zhang et al.⁶ study the effect of the temperature on surface modification of silica and properties of modified silica filled rubber composites (MSFC). They find that the grafting degree of modified silica at temperature of 50 °C is higher than any other temperatures. Researchers have found that the rheological properties are influenced by the particle concentration (volume fraction), surface activity of silica and the polarity of the matrix.^7,8 Wu et al.⁹ discuss a rheological study on temperature dependent microstructure changes of fumed silica in dodecane, and the remarkable increase of G′ with rising temperature is believed to be related to the restructuring of nanoparticle chain aggregates of fumed silica in gels. Fröhlich et al.¹⁰ find that the storage modulus of silica/rubber system will increase evidently after annealing them at intermediate and high temperatures. Recently, G. Heinrich et al. present one dimensional Ising model to discuss the steric interaction of filler particles in reinforced elastomers.¹¹

To ensure the safety and durability of rubber products, the research on crack propagation of rubber has a vital significance. Since tearing energy (T) as an analysis criterion of rubber crack-growth is proposed by Rivlin in 1953, the crack-growth approach has been widely applied.^12–14 Weng et al.^15,16 researched the crack-growth mechanism of natural rubber by real-time crack tip morphology monitoring method. The J-integral method is used to evaluate the resistance to crack initiation and propagation of silica/CB/NR composites.^17,18 Along with increasing silica/CB ratio, the crack initiation and propagation resistance improved. In 1980s, Gent et al. proposed micromechanics research of crack tip in elastomers.¹⁹ By the use of scanning electron microscopy, they studied the torn surfaces and the tips of propagating tears in elastomers. In their investigations, characteristic cracking morphologies were discovered. This study confirms the importance of research on the cracking morphology, which could be used to explain the microstructure changes at the crack tip.

However, these approaches can only solve part of the crack growth problem due to the complexity of rubber composites. The flocculation behavior of high content silica in styrene butadiene rubber has not been investigated enough, and the present studies focus on the effect of filler and silane coupling agent content on fatigue fracture, while the influence of silica flocculation with fixed silica content on fatigue crack growth rate is also scarcely involved. In this article, several new insights have been proposed to research the flocculation behavior of hydrophilic silica in polymer composites and fatigue fracture mechanism of vulcanized samples: (i) the continuous temperature dependent flocculation behaviors of hydrophilic silica in SBR composites have been investigated. (ii) The influence of silica flocculation on fatigue crack growth rate has received little attention ever before, and we propose the fatigue crack deflection growth mechanism for modified silica filled rubber composites with moderate silica flocculation. To the best of the authors' knowledge, the presence of silica particles in rubber composites induce crack deflection fracture mechanism has not been reported ever before.

2. Experimental

2.1 Materials

The emulsion styrene-butadiene rubber (SBR) is purchased from Lanzhou Petrochemical Company, China. The precipitated silica (Ultrasil VN3) with specific surface area ∼160 m² g⁻¹ was obtained from Degussa. The characteristic size of silica particle range from 2 to 30 microns, and the average particle size is 7 μm. 3-Octanoylthio-1-propyltriethoxy (NXT) provided by Crompton, is the silane coupling agent, which endows good compatibility and dispersibility of silica with rubber matrix. Stearic acid, zinc oxide (ZnO), N-cyclohexyl-2-benzothiazole sulfonamide (CZ), diphenyl guanidine (DPG), alcohol, acetic acid, and sulfur are all commercial grade.

2.2 Sample preparation

2.2.1 Surface modification of hydrophilic silica. The surface modification reaction of hydrophilic silica was carried out in a mixture of water/ethanol (1/3 by volume fraction). 10 g NXT was first introduced into 600 ml water/ethanol mixture and stirred for 30 min at 60 °C to obtain hydrolysis. Then about 100 g hydrophilic silica particles were added into above solution, and the reaction was realized under shearing for 4 hours at 80 °C. After that, the reaction product was filtered and washed 4 times using the water/ethanol mixture. Finally, the product was dried in vacuum oven for 12 hours at 60 °C.

2.2.2 Preparation of composites. The formulae of composites are detailed in Table 1. For easy to be remembered and concise, the designations of corresponding samples are also illustrated in it. After the rubber is plasticated in HAAKE rheometer for 1 min, 2/3 of silica (modified or unmodified) filler is incorporated into the rubber, and then the rest 1/3 of silica is incorporated in with ZnO and stearic acid 1 min later. The total mixing times are about 10 minutes at 40 rpm with varied temperatures for different samples. After cooling for a while, vulcanizators are added into the composites on open rollers at 80 °C for 5 minutes. Afterwards, all the samples are vulcanized at 151 °C under pressure of 10.0 MPa for t₉₀ measured on Curometer.

Table 1 Formulae of unmodified and modified silica filled ESBR composites^a

Code	Processing temperature in Haak	Mixing step 1					Mixing step 2
		E-SBR	Precipitated silica	Silane	Activator		Vulcanizators
		SBR1500E	Ultrasil VN3	NXT	ZnO	Stearic acid	CBS	DPG	Sulfur
		phr	phr	phr	phr	phr	phr	phr	phr
a The number of polymers, fillers, and additives are in phr (parts [in weight] per hundred rubber). b The “U” in the table denotes unmodified silica filled composites, and the “M” denotes modified silica filled composites. The numbers behind of the “U” and “M” mean the different processing temperature.
U90^b	90	100.00	65.00	—	3.00	1.00	1.50	1.40	1.40
U110	110	100.00	65.00	—	3.00	1.00	1.50	1.40	1.40
U130	130	100.00	65.00	—	3.00	1.00	1.50	1.40	1.40
U150	150	100.00	65.00	—	3.00	1.00	1.50	1.40	1.40
U170	170	100.00	65.00	—	3.00	1.00	1.50	1.40	1.40
M90	90	100.00	65.00	6.50	3.00	1.00	1.50	1.40	1.40
M110	110	100.00	65.00	6.50	3.00	1.00	1.50	1.40	1.40
M130	130	100.00	65.00	6.50	3.00	1.00	1.50	1.40	1.40
M150	150	100.00	65.00	6.50	3.00	1.00	1.50	1.40	1.40
M170	170	100.00	65.00	6.50	3.00	1.00	1.50	1.40	1.40

---

It should be pointed out that the samples without vulcanizators are prepared separately for rheological measurements to avoid the pre-vulcanization in higher testing temperatures. The hydrolysis of NXT and silica modification has been diagrammed in Fig. 1.

Fig. 1 Schematic illustration of the hydrolysis of 3-octanoylthio-1-propyltriethoxy (NXT) and modification of hydrophilic silica by NXT.

2.3 FTIR analysis

FTIR analysis is carried out on a Nicolet 6700 FTIR spectrometer at spectral range from 4000 to 500 cm⁻¹, and resolution of 4 cm⁻¹ is chosen. These modified and unmodified silica samples are pressed into pellets with KBr for FTIR transmission analysis.

2.4 Rheological and non-linear viscoelastic behavior

The rheological behavior of the samples are characterized by using rotational rheometer (AREX2000, TA instruments, USA) with parallel plates (8 mm in diameter), and the rheometer is equipped with accurate temperature control system to minimize possible error. Dynamical temperature sweeps are carried out at strain amplitude of 0.2% and a constant frequency of 6.28 rad s⁻¹. The test temperature interval is 40–140 °C, and heating rate is 2 °C min⁻¹. The dynamical strain sweeps are carried out at 100 °C, and the strain amplitude is 0.01–80%. The speed and the force of the compression process are strictly controlled by rheometer to produce an appropriate gap for rheological tests. The tests are repeated several times, and the experimental data are found to be reproducible with a relative error of less than 10%.

2.5 Scanning electron microscopy (SEM)

The morphologies of the samples after processing in HAAK are obtained on magnification factor of 20 000 by JEOL SJM-5900VL scanning electronic microscopy (SEM) with an acceleration voltage of 20 kV. The SEM-EDS sweeps have been implemented to verify the dispersing state of silica particles in the polymer matrix, and the silicon element has been identified by red dots in images. Samples are fractured in liquid nitrogen and plated with a thin layer of gold prior to SEM analysis. The samples after fatigue fracture are carefully cut off along the crack front-line. Then, rubber samples are cleaned using ultrasounds in a neutral solution for SEM imaging.

2.6 Atomic force microscope (AFM)

The micrographs of the samples are obtained in 1 μm × 1 μm area under tapping mode by Nanoscope multimode, Explore (Veeco Instruments, USA), and the phase contrast scale is 0° to 25°. Samples are annealed for 15 min at 170 °C before analysis, and the experiments are conducted at room temperature.

2.7 Fatigue crack growth properties

The fatigue crack growth performances are tested by MTS810 (MTS corporation, USA) material testing machine under 3 Hz using a displacement-controlled mode. The specimens have a gage length of 89.0 mm, width of 25.4 mm, thickness of 2.0 mm, and initial crack of length around 5.0 mm cut into the edge of the rubber specimens by a sharp razor blade. Crack-growth rates are measured under seven levels of T (J m⁻²). Then, the crack growth rate dc/dn can be determined by crack length and cycle increments. All the tests are conducted according to the GB/T13934-2006 at room temperature. During the test, the machine is periodically stopped in order to record the length of the crack measured by a digital camera.

2.8 Determination of cross-linking density

The cross-linking density is determined by equilibrium swelling. Specimens are swollen in toluene at room temperature for 72 h and then removed from the solvent, and the toluene on the specimen surface is quickly blotted off with tissue paper. The specimens are immediately weighted on an analytical balance to the tolerance of 1 mg, and then vacuum dried. The cross-linking density is determined on the basis of the Flory–Rhener:²⁰

−[ln(1 − Φ_r) + Φ_r + xΦ_r²] = V₀M_c[Φ_r^1/3 − Φ_r/2] (1)

where Φ_r is the volume fraction of polymer in the swollen mass, V₀ is the molar volume of the solvent (106.2 cm³ for toluene), M_c is cross-linking density, χ is the Flory–Huggins polymer–solvent interaction term. The value of χ for toluene–SBR is 0.41. The value of Φ_r is reached according to the method used by Bala et al.:²¹where w₁ and w₂ are the weights of the swollen and deswollen specimens, respectively, and ρ₁ and ρ₂ are the densities of the toluene and the cured rubber.

2.9 Mechanical properties

Tensile tests were performed with dumb-bell shaped samples according to the Chinese National Standard GB 528-82 on an Instron-5567 material tester at room temperature at a rate of 500 mm min⁻¹. For each sample five parallel measurements were carried out and the average value was taken.

3. Results and discussion

3.1 Characterization of surface-modified fillers

The FTIR spectra in Fig. 2 are used to characterize silica before and after treated by NXT. The peaks at 1380 cm⁻¹ can be assigned to the characteristic bending vibration peaks of C–H bonds. Furthermore, the peaks from 2850 cm⁻¹ to 2950 cm⁻¹ are ascribed to the stretch vibration peaks of C–H bonds, which confirm a coating of NXT has been formed on the surface of silica.²²

Fig. 2 FTIR spectra in the stretching region from 500 to 4000 cm⁻¹ of unmodified, modified silica, and NXT are in the transmission mode and registered with 16 scans and a resolution of 4 cm⁻¹.

3.2 Temperature dependent flocculation behavior

3.2.1 Dynamic rheological analysis. The rheological sweep can provide change of the microstructure in the system.⁹Fig. 3(a) exhibits an unusual nonlinear behavior of elastic modulus (G′) versus temperature for unmodified silica filled rubber composites (USFC). The G′ changes slightly below 60 °C and increases obviously up to 90 °C. Then, a gradual reduction of G′ appears from 90 °C to the end of heating. Analytically, the enhanced mobility of rubber matrix reducing the G′ compensates the enhanced silica flocculation contributing to the G′. As a results, the G′ changes slightly. Heating further, silica particles aggregate together to larger agglomerates through hydrogen bond and electrostatic interactions due to the abundance of silanol groups (Si–OH) on their surfaces,²³ which contributes to the increment of G′. Besides, rubber–filler interactions, namely, bound rubber, contributing to the elastic modulus, is constant almost in heating.³ Meanwhile, matrix mobility increases continuously. As a competitive result, G′ increases for more intensive fill–filler interactions than the enhanced matrix mobility. Heating from 90 °C, according to kinetic differential equations established by Georg,²⁴ as shown in eqn (3), the silica flocculation rate (dS/dt) reduces gradually with less new filler–filler networking forming due to the reduced aggregation driving force (S_R − S),where S_R is maximum aggregation values, S is instantaneous aggregation values, and k_R is aggregation coefficient. As the same time, enhanced matrix mobility reduces the G′ continuously in heating. As a consequence, the G′ declines up to the end of heating.

Fig. 3 Dynamical temperature sweeps for silica filled composites (a) U150, (b) M150. The experiments, consisting of a heating and cooling cycle, are carried out at strain amplitude of 0.2% and a constant frequency of 6.28 rad s⁻¹. The test temperature interval is 40–140 °C, and heating rate is 2 °C min⁻¹.

During cooling, G′ does not return to its initial level, but monotonically increases until the end of cooling. It should be interpreted still from the viewpoints of filler, polymer and filler–polymer interaction: (i) the mobility of rubber molecule reduce gradually during cooling, but does polymer supply extra contribution for increasing G′ in cooling such as orientation effect? Dynamic temperature sweep of neat SBR matrix has been conducted and inserted in Fig. 3(b). It can be seen that cooling curve almost covers heating curve, therefore matrix itself has not supplied extra elastic modulus contribution in cooling. (ii) Filler–rubber interactions remain constant for it is thermally activated.³ (iii) The irreversible filler flocculation completed in heating takes effect still in cooling. Therefore the irreversible silica flocculation produced in heating should be the predominant and unique reason for the extra increment of G′ (ΔG₄₀) in cooling.

In previous reports, researchers mainly focus on the one fold flocculation behavior of filler in low molecular matrix (such as dodecane),^9,25,26 and omit the important filler–polymer interactions, which have important significant role for the viscoelastic behavior of silica filled system. Fröhlich et al.¹⁰ have found that the storage modulus of silica/rubber system will increase evidently after annealing them at intermediate and high temperatures, while it cannot reveal the variation of elastic modulus (G′) versus variable temperature. In addition, in many investigating reports concerning silica filled rubber composites, researchers mainly focus on the strain and frequency dependent viscoelastic behavior,^2,27 while the temperature dependent viscoelastic behavior are scarcely concerned.

Hence, in this manuscript, the complicated filler–polymer, filler–filler, and polymer matrix interactions exist at the same time in the silica filled polymer system, and the temperature dependent dynamic temperature sweep have revealed more interesting variation of elastic modulus in heating and cooling, which reflect the change of microstructure in the polymer system. Therefore, we consider that unmodified silica filled rubber composites exhibits an unusual nonlinear behavior compared with previous reports.

Fig. 3(b) indicates the diagrammatic presentation of modified composites. The falling of elastic modulus is significant from the beginning of heating. Similar analysis, enhanced polymer mobility diminishes the G′, and the preferable filler–polymer compatibility between modified silica and matrix leads to less filler–filler interaction, contributing to the reduction of G′ also, therefore, the elastic modulus reduces. Further heating from 90 °C to 140 °C, curve gradient becomes flat owing to the hydrogen bond interactions coming from the residual silanol groups (Si–OH) on modified silica surfaces, which form a few filler–filler network in hindering the reduction trend of G′. The cooling process will not be discussed again because it has been interpreted in Fig. 3(a). Besides, there exists a critical temperature (T_cri) in both MSFC and USFC, which results in the definite transition of G′ due to filler–filler network, and the ΔG₄₀ of MSFC is apparently smaller than that of USFC for its slight flocculation contribution to elastic modulus (G′) in heating.

To exclude the degradation possibility of neat matrix, which perhaps has considerable effect in testing, dynamic temperature sweep is conducted and inserted in Fig. 3(b) with same test setting. The result indicates that the cooling curve of neat matrix almost accords with the curve in heating, indicating no degradation effect during temperature sweeps.

3.2.2 Flocculation contribution in rubber composites. As spherical filler particles, the enhancement of silica in rubber matrix can be expressed through the Smallwood–Guth–Gold equation shown in eqn (4).²⁸where G′ is the elastic modulus of the composite, G′₀ is the elastic modulus of the neat elastomer, ∅ is the filler volume fraction and the quadratic term accounts for the mutual disturbance of filler particles. On the one hand, it can be seen in Fig. 3 that the G′ of U150 and M150 increases markedly compared with the neat SBR due to the filler addition, which is consistent with eqn (4). On the other hand, filler volume fraction of M150 and U150 looks like the same, but the G′ of U150 is higher enough than that of M150 after the temperature sweep circle. Actually, ∅ should be revised by ∅′ (revised filler volume fraction) for unmodified composites, because evident filler networks forming and occluded rubber shielded by silica agglomerates both amplify the filler volume fraction in unmodified composites.¹⁰

In brief, the increment of G′ for USFC in heating should be mainly related to the reconstruction of silica aggregates in composites, which form loose clusters and interact with each other and form a filler network with a significant contribution to the G′ of the rubber material. The filler network is constituted of closely packed fractal flocculation, and this reconstruction of silica chains is irreversible because the closely packed aggregates will possess a stable structure with lower Gibbs free energy.²⁹

3.3 Nonlinear viscoelastic behavior

To study the flocculation behavior directly, the dynamical strain sweeps have been conducted and just parts of results are shown in Fig. 4 for clarity. Fig. 4 presents the Payne effect plots observed as softening of the rubber. The USFC series have a larger magnitude at strain smaller than 5% compared with MSFC series. The modulus magnitude increases when the processing temperature rises for USFC series due to its enhanced filler networking in higher processing temperature,¹⁰ which is consistent with Fig. 3(a). The modulus magnitude decreases when the processing temperature rises for MSFC series, which is similar with Fig. 3(b). Analytically, by using modifier (NXT), more rubber chains participate in the bound rubber layer at the vicinity of filler surface and the fixed rubber chains do not contribute to molecular slippage,³ and the preferable filler–rubber compatibility leads to less silica flocculation, accompanying with the reduction of modulus magnitude. The results confirm, that the temperature dependence of flocculation behavior of silica in USFC and MSFC is quite different, denoting the distinct filler–rubber interaction.

Fig. 4 Plots of storage modulus versus strain for USFC and MSFC, and the processing temperature of samples are 110 °C, 130 °C, and 150 °C, respectively. The experiments are carried out at strain interval of 0.01–50.0%, and a constant frequency of 6.28 rad s⁻¹. The test temperature is 100 °C.

3.4. Fatigue fracture

3.4.1 Fatigue crack growth. The fatigue fracture performance can provide direct evidence for microcosmic cross-linking and filler–polymer interaction,¹⁷ so the fatigue crack growth tests are conducted. Griffith proposed that crack growth is due to the conversion of a structure's stored potential energy to surface energy associated with new crack surfaces.³⁰ The peak energy release rate (T), namely, the tearing energy, which drives the crack at a given rate, is expressed aswhere U is the total elastic strain energy stored in the component, and A is area of one fracture surface of the crack. Lake and Lindley^31,32 have proved that the crack-growth rate can be divided into four stages based on various tearing energies. In regime 1, T is lower than the threshold tearing energy (T₀), and the expansion length per cycle (dc/dn) is usually regarded as 0. In regime 4, dc/dn becomes very large when the T is higher than the critical tearing energy (T_c). Between the above two values, in regime 2, dc/dn proceeds linearly with increasing T as shown in eqn (6) and proceeds exponentially with increasing T in regime 3 as shown in eqn (7).where c is the cracking length, n is the number of loading cycles, T₀, T_c, T_a, A, B, b are material constants, T_a is a transition value between T₀ and T_c. A, B, b should be determined experimentally.³³ In the simple tension specimen with the single edge cut, the peak energy release rate T depends on the gauge section strain energy density W, the length of the crack c, and a strain-dependent parameter k.³⁴

T = 2kW_c (8)

Lake³⁵ proposed an approximate relation for k

Thus, the peak energy release rate of a rubber sample containing a crack under uniaxial stretching could be determined as follows:

T = 2πλ^−1/2W_c (10)

where λ is the extension ratio, and W is the strain energy density, is calculated from the stress–strain curve to eliminate the Mullins effect as shown in diagrammatic sketch (see Fig. 5).

Fig. 5 Calculation of strain energy density via the cyclic stress–strain curve of silica filled composites.

The tearing energies are calculated from eqn (10) at different crack lengths. Then the crack growth rate, dc/dn, was plotted against the maximum energy release rate (see Fig. 6). All of which obey a power-law dependency in eqn (7). Then the fatigue parameters have been determined and listed in Table 2. To observe the difference of MSFC and USFC clearly, the crack growth rates at the fixed tearing energy (T) of 12 000 J m⁻² are displayed in Fig. 7. The results show that MSFC possess lower values of crack growth rate at a given tearing energy than that of USFC always, namely, the stronger resistance to crack growth. Still, U150 and U170 encounter the worst crack growth rate because of the largest value of the exponent, b, and the maximum dc/dn at a given tearing energy.³³

Fig. 6 The fatigue crack of MSFC and USFC as a function of the crack-growth rate (dc/dn) versus different tearing energy (T) with varied processing temperature respectively. The tests are carried out at room temperature with a frequency of 3 Hz.

Table 2 Fatigue parameters of MSFC and USFC with varied processing temperature

Composite	B	b	Composite	B	b
U90	1.03 × 10^−2	0.71	M90	2.15 × 10^−6	1.45
U110	3.84 × 10^−4	0.82	M110	1.17 × 10^−6	1.50
U130	4.76 × 10^−4	0.79	M130	1.34 × 10^−4	0.96
U150	7.63 × 10^−6	1.54	M150	6.35 × 10^−6	1.31
U170	1.62 × 10^−5	1.47	M170	3.79 × 10^−5	1.14

---

Fig. 7 Crack-growth rate for vulcanized MSFC and USFC at the fixed tearing energy (T) of 12 000 J m⁻².

Analytically, evident silica agglomerates formed in USFC leads to serious stress concentration interfaces and defects between silica agglomerates and rubber matrix, which deteriorates the mechanical properties.³⁶ That means intrinsic flaws existing in composites dominate the crack growth of USFC. Compared with USFC, more covalent bonds between filler and rubber in MSFC increases the energy necessary to break the crack tip.¹⁶ To confirm this analysis, the cross-linking density experiments determined by toluene equilibrium swelling are conducted (see Table 3). It can be seen that cross-linking density of MSFC are larger than that of USFC always, reflecting the superior cross-linking networks in MSFC.

Table 3 Cross-linking density of USFC and MSFC with varied processing temperature

Composites	Cross-linking density × 10^5 (mol cm^−3)	Composites	Cross-linking density × 10^5 (mol cm^−3)
U90	3.69	M90	7.39
U110	3.49	M110	8.18
U130	3.69	M130	8.32
U150	3.80	M150	8.48
U170	3.35	M170	8.90

---

Meanwhile, it is interesting that, M130 possesses the lowest value of the exponent, b, and the smallest crack growth rate at a given tearing energy, denoting the strongest resistance to crack growth. Based on Table 3, the cross-linking density of MSFC increase with the elevated processing temperature, which reveal the enhanced covalent bonds between filler and rubber, while the resistance to crack growth rate do not improve linearly and reveal a “V” shaped transition under the processing temperature of 130 °C.

To research this interesting phenomenon further, the atomic force microscope (AFM) analysis has been conducted and parts of samples have been shown in Fig. 8 for clarity. First of all, Fig. 8(a)–(c) show that the lighter area difference in micrographs is similar, reflecting the small difference in filler–polymer interactions owing to similar rubber response to the AFM tapping deformation, because there appears similar hardness (lighter area) in the intermediate region from 10° to 20° phase difference.³⁷ Compared with USFC shown in Fig. 8(a)–(c) and (a1)–(c1) reveal the significant rubber response to the AFM tapping deformation. That is to say, the filler–polymer interactions in MSFC is preferable than that of USFC at each same processing temperature, which is attributed to the better modification effect of NXT in MSFC, and the modification mechanism can be seen clearly in Fig. 1. More important, as shown in Fig. 8(a1)–(c1), the higher processing temperature in HAAK can enhance the filler–polymer interactions for modified composites, namely, less filler–filler networks forming, because the filler–filler and filler–polymer interaction are competitive.¹⁰ That means M130 possesses stronger silica flocculation than M150 and M170. Meantime, processing temperature beyond 110 °C is necessary to obtain fundamental filler–polymer interaction which contributes to mechanical properties (see Table 4). The tensile strength at break is increased from 24.8 MPa for M90 to 29.7 MPa for M110, while, higher processing temperature more than 130 °C will not provide more power in improving the mechanical properties further, which is consistent with the fatigue crack growth results shown in Fig. 7. According to these results, we propose that the silica flocculation gives contributions to crack growth resistance for modified composites.

Fig. 8 AFM images of 1 μm × 1 μm area for parts of USFC (a) U90, (b) U130, (C) U170 and MSFC (a1) M90, (b1) M130, (C1) M170. The micrographs are obtained in tapping mode and are phase contrast images, with harder regions appearing lighter. The phase contrast scale (0–25°) which is applied to both micrographs. Samples are annealed for 15 min at 170 °C before analysis.

Table 4 Mechanical properties for MSFC with varied processing temperatures

Properties	M90	M110	M130	M150	M170
Modulus, 100% (MPa)	3.41	3.38	2.97	3.34	3.81
Tensile strength (MPa)	24.8	29.7	29.6	28.4	29.9
Elongation at break (%)	653	732	743	694	666

---

According to the theories reported by Persson et al.,³⁸ the energy dissipation at a crack is consistent with the tearing energy. There are two contributions in rubber-like materials. The first contribution is associated with the energy necessary to break the bonds at the crack tip, which is related with cross-linking network. The second contribution is attributed to the viscoelastic dissipation in the linear viscoelastic region in front of the crack tip. Strong filler–rubber interactions restrain the crack growth resulting from the need of more energy to break the bonds at the crack tip, as shown in Fig. 9.¹³ However, these two energy dissipation contributions cannot interpret why M130 possesses preferable resistance to crack growth than others. On this basis, we propose the crack tip deflection mechanism model by a schematic diagram shown in Fig. 10.

Fig. 9 Two contributions of tearing energy. The first, crack propagation energy (T) is a product of a term T₀ derived from the bond breaking at the crack tip, and the second is derived from the bulk viscoelastic energy dissipation in front of the tip.

Fig. 10 Schematic diagram of crack tip deflection mechanism of filler flocculation in retarding fatigue crack growth rate on fatigue fracture surface.

The crack deflection theory was described by Faber and Evans.³⁹ It is assumed that a crack can be deflected at an obstacle and that it is forced to move out of the initial propagation plane by tilting and twisting. We propose that the modified composites with moderate silica flocculation possesses the longer crack growth routes (L1 + L2 + L3 + … + Ln − 1 + Ln > L). It can be interpreted that the crack growth direction will deflect when the crack tip collides with rigid silica flocculation, and the silica flocculation should be larger to be collided by crack tip to obtain the higher deflection probability. Actually, it has been reported that the presence of nanoparticles in epoxy induce the crack deflection fracture mechanism, which improves the stiffness, strength and toughness of epoxy.⁴⁰ Typical crack-side morphologies of U130 and M130 (T = 20 kJ m⁻²) are captured and shown in Fig. 11. It can be seen that the crack edge of M130 is scraggly obviously than that of U130, revealing the distinct microstructure between unmodified and modified samples. More importantly, Fig. 11(b) indicates that M130 subjects to more fracture routes or surfaces, which means that it requires more energy necessary to break the bonds at the crack tip. It can be regarded as one of the evidence for proposed crack tip deflection mechanism. In a word, the smooth flaw edge growth of U130 reflects the stress concentration mechanism originating from severe silica flocculation and the scraggly flaw edge growth of M130 reflects the crack deflection growth mechanism originating from moderate silica flocculation. Therefore the moderate silica flocculation may be beneficial to crack growth resistance for modified composites.

Fig. 11 Crack-side morphologies of (a) U130 and (b) M130 at T = 20 000 J m⁻².

3.4.2 Crack tip morphology study. To deeply understand the difference of crack propagation mechanism between MSFC and USFC, the fracture surfaces morphologies with varied tearing energy (T = 5, 10, 15, 20 kJ m⁻²) have been investigate by SEM (see Fig. 12). The initial crack front-line are marked out by the red dashed circles. Evidently, the fracture surface morphology of U130 and M130 are quite different, reflecting the different fracture fatigue mechanisms. As tear energy increases, the fracture surface becomes smoother no matter modified or not, indicating the transition of the fracture surface morphology. Compared with U130, the fracture surfaces of MSFC exhibit apparently layered structures and crease, revealing the stronger resistance to crack growth. The fracture surface of U130 at T = 10 kJ m⁻² presents loose avalanche-like structure, which can be attributed to a great deal of interfaces defects between filler and rubber, originating from excessive silica flocculation.

Fig. 12 Fracture surface morphology of U130 (a), (b), (c) tested at T = 5, 10, 15 kJ m⁻², and of M130 (d), (e), (f) tested at T = 10, 15, 20 kJ m⁻². The arrows indicate the crack growth directions.

SEM images of fracture surface for M130 is shown in Fig. 13 to deeply understand the crack tip deflection mechanism. Take a closer look at the fracture surface; we can see a number of crazes exist on the fracture surface (see the regions marked by the white circles), which is the results of coalescence of the voids.¹⁵ Still, there are many stairs-like structures on the fracture surface (see the areas pointed by the arrows). The stairs-like structures of the fracture surface may be attributed to the deflection in crack growth direction. The crack tip deflection phenomenon points out two features: (i) by tilting and twisting, crack tip deflection extends the crack growth route, which is almost deflective to the cracking direction; (ii) more tear energy necessary is dissipated to the more incremental crack route and more extending fracture surface areas. Thus, crack growth rate slows down. Actually, from the viewpoint of energy dissipation, the creation of more new surfaces reflects higher energy necessary to create new propagating cracks.³⁸ In short, the moderate silica flocculation in modified silica filled rubber composites benefits the resistance to fatigue crack growth.

Fig. 13 SEM image of the fracture surface of M130 tested at T = 15 kJ m². Features indicate (A) an apparent agglomerates, (B) stretching banding, and (C) stairs-like crack deflection structures. Crazes regions marked by the white circles. The black arrows indicate the cracking direction.

3.5 Microscopic morphology

Scanning electron microscopy (SEM) can give direct observation about the microscopic morphology of filler particles within rubber matrix. Thus, microscopic observations of samples, after mixing in HAAK, are investigated and displayed in Fig. 14. First of all, the dispersing state of silica particles in MSFC is preferable always than that of USFC at each same processing temperature, which is attributed to the better modification effect of NXT in MSFC, and the modification mechanism can be seen clearly in Fig. 1. More importantly, it can be seen that agglomerates are pronounced with elevated temperature for USFC, agreeing with the results in rheological sweeps. Meantime, the compatibility between filler and rubber matrix improves with elevated processing temperature for MSFC, and silica particles are homogeneously dispersed in rubber matrix without distinct silica agglomerates generating, which agrees with the nonlinear viscoelastic strain sweeps shown in Fig. 4 and mechanical properties shown in Table 4, etc. Hence, the silica flocculation created during processing in HAAK can be remained partly after milling and subsequent vulcanization, so the processing temperature cannot be neglected for its significant effects on resistance to fatigue crack growth and other performances.

Fig. 14 The microscopic morphologies of USFC and MSFC after processing in HAAK with varied processing temperature of (a) 90 °C, (b) 110 °C, (c) 130 °C, (d) 150 °C, (e) 170 °C, (a1) 90 °C, (b1) 110 °C, (c1) 130 °C, (d1) 150 °C, (e1) 170 °C, respectively.

For verifying the dispersing state of the silica particles in the polymer matrix evidently, further SEM-EDS element scans have been conducted (see Fig. 15), and the silicon element has been identified by red dots in images. According to the silica particles distribution results, it can be seen that, the silica flocculation behavior is not significant in lower processing temperature no matter modified or not, which is consistent with SEM images shown in Fig. 14(a)–(c) and (a1)–(c1). The dispersing state of the silica particles in U170 deteriorates significantly owing to the distinct flocculation behavior, revealing the consistency with SEM images shown in Fig. 14(e), and the homogeneity of the silica particles in M170 improves significantly owing to the preferable filler–polymer interactions, revealing the consistency with SEM images shown in Fig. 14(e1). Further, it also can be observed that the dispersing state of the silica particles in MSFC is preferable than that of USFC at each same processing temperature, which is attributed to the better modification effect of NXT in MSFC, which is consistent with the comparison between Fig. 14(a)–(e) and (a1)–(e1).

Fig. 15 The silica particles distribution results for parts of USFC and MSFC after processing in HAAK with varied processing temperature, (a) U90, (b) U130, (C) U170, (a1) M90, (b1) M130, (C1) M170. The red dots in images represent silicon element by SEM-EDS mapping scans with magnification factor of 20 000.

4. Conclusion

The flocculation behavior of hydrophilic silica in styrene butadiene rubber composites and its effect on fatigue crack growth have been investigated carefully. An evident increment of the elastic modulus (G′) can be observed over a critical temperature for unmodified composites, while the increment of G′ for modified composites is slight. The flocculation behavior in heating should be mainly related to the reconstruction of silica aggregates. The filler network is constituted of closely packed fractal flocculation, and this reconstruction of silica chains is irreversible because the closely packed aggregates will possess a stable structure with lower Gibbs free energy. Occluded rubber shielded by silica agglomerates amplify the filler volume fraction and contributes to the increment of G′ also. Fatigue crack growth changes from local stress concentration mechanism originating from severe silica flocculation within USFC to crack deflection growth mechanism originating from moderate silica flocculation within MSFC. Based on the crack tip morphology investigations, it can be proposed that the crack tip morphology has tear energy dependence, relating with flocculation behavior of silica closely.

Acknowledgements

This paper is financially supported by State Key Laboratory of Polymer Materials Engineering (Grant No. sklpme2014-2-08), the National Science of China (51421061), Sichuan Youth Science and Technology Foundation (2015JQ0012). The instructive comments and suggestions from the reviewers are also greatly appreciated. The authors also appreciate Yan Liu for her great help with the fatigue testing.

Notes and references

1. A. Le Gal, L. Guy, G. Orange, Y. Bomal and M. Klüppel, Wear, 2008, 264, 606–615.
2. K. Stockelhuber, A. Svistkov, A. Pelevin and G. Heinrich, Macromolecules, 2011, 44, 4366–4381.
3. C. Gauthier, E. Reynaud, R. Vassoille and L. Ladouce-Stelandre, Polymer, 2004, 45, 2761–2771.
4. G. Seo, S. M. Park, K. Ha, K. T. Choi, C. K. Hong and S. Kaang, J. Mater. Sci., 2010, 45, 1897–1903.
5. R. Scotti, L. Wahba, M. Crippa, M. D'Arienzo, R. Donetti, N. Santo and F. Morazzoni, Soft Matter, 2012, 8, 2131–2143.
6. Y. Li, B. Han, S. Wen, Y. Lu, H. Yang, L. Zhang and L. Liu, Composites, Part A, 2014, 62, 52–59.
7. S. R. Raghavan, H. Walls and S. A. Khan, Langmuir, 2000, 16, 7920–7930.
8. F. Yziquel, P. Carreau and P. Tanguy, Rheol. Acta, 1999, 38, 14–25.
9. X.-J. Wu, Y. Wang, W. Yang, B.-H. Xie, M.-B. Yang and W. Dan, Soft Matter, 2012, 8, 10457–10463.
10. J. Fröhlich, W. Niedermeier and H.-D. Luginsland, Composites, Part A, 2005, 36, 449–460.
11. G. Heinrich and T. A. Vilgis, eXPRESS Polym. Lett., 2015, 9, 291–299.
12. R. Rivlin and A. G. Thomas, J. Polym. Sci., 1953, 10, 291–318.
13. Y. Nie, L. Qu, G. Huang, B. Wang, G. Weng and J. Wu, Polym. Adv. Technol., 2012, 23, 85–91.
14. G. J. Lake and A. G. Thomas, Proc. R. Soc. London, 1967, 300, 108–119.
15. G. Weng, H. Yao, A. Chang, K. Fu, Y. Liu and Z. Chen, RSC Adv., 2014, 4, 43942–43950.
16. H. Yao, G. Weng, Y. Liu, K. Fu, A. Chang and Z. R. Chen, J. Appl. Polym. Sci., 2015, 132.
17. B. Dong, C. Liu and Y.-P. Wu, Polym. Test., 2014, 38, 40–45.
18. G. Song, K. Chandrashekhara, W. Breig, D. Klein and L. Oliver, Fatigue Fract. Eng. Mater. Struct., 2005, 28, 847–860.
19. A. Gent and C. Pulford, J. Mater. Sci., 1984, 19, 3612–3619.
20. D. A. Brant, W. G. Miller and P. J. Flory, J. Mol. Biol., 1967, 23, 47–65.
21. P. Bala, B. Samantaray, S. Srivastava and G. Nando, J. Appl. Polym. Sci., 2004, 92, 3583–3592.
22. G. Mathew, M. Y. Huh, J. Rhee, M. H. Lee and C. Nah, Polym. Adv. Technol., 2004, 15, 400–408.
23. V. Gun'ko, I. Mironyuk, V. Zarko, V. Turov, E. Voronin, E. Pakhlov, E. Goncharuk, R. Leboda, J. Skubiszewska-Ziȩba and W. Janusz, J. Colloid Interface Sci., 2001, 242, 90–103.
24. G. G. Böhm and M. N. Nguyen, J. Appl. Polym. Sci., 1995, 55, 1041–1050.
25. J. Nordström, A. Matic, J. Sun, M. Forsyth and D. R. MacFarlane, Soft Matter, 2010, 6, 2293–2299.
26. A. Kelarakis, M. J. Krysmann and E. P. Giannelis, Polymer, 2014, 55, 6278–6281.
27. J. Ramier, C. Gauthier, L. Chazeau, L. Stelandre and L. Guy, J. Polym. Sci., Polym. Phys. Ed., 2007, 45, 286–298.
28. S. Agnelli, V. Cipolletti, S. Musto, M. Coombs, L. Conzatti, S. Pandini, T. Riccò and M. Galimberti, eXPRESS Polym. Lett., 2014, 8, 436–449.
29. A. P. Weber and S. K. Friedlander, J. Aerosol Sci., 1997, 28, 179–192.
30. A. A. Griffith, Philos. Trans. R. Soc., A, 1921, 163–198.
31. J. Halpin, Rubber Chem. Technol., 1965, 38, 1007–1038.
32. W. Mars and A. Fatemi, Fatigue Fract. Eng. Mater. Struct., 2003, 26, 779–789.
33. G. Lake and P. Lindley, J. Appl. Polym. Sci., 1965, 9, 1233–1251.
34. W. Mars and A. Fatemi, Int. J. Fatigue, 2002, 24, 949–961.
35. G. Lake, Rubber Chem. Technol., 1995, 68, 435–460.
36. P. Mélé, S. Marceau, D. Brown, Y. de Puydt and N. D. Albérola, Polymer, 2002, 43, 5577–5586.
37. C. G. Robertson, C. Lin, R. Bogoslovov, M. Rackaitis, P. Sadhukhan, J. Quinn and C. Roland, Rubber Chem. Technol., 2011, 84, 507–519.
38. B. Persson, O. Albohr, G. Heinrich and H. Ueba, J. Phys.: Condens. Matter, 2005, 17, R1071.
39. K. Faber and A. Evans, Acta Metall., 1983, 31, 565–576.
40. B. Wetzel, P. Rosso, F. Haupert and K. Friedrich, Eng. Fract. Mech., 2006, 73, 2375–2398.

---