A commercial available vinylester resin (VE; Derakane 411 - 350 Momentum; Ashland) was used as a thermoset matrix and Metil Etil Ketone Peroxide (MEKP; Médano) was used as catalyst. Polycaprolactone (PCL), Mn = 42.500 gr/mol, was obtained from Sigma Aldrich. All materials were used as received.

The thermoset resin was heated to 100˚C with constant stirring (600 - 720 rpm). After that, the PCL was slowly added in different weight ratios varying from 0 to 36 wt%, that was the maximum amount that could be possible dissolved without the formation of aggregates. After 30 minutes, total dissolution of the PCL was achieved. The temperature was reduced to 60˚C and the catalyst was quickly added to the mixture in a 1% wt referred to the resin. The final catalyst weight ratios in the blends were 1%, 0.9%, 0.8%, 0.76%, 0.72%, 0.68% and 0.64%. Finally, plaques of 6 mm thickness were obtained in an aluminum mould.

Samples were cured in an oven. The curing cycle consisted in 1 hour at 50˚C in order to eliminate bubbles, followed by 2 hours at 110˚C as a poscuring cycle.

Blends of increasing PCL wt% were named as VE for 100%VE and respectively until VE64 (64%VE + 36% PCL).

Fracture characterization was carried out in three-pointbending configuration in an Instron dynamometer 4467 at a crosshead speed of 1 mm/min in accordance with ASTM 5045-93 standard recommendations. Critical stress intensity factor (K_IC) values and energy release rate (G_IC) values were obtained from these tests. Single-edge notched bend SENB specimens were cut out from thick plaques (B = 5 mm). Sharp notches were introduced by sliding a fresh razor blade into a machined slot. Crack-to-depth (a/W), thickness-to-depth (B/W) and span-todepth (S/W) ratios were kept equal to 0.5, 0.5 and 4, respectively.

Fracture surfaces of specimens broken in fracture tests were also analyzed by scanning electron microscopy (SEM, model JEOL JSM 6460 LV) after they had been coated with a thin layer of gold.

Differential scanning calorimetry (DSC) analysis was performed for the different materials in a Perkin Elmer Pyris 1DSC analyzer. Samples were heated from –65˚C up to 200˚C at a rate of 10/min. This study was carried out in order to verify the possible existence of phase separation.

Viscoelastic properties were measured using a dynamic mechanical analyzer Perkin Elmer DMA 7e with three-point-bending geometry. Tests were performed in temperature scan mode from 20˚C to 180˚C at a fixed frequency of 1 Hz and a heating rate of 10˚C/min. Rectangular specimens having a size of 20 mm × 3.5 mm × 2 mm were used for the evaluation of storage modulus (E’), loss modulus (E”) and damping parameter (tan δ). The static stress was 500 MPa and dynamic stress 300 MPa.

Flexural tests were done in accordance with ASTMD 790-03 standard over samples of 10 mm × 5 mm × 100 mm in an Instron dynamometer 4467 at a crosshead speed of 1 mm/min, and the span used was 80 mm. Flexural modulus was determined from the initial slope load-displacement plot.

Izod tests were also conducted on notched specimens according to ASTM D256-04 at 1 m/s in a falling weight Fractovis of CEAST. Impact strength values were obtained from numerical integration of the load-displacement records and they were normalized to the unnotched ligament area.

Figure 1 shows the damping factor (tan d) as a function of temperature for blends with different compositions. It is clear that the tan d peak was shifted to lower temperatures when PCL content was increased. From this peak, T_g was determined. Figure 2 summarizes the glass transition temperature, T_g, of each system as a function of PCL content. As it can be observed form the previous figure, the T_g of the vinylester resin decreased as a function of thermoplastic content. This result is in accordance with other thermoset-thermoplastic bleds reported in the literature [12].

There are several models [13,14] available in literature to correlate the T_g of blends. One of them is Fox equation [13] that, for a binary system 1 + 2, can be represented by:

where T_g is referred to the glass transition temperature of the blend, T_g,i to pure component i, and x_i is the weight fraction of one component. Fox equation does not take in account the interaction between the blend components and therefore it did not fit the experimental data obtained.

Gordon-Taylor (GT) equation is useful to predict the

T_g of a blend based on the interaction between the components [14]. The glass transition temperature of the blend can be written by the GT equation as follows:

where T_g is the glass transition temperature of the blend, T_g1 and T_g2, are the glass transition temperatures of the components 1 and 2, respectively, W is the weight fraction of the components and k is a constant. The curve in Figure 2 was drawn using Gordon-Taylor equation obtaining a k value of 0.65 and showing a good correlation with experimental data. Several authors [15-17] have suggested that k can be taken as a fitting parameter that measures the strength of the interaction between the components of the blend. Belorgey, and Prud’homme. [15], modelled the change in T_g with composition of poly (e-caprolactone) with chlorinated polyethylene, poly (vinyl chloride) (PVC) and chlorinated PVC. They found values of k varying from 0.26 to 1.0. When this approach is used, a k value of 0.26 suggests a weak interaction whereas a k value of 1.0 suggests an important interaction.

In our case, the k value of 0.65 suggests intermediate interaction. This fact along with the transparency observed in all systems suggests that no phase separation occurred.

The morphology of the blends was studied by several methods. DSC analysis showed only one T_g peak, and no melting peak of PCL was observed, suggesting that only one phase is present in the blend. Furthermore, fracture surfaces of samples broken in SENB tests were treated with chloroform during 15 minutes and observed with an optical microscope and SEM. Chloroform dissolves PCL and in case of phase separation it reveals the surface microstructure. In both cases no evidence of phase separation was found.

Figure 3 shows the storage modulus E’ values as a function of temperature for the different materials. As it can be observed in this figure, a sharp drop of E’ in a small range of temperature was displayed for all systems, indicating the presence of the T_g. As the amount of second phase was increased, the drop became less abrupt. A marked decrease in the storage modulus as function of the polycaprolactone in the blend was also noted.

Experimental values of flexural modulus as a function of polycaprolactone content are presented in Figure 4.

As it may be clearly observed, increasing polycaprolactone presence significantly affects the flexural properties of the vinylester matrix. These results are in accordance with previous results [17].

The reduction of flexural modulus exhibited by specimens as function of thermoplastic content is in accordance with the results of storage modulus presented in Figure 3.

Critical stress intensity factor (K_IC) and energy release rate (G_IC) values of VE-PCL blends are presented in Figures 5 and 6 respectively along with their deviations. In these figures, it can be observed that modifying the VE with PCL results in an increase in fracture toughness. K_IC monotonically increased with polycaprolactone content until 28 wt.% whereas a maximum in (G_IC) was found around 32 wt.% of polycaprolactone added. It should be also noted that all composites exhibited fracture toughness values significantly higher than that of the vinylester matrix, reaching an increase in K_IC as high as 240% for the blend with 28 wt.% PCL. This result is in accordance with the literature. Yang et al. [12] have reported an important increase in K_IC with the incorporation of a thermoplastic copolymer. Nimura et al. [18] also reported an increase of 2.5 times on K_IC of an epoxy resin modified with a thermoplastic polymer.

Scanning electron microscopy of the fracture surfaces of specimens broken in fracture tests show that the pure vinylester matrix (VE) presents a flat fracture surfaces as seen in Figures 7. The only remarkable features are low ridges and shallow grooves. The observed morphology is indicative of a brittle fracture [19] and it is typical of thermoset systems [6]. For the 90 wt.% VE blend (VE90; Figure 7(b)) the fracture surface was quite similar, indicating just little plastic deformation in accordance with the low increase on the fracture energy found [18].

As the PCL content increased, a change in the failure mode was observed, switching to stable crack propagation and a plastic failure mode was found. Plastic deformation features can be observed for VE80 (Figure 7(c)) and VE64 (Figure 7(d)) systems. Peaks and fibril structures were clear, stress whitening also indicates new energy absorption mechanisms present in the blends. Finally the branced fracture paths results in high energy absorption [19].

branced fracture paths results in high energy absorption [19].

Figure 8 shows the experimental values of impact energy as a function of thermoplastic content. As it may be clearly observed in this figure, the addition of PCL significantly affects the impact behaviour of vinylester matrix [20].

In agreement with the fracture patterns observed in SEM, impact properties were improved by the second phase addition, mainly due to the presence of energy absorption mechanisms such as branced fracture paths and plastic deformation.