Epoxy/Silica Nanocomposites: Nanoparticle-Induced Cure Kinetics and Microstructure


  • Patrick Rosso,

    Corresponding author
    1. Centre for Advanced Materials Technology, School of Aerospace, Mechanical and Mechatronic Engineering, The University of Sydney, NSW 2006, Australia
    • Centre for Advanced Materials Technology, School of Aerospace, Mechanical and Mechatronic Engineering, The University of Sydney, NSW 2006, Australia. Fax: +61 2 9351 7060
    Search for more papers by this author
  • Lin Ye

    1. Centre for Advanced Materials Technology, School of Aerospace, Mechanical and Mechatronic Engineering, The University of Sydney, NSW 2006, Australia
    Search for more papers by this author


This study reveals the influence of silica nanoparticles on the cure reactions of a diglycidyl ether of bisphenol A epoxy resin. As soon as the silica nanoparticles are added to the neat resin (1, 3, and 5 vol.-%), the total degree of conversion increases with an increasing amount of nanoparticles, and the cure reaction shows a more complex autocatalytic behaviour, which can not be described by a traditional kinetic model. Results from subsequent thermo-mechanical analyses confirm an alteration in the microstructure attributable only to the presence of the nanoparticles in the curing stage. An amino-rich interphase around the reactive treated particles is formed, which shifts the resin/hardener ratio, and benefits the homopolymerization of the epoxy and leads to a more highly crosslinked epoxy network. At the same time, the nanophase consists of a core-shell structure with the rigid particle inside and a rubber-like shell because of the excess hardener in this region.

original image

TEM image of two neighboring silica nanoparticles in the epoxy matrix showing a 2–3 nm altered interphase region.


Over the past decade there have been intensive studies of the development of polymer nanocomposites with nanometer-sized additives, and these materials have demonstrated their unique properties with improvements in stiffness, toughness, and tribological behaviour.1–3 Epoxy nanocomposites have been extensively investigated for use as adhesives, functional coatings, and packaging materials for electronic devices, and in particular, matrices of advanced fibre composites with improved performance. However, there is still a lack of knowledge about the mechanisms responsible for the dramatic changes in properties. The current procedures for the development of these materials are mostly based on ‘trial and error’ approaches, adopting the methods commonly used for micrometer-sized particulate composites. Fundamental understanding is very limited in the field and often leads to the conclusion that approaches commonly used for particulate composites are not suitable when a nanometric phase is involved.4

There have been significant efforts over the past decade to develop modified epoxies with various nano-additives including carbon nanotubes/fibres, exfoliated nanoclays, and organic/inorganic nanoparticles. However, few systems have achieved uniform dispersions of nanoparticles without agglomeration. Recently, a novel nanosilica system was developed by a German company (Nanopox F, nanoresins AG), which can achieve a uniform dispersion of nanoparticles of a fraction up to 40–50 wt.-% in epoxies.5 Nanopox F products are colloidal silica sols in a resin matrix with surface-modified, spherically shaped silica nanoparticles that have an extremely narrow particle size distribution. These nanospheres (15–50 nm in diameter) are dispersed agglomerate-free in the resin matrix as it can be seen from a transmission electron microscopy (TEM) picture in ref.1. The nanoparticles are chemically synthesized from aqueous sodium silicate solution. The OH groups on the surface of the silica particles are reacted with organosilanes, which are selective with regard to the bisphenol-A epoxy in which the particles are incorporated. Thus a hydrophobic organic surface coating is formed.

The thermal and mechanical properties of epoxy resins are highly dependent on the crosslinked three-dimensional microstructure formed during the curing process. Different factors may influence the curing process, such as the temperature profile, time, epoxy/hardener ratio, and additives.6,7 DSC is a technique commonly used to characterize the cure kinetics based on heat evolution and activation energies. Conversion studies resulting from isothermal and dynamic DSC experiments can quantitatively define the formation of the epoxy network. The cure kinetics can provide information about the autocatalytic reactions between epoxy and amine, and thus facilitate capture of separate events (e.g., side reactions) as a result of the addition of the nanoparticles and their interaction with the epoxy molecules. The present study confirms that the presence of nanosilica particles in the bulk epoxy induces a change in curing behaviour, and a core-shell structure is formed around the nanoparticles.

Experimental Part


The raw materials used in this project included nanosilica of 15–50 nm in diameter (Nanopox F400, nanoresins AG, Germany), Araldite F epoxy (AF, Ciba-Geigy, Australia), and piperidine (Sigma Aldrich), a cycloaliphatic secondary amine, which was one of the first amines to be used with the epoxy resins on a commercially significant scale. The diglycidyl ether of bisphenol A (DGEBA) it is customarily used at 5 phr. The neat epoxy (AF) and nanoparticles were mixed using a simple laboratory stirrer for 15 min at 50 °C. The nanosilica content in the epoxy was 1, 3, and 5 vol.-% in composition, and referred to as AF1, AF3, and AF5, respectively. The mixtures were degassed in a vacuum oven (about −100 kPa) at 100 °C for about 1 h. The vacuum was then removed and piperidine was added while stirring slowly. The resin was then cast into preheated aluminium moulds and cured at 120 °C for 16–20 h.


In kinetic studies using DSC, the degree of cure (α, also conversion) is defined as the heat evolved at a certain time during cure (ΔH, the area under an isothermal DSC curve up to time t) divided by the total heat of reaction (HR) from a subsequent dynamic scan:

equation image((1))

It is assumed that the rate of heat released during cure is directly proportional to the rate of reaction:

equation image((2))

The reaction rate (dα/dt) is a function of time and temperature (or α) and can also be described by different empirical models. Considering the cure of epoxies, the most commonly used expression for reaction kinetics is the autocatalytic model by Kamal:8

equation image((3))

where m and n are the reaction orders, and k1 and k2 are temperature-dependent rate constants that correspond to the catalysis by hydroxy groups initially present in the epoxy resin and the catalysis by hydroxy groups formed in the curing reaction, respectively. In two part epoxy/amine systems this model mostly fits the experimental data very well.

A differential scanning calorimeter (TA Instruments DSC 2920) was used to monitor the heat flow during dynamic and isothermal scans. Samples of 10–20 mg weight were sealed in hermetic pans and heated from room temperature to 350 °C at a rate of 2 °C · min−1 for the dynamic scanning tests. Since the Araldite F/piperidine mixture is very slow to react, higher heating rates failed to capture all the heat released from the reaction. Isothermal experiments were conducted at 120 °C for 16 to 20 h. In addition, dynamic mechanical analysis (DMA) experiments were conducted on a TA Instruments DMA 2980. Viscoelastic material parameters such as mechanical loss factor and storage modulus (tan δ and E′, respectively) were determined in single cantilever mode starting at ambient temperature and increasing to 140 °C at a heating rate of 2 °C · min−1 and a frequency of 1 Hz. Furthermore, the dimension change of a cubic sample (z = 5 mm) was recorded under a static compression load of 0.01 N from ambient temperature to 200 °C. The subsequent calculation of the linear coefficient of thermal expansion β was based on the following definition:

equation image((4))

where l means length, P means pressure, and T means temperature.

Results and Discussion

Cure Kinetics

The DEGBA/piperidine system is known for the ill-defined nature of its chemistry of cure. The network is generated by the chainwise anionic polymerization of epoxy groups. In spite of its high crosslink density, the flexibility of polyether chains leads to a low-Tg epoxy network.9 Figure 1 shows both the conversion as a function of time and the reaction rate during the cure. It can be seen that the neat epoxy (AF) reaches the maximum conversion of 0.85 after 14 h of curing, whereupon it reaches a plateau value. Within the first hour, the absolute conversion shows no significant changes, but after 8 h of cure the nanosilica-filled epoxies show higher conversion than the neat system. With increasing content of particles the absolute conversion is enhanced proportionally, as shown in Figure 1 and Table 1. The maximum conversion of AF5 is reached after 19 h. This suggests the formation of a more highly crosslinked system since more epoxide groups have been able to react. On closer examination of the conversion rate (Figure 1b), a dramatic change in the reaction behaviour is noticeable as soon as the particles are added. The graph illustrates the reaction rate of the neat AF, which can be perfectly described by the Kamal-Sourour model (Equation (3) with k1 = 0.00358, k2 = 0.00363, m = 0.57647, and n = 1.01765), as shown by the solid line. In contrast, the model is not able to describe the curing behaviour of the nanosilica-filled systems. At low conversion, AF1, AF3, and AF5 show a sharp peak in the conversion rate, which indicates an intensive reaction at that stage of cure. Furthermore, this indicates a more complex mechanism for autocatalysis than a bi- or termolecular one. Barton10 has suggested that ether-forming reactions such as homopolymerization or epoxy/hydroxy reactions are not accounted for in kinetic models. Etherification has been shown not to be a problem at low conversions, usually becoming dominant near and beyond vitrification.11 However, for epoxy-rich systems, the etherification reaction might dominate the curing reaction after most of the secondary amines are consumed during the initial curing stage.12 During the later reaction stage, the conversion rate of the nanoparticle-filled systems becomes lower than that of the neat system, showing almost stepwise behaviour for the higher filled epoxies AF3 and AF5. This behaviour is similar to that of an epoxy-rich system. Furthermore, Table 1 illustrates the calculated exothermic heat per mol of epoxide, which decreases with an increasing volume fraction of particles. This normally indicates that a lower content of hardener inhibits the completion of epoxy group reactions, which results in lower exothermic heat per mol of epoxide. The neat AF cannot reach a higher conversion than 0.85, a phenomenon that is related to the dominant etherification. The lack of a further increase in secondary amine conversion shows that although there are still many remaining reactive secondary amine sites, they are unable to react with epoxide groups because of already being immobilized in the rigid network.13 Since the oxygen atom present in the polyether chains of DEGBA-piperidine acts as a hinge in the network structure,9 it could facilitate the formation of an amino-rich region around the silica particles. After the reaction of the silica with the amine, the reaction rate slows down and can gradually provide more amine at the interphase towards the end of the curing reaction. The TEM image (Figure 2) substantiates the formation of an altered interphase, which is depicted by the brighter region around two neighbouring nanoparticles. This is in good agreement with results from Olmos et al.,14 where a concentration of amino groups was detected at the silica/epoxy interface using FT-IR spectroscopy. However, direct testing methods on our own materials are currently being carried out in order to provide more evidence of the amine accumulation at the interface.

Figure 1.

a) Conversion and b) conversion rate diagrams obtained from DSC measurements.

Table 1. Properties of neat (AF) and nanoparticle reinforced matrix (AF1, AF3, and AF5).
MatrixDensityYoung's modulus (23 °C)CTE (40–80 °C)TgConversionMolecular WeightExothermic heat (of epoxide)
g · cm−3MPa10−6 mm · K−1°Cg · mol−1kJ · mol−1
AF1.171.172 50048.494.50.8584869.4
AF11.191.182 7542 67532.747.996.30.9371667.8
AF31.21.22 8802 90230.14795.80.9673566.4
AF51.221.222 9953 08224.94694.90.9965560.8
Figure 2.

TEM image of two neighboring silica nanoparticles in the epoxy matrix showing a 2–3 nm altered interphase region.

Thermomechanical Analyses

The temperature-dependent storage modulus of the tested material systems is illustrated in Figure 1a. It is evident that the addition of silica nanoparticles leads to an enhancement in stiffness as expected. Table 1 compares the values at room temperature obtained by DMA with the calculated values from the Counto Model15 for the traditional method of stiff and rigid particle reinforcement. The values are in good agreement and, therefore, the same mechanisms of reinforcement can be assumed for the nanoparticles. However, the AF5 shows a post-cure effect when reaching the Tg. The Tg values do not vary greatly, as can be seen from Table 1. Nevertheless, the AF5 still seems to have the potential of increasing the crosslink density. At this stage, we can only assume that for a higher particle content the network structure still allows nodular epoxy formation and, in general, definitely needs a longer curing time. Applying the theory of rubber elasticity, the mean molecular mass between the crosslinks can be obtained:

equation image((5))

Where E is the dynamic storage Young's modulus, q is the front factor (usually equal to 1), n is the apparent crosslink density, R is the universal gas constant (8.314 J · (mol−1 · K−1), d is the density of the material at T, and T is the absolute temperature at a selected point of the rubbery plateau (Tg + 30 °C). For the estimation of MC, E′ values are read from the DMA spectra in the rubbery state at Tg + 30 °C.16 This result substantiates the suggested alteration in the microstructure, namely that a lower molecular weight indicates a higher crosslink density. The MC values may be used for the sake of comparison only, but are also in good agreement with previously published results from others.17 Although more highly crosslinked epoxies normally correspond with a more rigid structure, some others18 have also reported greater deformation capacity in more highly crosslinked epoxy systems because of resin/hardener ratio shifts. In our case, the microstructure formed with the presence of nanoparticles could be composed of nodules with a lower crosslink density in the interphase region (locally) embedded in a more highly crosslinked matrix as a result of the consequent global lack of hardener. Furthermore, Table 1 shows the measured and calculated density of the various systems, which are almost identical, which indicates no changes in free volume. Various research groups19–21 have reported a decrease in Tg when silica nanoparticles are added to epoxy resins. This behaviour has been attributed to a plasticizing effect or to a limited amount of unreacted epoxy. The present study shows that this effect is clearly attributable to the shifting of the epoxide/amine group reaction and can only be controlled by individual stoichiometric considerations. The influence of the amino-rich interphase built around the particles can also be captured when the coefficient of thermal expansion is measured by DMA and calculated by the linear rule of mixture. Comparison of both values for each material (Table 1) shows a large discrepancy. The measured values (averaging between 40 and 80 °C) are significantly lower than the calculated values. In Figure 3b, the linear thermal expansion behaviour of a cubic sample is illustrated. The neat AF and the different mixtures show two levels, before and after Tg. The AF reaches a stable plateau value after Tg, whereas AF1, AF3, and AF5 show a further increase after passing Tg. This could indicate a slight post-cure effect, which in the case of AF5 can take place only at the interphase region since a conversion of 0.99 is measured by DSC. The discrepancy, however, must be explained by a completely altered microstructure and not only by the presence of the nanoparticles.

Figure 3.

a) Plot of temperature-dependent storage modulus and b) coefficient of thermal expansion β (both from DMA measurements).


The cure kinetics of silica-nanoparticle-reinforced epoxies is successfully studied by means of differential scanning calorimetry. The addition of nanoparticles in the range of 1–5 vol.-% into the neat epoxy is observed to lead to a higher reactivity in curing and, thus, an alteration in the crosslinking. Furthermore, TEM pictures and additional results from thermomechanical analyses suggest the formation of an amino-rich interphase region around the particles. The epoxy network in these composites shows higher crosslinking with an increase in content of nanoparticles, because of the enhanced homopolymerization of epoxide groups. On the other hand, a softer, less crosslinked region is found to control the nanophase around the nanoparticles. The change in the microstructure triggered by a clearly changed curing behaviour contributes to the physical properties of nanoparticle reinforced epoxies.


The authors express their thanks to the Alexander von Humboldt foundation, Bonn, Germany, for the Feodor Lynen fellowship for Patrick Rosso, which enabled him to conduct the research work described at the University of Sydney. Further, the authors gratefully acknowledge the support of nanoresins AG (Germany) for providing the material samples.