The effect of bound rubber on vulcanization kinetics in silica filled silicone rubber

Abstract

The effects of bound rubber on vulcanization kinetics were studied on poly(methyl vinyl) silicone rubber (VMQ) filled with three different types of silica, which was surrounded by different amount of bound rubber due to their morphological differences. Higher specific surface area (SSA) and content of silica corresponds to the more favorable formation of bound rubber. The results of cure kinetic behavior demonstrate that the loading of silica suppresses the vulcanization rate compared to neat rubber. Under the same filler content, the higher the bound rubber content, the more obvious the inhibitory effect on vulcanization reaction. A structure model of filler network mediated by bound rubber was proposed to account for the influence of the reinforcing particles on vulcanization kinetics and rheological behavior of silicone rubber, which may be adopted to analyze the properties of nanocomposites with different fillers.

---

1. Introduction

Silicone rubber is one of the most fascinating and important industrial polymers. It is irreplaceable in many important and key areas due to its unusual flexibility and other excellent properties as thermal stability, weather resistance, transparency, hydrophobicity and biocompatibility.^1,2 Silica is one of the most effective fillers to tailor rheological behaviors and enhance mechanical properties of polysiloxane.³ Such reinforcements are related to the secondary structure of filler particles (aggregate)^4–6 and the filler–rubber interactions.^7,8 Polymer–filler interaction is mainly determined by the type and amount of adsorption sites, leading to different interfacial structure between filler particles and polymer matrix.^9,10 Previous researchers have investigated the nature of polymer–filler interactions in rubber–silica mixtures,^11–14 during which the relationship of macroscopic mechanical properties and microstructures are emphasized.^15,16 Additionally, some recent studies showed that the presence of reinforcing fillers can also influence vulcanization kinetics of rubber.^15,17,18 Majority of studies on the kinetics of rubber vulcanization were performed on the chemical effects of the filler, while it seems that chemical characteristics of the fillers are not the only affecting parameters. A systematic study of filler–polymer interactions and their role in the vulcanization of rubber is required.^15,19

The rubber–filler interaction is reflected macroscopically in the formation of bound rubber, which involves physical adsorption, chemisorption and mechanical interaction.²⁰ Here, “Bound rubber”,^1,8,20,21 defined as the polymer chains that can't be extracted from suspensions by a good solvent at room temperature, provides an important contribution on the network structure of rubber composites. The presence of bound rubber is direct evidence for filler–rubber bonds, which has been studied by many scholars.^22–24 Therefore the research on the role of bound rubber in vulcanization process might provide a new perspective to understand the structure of the rubber filler network system.

In the present work, the relationship between vulcanization kinetics and bound rubber was studied with different techniques, where the effects of silica loading and structure on the kinetics of methyl vinyl silicone rubber vulcanization were investigated systematically. Bound rubber was found to play a major role in altering the kinetics of vulcanization reactions of PVMQ. A structure model of bound rubber was proposed to explain the rubber–filler interaction.

2. Material and experiment

2.1 Materials

Silicone rubber (methyl vinyl silicone rubber with 0.13–0.20% vinyl, VMQ 110-2) was purchased from Nanjing Dongjue Silicone Group Co. Ltd. The molecular weight of the silicone rubber was about 450 000–700 000. Fumed silica (A) and precipitated silica (Z) were obtained from the Evonik-Degussa AG, Germany. Their specific surface area measured by N₂ adsorption was 200 m² g⁻¹ and 140 m² g⁻¹ respectively. Another monodispersed silica (G) with a specific surface area of 16 m² g⁻¹ was supplied by the Technical Institute of Physics and Chemistry, CAS. All silica fillers were not surface-modified by any organic molecules. As a coupling agent and a structure controlling agent, hydroxyl silicone oil (GY-209-3) was purchased from Chenguang Research Institute of Chemical Industry, China. Dicumyl peroxide (DCP, 99%, AR) as a radical initiator was purchased from Sinopharm Chemical Reagent Co. Ltd, China.

Compounds starting with SG are the ones containing G. The other compounds starting with SA and SZ are the ones containing silica A and Z, respectively. The compound with no silica is labeled as VMQ. Filler loadings ranged from 0 to 70 phr (at intervals of 10 phr). Here, 1 phr means 1 g of silica nanoparticles dispersed in 100 g of silica rubber. The ratios of GY-209-3/VMQ and DCP/VMQ are 8 : 100 and 3 : 100, respectively.

2.2 Mixing technique and sample preparation

Mixing experiments were performed on a Poly Lab System Rheocord 400P with a 75 cm³ mixing chamber Rheomix 610P (Thermo Haake) and the compounds were prepared by a two-step mixing method. A master batch of rubber and silica was prepared in the internal mixer at 105 °C for 30 min. The rotor speed was 90 rpm and the fill factor was 0.7. Then the VMQ-silica was kept at room temperature for 2 weeks in order to reach an equilibrium adsorption. Finally, the crosslinking agent DCP was added into the compound at room temperature and mixed in the internal mixer for 15 min with rotor speed of 60 rpm. The final compound was used for the test of vulcanization.

2.3 Collection and measurement of bound rubber

Bound rubber is defined as the percentage of polymer that cannot be extracted from an uncured filled rubber using a good solvent, usually, at room temperature.^21,25 Samples of 2 g of uncured rubber compound were cut into small pieces and introduced in a stainless-steel cage with 600 mesh size. The cage was closed and immersed in 200 mL toluene solvent. The solvent was renewed every 24 h. After 96 h the cage (with bound rubber and the filler) was slowly removed from solvent and dried for a few hours at room temperature, then kept to stand in a vacuum oven at 60 °C until constant weights were achieved. The obtained dry solids labeled as PA, PZ and PG are the ones containing the bound rubber and corresponding silica A, Z and G, respectively. The obtained mixture were used to determine the amounts of bound rubber by thermogravimetric analysis (TGA).^1,26 Samples were heated from 25 °C to 800 °C at a rate of 20 °C min⁻¹ under atmosphere, using a TG apparatus (PerkinElmer, USA).

According to the thermal degradation,^27,28 hydroxyl terminated VMQ chains depolymerize through Si–O bond scission in a stepwise fashion from chain ends, and the resultant volatile cyclic oligomers lose weight completely in an inert gas atmosphere.¹ Ignoring the weight loss caused by silica, the amount of silica and unextracted VMQ chains can be determined from the remained weight and lost weight respectively. The amount of adsorbed polymer (bound rubber) is calculated as the difference between the weight of the recovered solids and the test sample. The percentage of bound rubber (P_BR) refers to the weight per unit weight of pure rubber in the sample.

2.4 Characterization

The bound rubber including silica particles was dispersed in ethanol and observed by transmission electron microscopy (TEM). TEM observations were performed on an H-800 transmission electron microscope (Zeiss, Libra 200FE) with an acceleration voltage of 200 kV.

A RPA2000 rubber process analyzer (Monsanto Co. Ltd, USA) was used for rubber processing performance test. For a detailed study on the effect of filler during the curing process, RPA vulcanization scanning conditions were set to track the changes of network structure in the composites: the strain was about 0.7%, the temperature was 160 °C and the time was 30 min.

3. Results and discussion

3.1 Effect of silica concentration and type on bound rubber

To investigate the effects of silica concentration and type on bound rubber, the obtained mixtures including bound rubber and silica were analyzed by the TG apparatus. Here, the mixtures of bound rubber and corresponding silica particle were noted as PA, PZ and PG, respectively.

The amounts of bound rubber obtained from TGA curves are shown in Fig. 1. With the increases of filler content, the percentages of bound rubber P_BR in PA and PZ increase significantly, which are in line with previous reports. This can be mainly attributed to the formation of bridging chains which occupy more than one adsorption sites. As the number of filler particles increase, distance between particles is shortened, which favors the formation of bridging chains. At the same time, as filler content increases, more active sites are available for adsorption because of the growth of direct polymer–filler interfacial surface area. Among the three types of filler, the percentage of bound rubber shows a trend of PA > PZ > PG. For PG, P_BR increases weakly with the increase of filler content, especially when the filler content is less than 30 phr. With the same filler content, the bound rubber percentages P_BR in PG is far lower than that in samples PA and PZ. It was may due to low surface area and structure of G as depicted in follow. Since filler G with lower surface areas has a smaller interfacial area with the polymer per unit volume of compound at the same loading.^1,8,29–31

Fig. 1 Changes of P_BR with silica concentration and type.

3.2 TEM micrographs of bound rubber and corresponding fillers

The results of the previous section demonstrate that not only the silica contents, but also the type of silica has great influences on bound rubber. TEM was used to determine the morphology and dispersibility of the bound rubber entrapped around silica particles. Fig. 2 illustrates TEM pictures of pure fillers (a–c) and corresponding bound rubber (d–f) with a concentration of 40 phr. Fig. 2a–c correspond pure filler A, Z and G, respectively. It can be seen that silica A is typical amorphous aggregate composed of the torispherical silica primary particles with grain size of about 10 nm, which is named as ‘high structure’. The morphology of silica Z is close to A and its grain size is about 20 nm. The disorder degree of particle Z aggregates is lower than the particles A due to the different preparation methods. Different with A and Z, morphology of particles G is a typical monodisperse solid sphere with grain size of about 200 nm, which is known as ‘low structure’ with a larger particle size and smaller specific surface area.

Fig. 2 TEM images for silica particles and corresponding bound rubber: (a) A; (b) Z; (c) G; (d) PA; (e) PZ; (f) PG.

Fig. 2d–f show the micrographs of samples PA, PZ and PG, respectively. It can be seen that a large amount of bound rubber is absorbed around A aggregates in sample PA to form a homogeneous continuous structure. For the sample PZ, transmission electron microscope results show that Z aggregates are also covered by the bound rubber. But the uniformity of sample PZ is not as good as sample PA. Whilst images of PG indicate that only small amount of bound rubber exist around the monodisperse solid particles G, and a continuous structure is not formed between the particles and the rubber matrix. The contents of bound rubber in PA, PZ and PG estimated from TEM images are in line with the above results obtained from TGA. Evidently higher structure filler will more easily form a continuous structure between the filler and rubber.

3.3 Rheological curves and vulcanization characterizations

In order to study the influences of type and concentration of fillers on the vulcanizing properties of silica rubber composites, the rheological curves during curing were obtained by RPA2000 as shown in Fig. 3a–c. As can be seen from Fig. 3a and b the changing trend of the composites filled with A and Z is similar. It is clear that the values of torque (M) increases with increasing silica loading. However, it is worth noting that the trend of the composite filled with G is different from the first two composites filled with A and Z. The values of torque vary a little between filled and unfilled compounds, especially when the concentration of filler is low.

Fig. 3 Curing curves of silica rubber composites: (a) torque–time curves of SA filled compounds at various loadings; (b) torque–time curves of SZ filled compounds; (c) torque–time curves of SG filled compounds.

To obtain more specific information, some useful physical parameters which were suitable to describe the structural characteristics of the rubber composites were extracted from the rheological curves. These were the minimum torque (ML), the maximum torque (MH) and effective torque (ΔM = MH − ML). Fig. 4a and b displays the influence of reinforcements loading on the ML and ΔM. It can be seen from Fig. 4 that the values (ML and ΔM) of compounds loaded A and Z in Fig. 4a and b increase with increasing filler content, but these values of compounds loaded G in Fig. 4c almost remains constant compared with the neat rubber.

Fig. 4 Vulcanization parameters of different type fillers and contents. (a) Minimum torque (ML); (b) the difference between minimum and maximum torque (ΔM). The lines are the guides to the eyes.

Actually, the ML exhibits the torque value of the compounds in the cure curves where no vulcanization reaction has started. Therefore it can be a measure of the viscosity of unvulcanized compounds. The increment of ML value by incorporation of reinforcing particle has already been reported by others in many rubber–filler systems,^32,33 which is often attributed to the stronger interaction at the rubber–particle interface. The stronger interaction at the rubber–particle interface, the more conducive it is to form bound rubber. Thus the value of ML is closely related to bound rubber. The different change trend of G filled compounds demonstrates that the interaction between G filler and rubber is weak, the result is in accordance with a small amount of bound rubber in the SG.

Fig. 4b exhibits the role of reinforcing particles on the effective torque ΔM as a parameter that indicates the evolution of the network structure in the rubber compounds.³⁴ Similar to the ML, high structure fumed silica, i.e. SA, shows the highest influence on the enhancement of ΔM, the precipitated silica (SZ) stands in the second rank, and the monodispersed SG shows the minimum influence. The influence of reinforcing particles on the enhancement of ΔM has already been reported by other researchers.³² It is revealed that all three silica particles used in this study enhance the value of ΔM, which is in agreement with the Flory–Rehner model. Therefore, the ΔM of the filled rubber suggest that the reinforcements may change the curing system. Different reinforcements used in this study lead to different levels of effective ΔM. Meanwhile, the value of ΔM is also associated with modulus of rubber composites.^19,35 The overall modulus of silica filled rubber is determined by the synergistic effect of rubber molecular network and filler network connected with bound rubber. Since pure rubber network has a relative low modulus, ΔM is mainly originated from filler network in the case of rubber filled with silica A and Z. On the other hand, adding silica G filler leads to a weak increase of torque, indicating rubber molecular network is mainly responsible for ΔM. Then filler network connected with bound rubber is speculated to play a critical role on the evolution of network structure.

3.4 Analysis of vulcanization kinetics

Kinetic parameters for samples were obtained from RPA data. In this study, the cure kinetics of neat rubber and its compounds with all three types of silica particles were simulated using an appropriate kinetic model. Similar to classical chemical reactions, the vulcanization kinetics can be modeled by a differential equation related to time and temperature, regardless of the reaction mechanism.^16,19,36 The equation can be written as follows:

dα/dt = K(T)f(α) (1)

where α is the conversion rate, t is the time, dα/dt is the vulcanization rate, K is a kinetic constant which is a function of temperature T, and f(α) is a function corresponding to the phenomenological model.

When a curemeter is used to study the vulcanization kinetics, conversion rate α is defined as follows:¹⁶

α = (M_t − M_L)/(M_H − M_L) (2)

where M_L is the initial torque of vulcanization, M_t is the torque value corresponding to time t during vulcanization, and M_H is the torque value at the end of reaction.

An autocatalytic kinetic model is normally used to predict vulcanization reactions in rubber compounds.^37,38 In these reactions, the conversion rate is related not only to the un-reacted materials, but also to the reaction products. According to this model, f(α) is given as:^16,32,36

f(α) = α^m(1 − α)ⁿ (3)

Thus eqn (1) can be given as:

dα/dt = K(T)α^m(1 − α)ⁿ (4)

where m is the reaction order of autocatalytic reaction, and n is the reaction order of non-autocatalytic reaction.

Vulcanization rate curves of three type compounds were plotted against conversion rate, as shown in Fig. 5a–c. A non-linear regression fitting is adopted on the experimental data. Though eqn (4), and the solid lines correspond to the obtained results.

Fig. 5 Vulcanization rate as a function of conversion: (a) SA, (b) SZ and (c) SG. The lines are the result of fitting data with eqn (4).

As shown in Fig. 5, the data are well fitted, regardless of the concentration of fillers. The values of K(T), m and n for all vulcanizations are listed in Table S1.† It can be found that the K(T) values of compounds filled with SZ and SA decrease with the increase of filler content. As will be discussed latter, this effect is attributed to the formation of a filler network, mediated by the bound rubber.

However, the change trend for K(T) values of SG compounds is obviously different from SA and SZ. These values fluctuate around 5.0 min⁻¹ which is close to the value of neat rubber. The G fillers dispersed independently in rubber matrix and didn't have obvious effect on the rate of vulcanization even though 70 phr silica is added. It can be speculated that stable filler–filler and filler–rubber network structure was probably seldom formed in SG.

In order to investigate the effects of different fillers on the vulcanizing process, corresponding data was extracted from curing curves in Fig. 3. Because most vulcanizing reactions have been completed in 6 minutes, the corresponding time range is used to analyze. Curve are obtained by plotting ln(M_H − M_t) against time t (see Fig. 6). Here the slope of curve is known as the vulcanization rate at the time t. It can be seen that the curing action can be divided into four stages. In the first stage (t < 0.4 min), the reaction rate is about 0 and all curves are almost the same, which means curing reaction has not yet started at this stage, the sample is in a heat accumulation phase. In the second stage (0.4 min < t < 1.5 min), the vulcanization rates of different filling system are rapidly increasing and the rates of different filling system have the similar trends. It is speculated that sulfurization reaction occurs mainly in the rubber matrix during this process. In the third stage (1.5 min < t < 2.5 min), the rates of reaction decrease gradually. It is noted that different fillers has different inhibitory effects on the curing reaction. The more content of bound rubber has the more obvious inhibitory effect. So, it is supposed that the curing reaction may mainly occurs in the interface between the filler and the rubber matrix (i.e. bound rubber) in this stage. In the fourth stage (2.5 min < t < 6 min), all reaction rates are near 0 and most of the reaction process has completed.

Fig. 6 Vulcanization rate (plotting ln(M_H − M_t)) against time t.

4. Discussion on the bound rubber effects on filler network associated with vulcanization kinetics

The above results show that the morphology and SSA of silica filler have strong influence on bound rubber content which further affects the vulcanization kinetics. So, we speculate the effect of bound rubber on vulcanization may be attributed to the formation of three dimensional filler network, which reduces the spatial connectivity of rubber matrix. A network model of rubber compounds filled with different types reinforcing particles is proposed, as schematically illustrated in Fig. 7.

Fig. 7 Schematic representations of bound rubber effects on filler network.

Fig. 7a shows the structure of the compound SA. The rubber chains are attracted either physically or chemically to form bound rubber on the surface of the silica aggregates. Then the continuous filler network structure mediated by bound rubber is formed between filler aggregates. The continuous network structure, which divides the rubber matrix, will limit the diffusion of the crosslinking agents and reduce the availability of the reaction sites of rubber matrix to these vulcanizing agents. This effect partially blocks the chemical reaction pathway and leads to the decrease of the vulcanization rate. The correlation between the content of bound rubber and vulcanization rate is in line with our interpretation based on the network model. The higher content of bound rubber in SA corresponds to the most suppression of vulcanization rate.

Unlike the compound SA, the compound SG does not form the continuous network architectures due to the characteristics of reinforcing particles. As illustrated in Fig. 7c, most filler particles are separated points due to low content of bound rubber. The vulcanization reaction spatial pathway is nearly unaffected by the presence of filler G, which shows little impact to vulcanization rate even though 70 phr silica is added. For the compound SZ filled with Z, its content of bound rubber is between SA and SG. The continuous filler network structure mediated by bound rubber is sparser than SA, as shown in Fig. 7b. Accordingly, the compound SA shows the largest influence on the curing rate. The compound SZ and SG stand in the second and third rank, respectively, which is consistent with the content of bound rubber.

In summary the influence of bound rubber on vulcanization rate essentially stems from the blockage of the spatial pathway for vulcanization reaction due to the continuous filler network mediated by bound rubber. Higher connectivity of filler network corresponds more blockage of the reaction pathway and lower vulcanization rate. Thus the content of bound rubber can be considered as a reference for the properties of compounds filled with different fillers.

5. Conclusion

The role of three different types of silica based reinforcements on the bound rubber and cure characteristics of VMQ blend was investigated. The results confirmed that higher structure and content of silica, the more bound rubber was obtained in the VMQ/silica compounds. The more content of bound rubber, the easier the formation of a continuous filler network structure was. The continuous filler network structure mediated by bound rubber influences the kinetics and conversion of the vulcanization reactions in rubber. The samples of SA and SZ with more bound rubber received greater impact in the vulcanization reaction process. While the filler G has small influence on the vulcanization characteristic of compound SG which might not form a continuous filler network structure mediated by the bound rubber.

Finally, we proposed a possible mechanism to interpret the influence mechanism of fillers on rubber vulcanization. The mechanical properties of three type rubber compounds will be fully studied in future and the results may improve the model proposed in this paper.

Acknowledgements

This work is supported by National Natural Science Foundation of China (51473151), military pre-research project of Southwest University of Science and Technology (13zxpt07), the project of Department of Education of Sichuan Province (15ZB0115), the Open Project of State Key Laboratory Cultivation Base for Nonmetal Composites and Functional Materials (11zxfk26) and the Project 2013BB05 supported by NPL, CAEP.

References

1. Y. Yue, H. Zhang, Z. Zhang and Y. Chen, Compos. Sci. Technol., 2013, 86, 1–8.
2. R. G. Jones, W. Ando and J. Chojnowski, Silicon-containing polymers: the science and technology of their synthesis and applications, Springer Science & Business Media, 2013.
3. H. Barthel, Colloids Surf., A, 1995, 101, 217–226.
4. A. Payne and R. Whittaker, Rubber Chem. Technol., 1971, 44, 440–478.
5. N. Suzuki, M. Ito and F. Yatsuyanagi, Polymer, 2005, 46, 193–201.
6. M.-J. Wang, Rubber Chem. Technol., 1999, 72, 430–448.
7. S. S. Choi, J. Appl. Polym. Sci., 2002, 85, 385–393.
8. E. Dannenberg, Rubber Chem. Technol., 1986, 59, 512–524.
9. J. Fröhlich, W. Niedermeier and H.-D. Luginsland, Composites, Part A, 2005, 36, 449–460.
10. V. Litvinov, H. Barthel and J. Weis, Macromolecules, 2002, 35, 4356–4364.
11. J. E. Mark and B. Erman, Rubberlike elasticity: a molecular primer, Cambridge University Press, 2007.
12. A. Camenzind, T. Schweizer, M. Sztucki and S. E. Pratsinis, Polymer, 2010, 51, 1796–1804.
13. J. Diani, P. Gilormini, Y. Merckel and F. Vion-Loisel, Mech. Mater., 2013, 59, 65–72.
14. J. Otegui, G. Schwartz, S. Cerveny, J. Colmenero, J. Loichen and S. Westermann, Macromolecules, 2013, 46, 2407–2416.
15. F. Avalos, J. C. Ortiz, R. Zitzumbo, M. A. Lopez-Manchado, R. Verdejo and M. Arroyo, Eur. Polym. J., 2008, 44, 3108–3115.
16. J. Wu, W. Xing, G. Huang, H. Li, M. Tang, S. Wu and Y. Liu, Polymer, 2013, 54, 3314–3323.
17. M. A. Kader and C. Nah, Polymer, 2004, 45, 2237–2247.
18. S. Lakshminarayanan, G. A. Gelves and U. Sundararaj, J. Appl. Polym. Sci., 2012, 124, 5056–5063.
19. S. M. Hosseini and M. Razzaghi-Kashani, Polymer, 2014, 55, 6426–6434.
20. J. L. Leblanc, Prog. Polym. Sci., 2002, 27, 627–687.
21. M. I. Aranguren, E. Mora and C. W. Macosko, J. Colloid Interface Sci., 1997, 195, 329–337.
22. C. Blow, Polymer, 1973, 14, 309–323.
23. B. Meissner, Rubber Chem. Technol., 1975, 48, 810–818.
24. S. Sarkawi, W. Dierkes and J. Noordermeer, Eur. Polym. J., 2014, 54, 118–127.
25. B. P. Kapgate, C. Das, A. Das, D. Basu, S. Wiessner, U. Reuter and G. Heinrich, J. Appl. Polym. Sci., 2016, 133, 43717.
26. G. Camino, S. Lomakin and M. Lazzari, Polymer, 2001, 42, 2395–2402.
27. N. Grassie and I. Macfarlane, Eur. Polym. J., 1978, 14, 875–884.
28. T. H. Thomas and T. Kendrick, J. Polym. Sci., 1969, 7, 537–549.
29. S. Wolff, M.-J. Wang and E.-H. Tan, Rubber Chem. Technol., 1993, 66, 163–177.
30. S.-S. Choi and E. Ko, Polym. Test., 2014, 40, 170–177.
31. H. Zhou, L. Song, A. Lu, T. Jiang, F. Yu and X. Wang, RSC Adv., 2016, 6, 15155–15166.
32. M. Ahmadi and A. Shojaei, Thermochim. Acta, 2013, 566, 238–248.
33. M. J. Wang, S. X. Lu and K. Mahmud, J. Polym. Sci., Part B: Polym. Phys., 2000, 38, 1240–1249.
34. A. Marzocca, Eur. Polym. J., 2007, 43, 2682–2689.
35. G. Sui, W. Zhong, X. Yang and S. Zhao, Macromol. Mater. Eng., 2007, 292, 1020–1026.
36. M. Kader, D. Choi, S. Lee and C. Nah, Polym. Test., 2005, 24, 363–366.
37. A. Shanmugharaj, J. Bae, K. Y. Lee, W. H. Noh, S. H. Lee and S. H. Ryu, Compos. Sci. Technol., 2007, 67, 1813–1822.
38. D. Yang, A. Velamakanni, G. Bozoklu, S. Park, M. Stoller, R. D. Piner, S. Stankovich, I. Jung, D. A. Field and C. A. Ventrice, Carbon, 2009, 47, 145–152.

---

Footnote

† Electronic supplementary information (ESI) available. See DOI: 10.1039/c6ra20063j

---